LinkedIn Hiring Assistant 2 (2026): Features, Cost, Limits

What changed in LinkedIn's AI recruiter, what it really costs to get it, and the limits no launch keynote mentions.

Over 20,000 companies now use LinkedIn's agentic hiring products, and on September 29, 2026 LinkedIn announced the version they will all be moved onto next - LinkedIn Talent Solutions. Hiring Assistant 2 adds memory of how you hire, personalized fit signals, identity and employment verification, voice pre-screening and the ability to evaluate candidates sitting in your applicant tracking system. It rolls out automatically from November 2026, at no extra cost, to every existing Hiring Assistant customer working in English.

That is the good news, and it is genuinely good. The problem is everything the announcement leaves out. Version 2 is free only if you already pay for version 1, and version 1 is a negotiated add-on to an enterprise LinkedIn Recruiter contract that LinkedIn does not price publicly. The agent still finds only LinkedIn members and people already in your own applicant tracking system. Its memory makes it smarter every month you use it, and harder to leave every month too. And almost every performance number attached to it is self-reported.

This guide gives you the full picture in the order you need it. It starts with a plain-English verdict on what is new, then walks through every new feature in detail, how the agent works day to day, what it actually costs (including LinkedIn's own published government rate card), the hard limits, the compliance risks, where it succeeds and fails, how it compares to the alternatives, and a practical pilot plan. If you only read the first section, you will still know whether Hiring Assistant 2 deserves a place in your 2027 budget.

Written by Yuma Heymans (@yumahey), founder of HeroHunt.ai, who has been building AI sourcing agents since 2021 and now competes with LinkedIn's agent for the same recruiter budgets, which is exactly why he reads its release notes for what did not change as closely as for what did.

Contents

1. Hiring Assistant 2 at a Glance: What Changed and What Did Not
2. From Charter Pilot to Version 2: How LinkedIn Got Here
3. Every New Feature in Hiring Assistant 2, Explained
4. How Hiring Assistant 2 Works Day to Day
5. What Hiring Assistant 2 Costs in 2026
6. The Limits: What Hiring Assistant 2 Still Cannot Do
7. Compliance, Fairness and Data Risks
8. Where Hiring Assistant 2 Works Best, and Where It Fails
9. Hiring Assistant 2 vs the Alternatives
10. How to Evaluate, Pilot and Roll Out Hiring Assistant 2
11. The Future: Where Recruiting Agents Go After Version 2
12. The Verdict: Should You Buy Hiring Assistant 2?

1. Hiring Assistant 2 at a Glance: What Changed and What Did Not

Hiring Assistant 2 is a real upgrade to the agent, not to the deal. The agent gets meaningfully smarter: it remembers what you have hired for before, explains why each candidate fits, flags who is likely to respond, shows verified identity and employment data, runs voice or text pre-screens, and can now evaluate candidates who live in your applicant tracking system rather than only on LinkedIn. The deal stays exactly where it was: you still need an enterprise Recruiter contract plus the Hiring Assistant add-on, the talent pool is still LinkedIn's, and the price is still negotiated behind closed doors.

That distinction matters because most of the coverage of the launch focuses on features, and features are not the hard part of the buying decision. For a team already paying for Hiring Assistant, version 2 is close to a free lunch: it arrives automatically in November 2026 and costs nothing extra - LinkedIn Talent Solutions. For a team that does not have it yet, version 2 changes the quality of what you would be buying, but not the price of entry, the contract structure, or the walled garden. Those teams should evaluate it as a five-figure commitment, not as a feature release.

LinkedIn's own launch film is short and worth two minutes of your time, because it shows the new candidate cards, the memory-driven search and the pre-screening flow in the actual product interface.

The video is marketing, so it shows the happy path, but it is useful for one thing text cannot convey: how much of the recruiter's screen the agent now occupies. The candidate card has become the agent's explanation surface, with fit, interest and verification signals stacked next to the profile. That design choice tells you where LinkedIn thinks the recruiter's job is going: less searching, more reviewing the agent's reasoning and deciding.

The short version: version 1 vs version 2

The fastest way to understand the release is to compare it feature by feature with the version most customers have been using since general availability in 2025. The table below separates what is genuinely new from what is improved and what is unchanged. Everything in the "new" rows comes from LinkedIn's announcement; everything in the "unchanged" rows is a structural fact about how the product is sold and what it can see.

The comparison is intentionally blunt. A launch announcement naturally emphasizes the left-to-right improvement in each row, while a buyer needs to see which rows did not move at all, because those are the constraints you will live with for the life of the contract.

Capability	Hiring Assistant (v1, 2025)	Hiring Assistant 2 (Nov 2026)
Search and sourcing	Turns a job brief into searches on LinkedIn	Reasons about intent, "searches based on what you mean"
Memory	"Experiential memory" of recruiter preferences (since 2024)	Persistent recruiter- and company-level memory of your hiring history
Candidate explanation	Why a candidate fits, qualifications met	Fit signals plus a separate candidate interest indicator
Trust signals	Profile data	Verified identity and employer, connected apps, pre-screen answers
Pre-screening	Text pre-screening for job applicants	Voice or text, extendable to sourced candidates, median six minutes
Talent pool	LinkedIn members and LinkedIn applicants	Adds candidates in your connected ATS (Greenhouse, Lever, SmartRecruiters, Zoho Recruit, Tracker; iCIMS and Workday coming)
Price	Negotiated add-on to Recruiter	No extra charge for existing customers
Who gets it	Recruiter Corporate and RPS+ customers, 7 languages	Existing users working in English first, other languages vary by feature
Talent visible	LinkedIn members	Still LinkedIn members, plus people already in your ATS

Three rows deserve attention. The memory row is the biggest functional change and also the biggest strategic one, because it raises both the value of staying and the cost of leaving. The ATS talent pool row is the first real crack in the walled garden: for the first time the agent can evaluate people LinkedIn did not source, but only people you already have, not people you have never met. And the price row is generous for existing customers and irrelevant for everyone else, because the entry ticket is unchanged. The rest of this guide unpacks each of those rows in detail, starting with how the product got here.

2. From Charter Pilot to Version 2: How LinkedIn Got Here

Hiring Assistant went from experiment to business line in under two years, and that speed explains both the ambition of version 2 and its rough edges. LinkedIn first showed the agent in October 2024 as a charter program for a handful of large customers, made it generally available in English in September 2025, and was reporting it as a meaningful revenue line by spring 2026. Each step changed what the product does, and the numbers LinkedIn used to sell it.

Knowing the history is practical, not academic. Many features marketed as new in version 2 (memory, hiring-manager feedback in Microsoft Teams, ATS-connected projects) existed in earlier forms, and several of the headline metrics have been re-cut between releases. If you are evaluating version 2, you are really evaluating the fourth or fifth iteration of the same agent, and the trajectory tells you more than any single launch claim.

October 2024: the charter launch

LinkedIn introduced Hiring Assistant on October 29, 2024 as its first AI agent, built to take on the most repetitive parts of recruiting: turning intake notes into a job description and search, sourcing candidates and drafting outreach - TechCrunch. The charter customers named at launch included AMD, Canva, Siemens and Zurich Insurance. Even then, LinkedIn's engineers described an "experiential memory" that learns a recruiter's preferences and an orchestration layer that coordinates the agent's steps - LinkedIn Engineering.

That detail matters for reading the version 2 launch correctly. Memory is not a brand-new concept in Hiring Assistant; it has been part of the design since day one. What version 2 changes is the scope and persistence of that memory, moving from preferences inside a project toward a lasting model of how a specific recruiter and company hire.

September 2025: general availability

After roughly a year in charter, LinkedIn announced on September 3, 2025 that Hiring Assistant would be globally available in English by the end of that month, citing early results of more than four hours saved per role, 62% fewer profiles reviewed and 69% better InMail acceptance, with named users including Microsoft, Expedia, Siemens and Wipro - LinkedIn Newsroom. Behind the scenes the agent was rebuilt around a planner that breaks a recruiting task into steps and hands them to specialized sub-agents, with one agent instance per recruiter.

The diagram LinkedIn's engineering team published at the time is still the clearest picture of how the system is put together, and version 2 builds on the same skeleton.

The important element in that architecture is the separation between planning and execution: a supervisor decides what needs doing, and sub-agents for intake, sourcing, evaluation and outreach carry it out with their own tools - LinkedIn Engineering. Version 2's "better reasoning" is largely an upgrade to the planning layer, which is why LinkedIn describes it as understanding what you mean rather than following what you typed.

Late 2025 to mid 2026: features, languages and revenue

Between October 2025 and March 2026, LinkedIn shipped a "Wave 1" release that added AI-generated applicant targeting, AI-assisted follow-ups, a Microsoft Teams integration and ATS Connected Projects - LinkedIn Talent Solutions. French and German support followed in June 2026, and by the time version 2 was announced the agent worked in seven languages. The commercial results came quickly: on Microsoft's April 2026 earnings call, LinkedIn reported that its agentic hiring products had surpassed a $450 million annual revenue run-rate - LinkedIn Newsroom.

Two caveats keep that number honest. The run-rate covers LinkedIn's agentic hiring products as a group, which includes Hiring Pro, the small-business hiring agent, not Hiring Assistant alone. And it measures revenue, not results: it proves companies are paying, not that the agent outperforms a good human sourcer. By July 2026, recruiters at more than 20,000 companies were using LinkedIn's AI-powered hiring solutions, with seats across the enterprise products up 140% quarter over quarter - LinkedIn Newsroom.

How the headline metrics have moved

The metric LinkedIn uses most consistently is the reduction in profiles a recruiter must review to find a qualified match, and it is a useful case study in how vendor numbers evolve. The chart below plots the figure each time LinkedIn published it, from the July 2025 earnings update through the version 2 press release.

The upward trend is plausible, because the agent's models have genuinely improved and LinkedIn credits specific intake, retrieval, ranking and evaluation updates for recent gains. But every point on that line is LinkedIn's own measurement, against LinkedIn's own baseline, and other metrics have moved in the opposite direction: the time saved per role went from "more than four hours" in 2025 to "an average of 1.5 hours" in 2026, partly because the measure itself changed. The lesson for buyers is not that the numbers are wrong. It is that they are directional marketing evidence, and your own pilot data should carry more weight than any of them.

September 2026: Hiring Assistant 2

LinkedIn announced version 2 at its Talent Connect conference in New York, publishing the details on September 29, 2026, and LinkedIn's VP of Product Dan Reid unveiled it in a keynote - LinkedIn Talent Connect. Dan Shapero, who became LinkedIn's CEO in April 2026, was candid about version 1's reception, telling ERE that after the team shipped a new model and iterated for four weeks, "the expletives in the product from recruiters have gone down" - ERE. That is an unusually honest admission, and it matches what many early users reported: version 1 was useful but uneven, and customers mostly wanted it to "just be smarter."

Version 2 is LinkedIn's answer to that request. The next section breaks down every new capability in turn, what it does, and what it needs from you to work.

3. Every New Feature in Hiring Assistant 2, Explained

Hiring Assistant 2's features fall into four groups: smarter reasoning and memory, richer candidate signals, trust and verification, and agentic workflows that reach into your ATS and run pre-screens. Each group addresses a specific complaint recruiters had about version 1. The agent felt generic, so LinkedIn added memory. Recruiters did not trust its picks, so LinkedIn added explanations. Applicant fraud exploded, so LinkedIn added verification. And recruiters were still juggling two systems, so LinkedIn connected the agent to the ATS.

Understanding the groups separately helps you get more out of the product, because each one needs different input from you. Memory needs consistent feedback. Fit signals need a precise brief. Verification needs candidates to have verified themselves. ATS evaluation needs an integration that someone in your operations team has to switch on. A team that expects all four to "just work" out of the box will see a fraction of the improvement the launch promises.

Memory and reasoning: searching for what you mean

The headline change is memory. LinkedIn's own example makes the point well: previously a recruiter asked for a product management candidate and got one. Now, the agent finds one close to the office, in the retail industry and at mid-level seniority, because those are the things that recruiter has hired for in the past - Social Media Today. In LinkedIn's words, the agent now "searches based on what you mean" rather than only what you typed.

Better reasoning is the other half of the change. "It does what I meant to tell it. That's a big difference," LinkedIn's VP of Product Dan Reid said when he unveiled the release, adding that it understands "how you hire, not how everybody hires" - Recruiting News Network. The agent can also apply company rules, such as never recruiting from your own subsidiaries, and read intent behind non-keyword requests like "a fresh perspective". In practice that means fewer clarifying back-and-forths at intake and fewer irrelevant candidates in the first batch, especially for roles you have hired before.

Memory is genuinely useful, and it carries three practical implications that the launch does not spell out:

• It rewards consistency, so ratings and feedback need to follow the same standard across recruiters
• It inherits your history, including any skew in who you hired before
• It increases switching costs, because the learned preferences live in LinkedIn's product
• It needs overrides, so recruiters must state when a search should break from the past

The first two points are operational, and they are the difference between a memory that makes the agent sharper and one that makes it narrower. If three recruiters on the same team rate candidates by different standards, the agent learns an average of three standards, which is no standard at all. The last two points are strategic. A memory that encodes years of hiring decisions is an asset, but it is an asset you cannot export, and when you want to change your talent profile (a new market, a new seniority level, a more diverse slate), you have to tell the agent explicitly to ignore the patterns it was built to follow.

Fit signals and candidate interest

Version 2 replaces the generic match summary with two explicit indicators. Fit signals give what LinkedIn calls a holistic, personalized read on how well a candidate aligns to the role, split into strengths and points to consider. Candidate interest estimates how receptive someone is likely to be, based on signals such as platform engagement, Open to Work details and InMail responsiveness.

These indicators are the agent's answer to the trust problem. Recruiters will not act on recommendations they cannot explain to a hiring manager, and the new cards are designed to be explainable at a glance. LinkedIn says recruiters are now 4x more likely to contact a candidate sourced by Hiring Assistant than one found through traditional methods, and 27% more likely than just six months ago - LinkedIn Talent Solutions.

The screenshot shows the design logic clearly: a "Top match" label, a fit rating with a "Strengths" note and a "Consider" note, and a separate candidate-interest rating. Keeping fit and interest separate is the right call, because a perfect-fit candidate who never answers messages and a lukewarm-fit candidate who is actively looking require completely different outreach strategies.

The interest signal also deserves scrutiny. It is built from behavior on LinkedIn, which means people who are less active on the platform (often senior, busy or simply private) will tend to score lower regardless of how open they would be to the right offer. As one commenter on LinkedIn's own announcement put it, platform activity or InMail responsiveness should not become a proxy for ability or a quiet reason to screen someone out. Use interest to prioritize who to message first, never to decide who is worth messaging at all.

Trust and verification

The third group responds to a problem every recruiter now recognizes: fake and inflated candidates. LinkedIn cites research that 39% of recruiters say one of their top challenges is not knowing who is real or whether an applicant's skills are real, and version 2 puts verification front and center in response - LinkedIn Talent Solutions. Candidate cards can now show identity verification, verified employers, pre-screening responses and work history alongside the agent's assessment.

The scale behind this is significant. LinkedIn reports 115 million verified members, and verification badges will also appear inside partner applicant tracking systems including iCIMS, JazzHR, Jobvite, Lever, SmartRecruiters and Workable. That second point matters for teams whose recruiters live in the ATS rather than in LinkedIn Recruiter: the trust signal travels with the candidate into the system where decisions are actually recorded.

The screenshot shows two layers working together. At the top, identity is verified by government ID and the current employer is verified by work email. Below that, the agent's checklist marks each required and nice-to-have qualification as met, partially met or missing. Verification answers "is this person real?", while the checklist answers "does this person fit?", and recruiters need both before they spend a phone screen on someone.

Two caveats apply. Verification is voluntary for members, so an unverified profile is not evidence of fraud, and recruiters should not treat the absence of a badge as a red flag on its own. And verification confirms identity and employment, not skill: a verified senior engineer can still be a weak senior engineer. Treat the badges as a fraud filter that removes the worst risks quickly, not as a quality signal.

Connected apps: evidence beyond the profile

Connected apps let members attach verified evidence of their work from outside platforms to their LinkedIn profile. Version 2 expands the list with new partners including Chess.com, ElevenLabs and Medium, alongside existing portfolio integrations such as Behance and Replit. LinkedIn says members with a connected app receive up to 2.9x more recruiter messages than members without one.

For recruiters, connected apps are the closest LinkedIn gets to the open-web evidence that independent sourcing tools have relied on for years: a design portfolio, a coding environment, published writing. The difference is that the candidate must opt in by connecting the app, so the evidence covers only members who know about the feature and chose to use it. Expect it to be most useful in creative, content and early-career technical roles, and least useful for senior specialists who have no reason to curate their LinkedIn profile.

Voice and text pre-screening

Hiring Assistant 2 can now run pre-screening on its own, in text or by voice. A candidate receives an invitation to a short voice call with the agent (with an option to try a practice call first), answers the role's screening questions, and the responses feed back into the candidate card. LinkedIn says pre-screening is completed in a median of six minutes, compared with the roughly three and a half hours it typically takes a candidate to receive a recruiter's first response.

The short clip below, published by LinkedIn with the announcement, shows the candidate's side of that flow, which is the part most recruiters never see.

Voice pre-screening is the most consequential feature in the release from a compliance standpoint, and it should be configured with care. A pre-screen is the first point at which an automated system's evaluation of a candidate directly shapes whether a human ever looks at them. Use it to collect structured information (availability, work authorization, salary expectations, must-have experience) rather than to score soft qualities, keep a recruiter reviewing the outcomes, and tell candidates clearly that they are speaking with an AI. Section 7 covers the legal reasons in detail.

ATS integration: evaluating your whole talent pool

The final group extends the agent beyond LinkedIn's own pipeline. When you connect your applicant tracking system, Hiring Assistant 2 can evaluate your entire talent pool, including candidates who applied through your careers site or came from other sources, and changes made in LinkedIn sync back to the ATS automatically. The initial ATS partners are Greenhouse, Lever, SmartRecruiters, Zoho Recruit and Tracker, with iCIMS and Workday named as coming next.

This is the most underrated part of the release. Most companies already have tens or hundreds of thousands of past applicants sitting unused in their ATS, many of them silver-medal candidates who nearly got an offer. An agent that can rediscover them for a new role, with the same fit and verification signals it applies to LinkedIn members, attacks the cheapest source of hires most teams ignore.

Look closely at the evidence tags in that screenshot: each qualification is attributed to a source, such as the LinkedIn profile, an uploaded resume, or a connected app like HubSpot or ElevenLabs. That attribution is what makes the ATS integration valuable, because a recruiter can see whether a claim came from the candidate's own profile or from a document in your system. Customer quotes in the announcement point the same way: Porch Group's talent leader credits the Workday integration with letting recruiters evaluate top talent much faster.

The catch is coverage. If your ATS is not on the partner list, this feature does not exist for you yet. LinkedIn's partner directory still marks iCIMS and Workday, along with Avature, Teamtailor, Recruitee, Workable and several others, as "coming soon" for Hiring Assistant, which suggests the Porch Group deployment is early access rather than general availability - LinkedIn Talent Solutions. Check the current list against your own ATS before planning around this feature. Even with a supported ATS, the agent evaluates only people already in your system. It rediscovers candidates you have met; it does not find candidates nobody on your team has ever encountered.

4. How Hiring Assistant 2 Works Day to Day

In daily use, Hiring Assistant 2 behaves less like a search tool and more like a tireless junior sourcer who always asks permission. A recruiter opens a project, hands over the job description and intake notes, and the agent turns them into a set of qualifications, a search plan and a first batch of candidates. From there, the recruiter's job is to review, correct and approve, while the agent handles the volume work of finding, evaluating, messaging and pre-screening.

The workflow has not changed shape since version 1; what changes in version 2 is how much the agent already knows when it starts, and how much evidence it shows you when it recommends someone. A recruiter who has run fifty projects through the agent should find the first batch noticeably closer to the mark than a new user does, because the memory layer carries context from every previous search, rating and hiring decision.

The end-to-end flow

The diagram below shows a typical project from intake to interview, with the points where the recruiter stays in control. LinkedIn's own description of the full workflow covers sourcing and applicant evaluation, outreach, screening, hiring-manager feedback, scheduling and ATS workflows - LinkedIn Talent Solutions.

Two loops in that diagram explain why the agent improves over time. The first runs from the recruiter's ratings back to the qualifications: every thumbs-up and thumbs-down tells the agent what "good" means for this role. The second is invisible in a single project but matters most: those ratings, plus the eventual hire, feed the memory that shapes the next project. That is the mechanism behind LinkedIn's claim that version 2 understands how you hire "not how everybody hires."

Intake: the step that decides everything

The quality of a Hiring Assistant 2 project is decided in the first ten minutes. The agent turns your intake into up to 35 qualifications, splits them into required and preferred, and uses them both to search and to evaluate. A vague brief produces vague qualifications, and vague qualifications produce a shortlist that looks reasonable and is not. The new memory layer fills some gaps with your history, which helps, but it cannot read a hiring manager's mind.

The practical habits that make intake work are simple. Paste the full job description and the hiring manager's real notes, even if they are messy. State the deal-breakers explicitly. Name two or three companies or profiles that represent the target, and say plainly when the search should break from past patterns (a new market, a different seniority level). Then read the qualifications the agent drafts before approving them, because that list becomes the yardstick for every fit signal you will see later.

Sourcing, evaluation and outreach

Once the brief is approved, the agent sources from LinkedIn's member base and, if your ATS is connected through LinkedIn's Recruiter System Connect integration (RSC+), evaluates applicants and past candidates in your system against the same qualifications. Each candidate card then carries the fit signals, the candidate-interest indicator and any verification badges, so the recruiter can review a batch quickly and rate it. Hiring managers can weigh in through Microsoft Teams without needing a Hiring Assistant licence of their own, which removes one of the most common bottlenecks in any sourcing process - LinkedIn Help.

Outreach is where most teams set their autonomy boundary. The agent can draft, send and follow up on messages on the recruiter's behalf, in batches of up to 25 candidates; the messages are labelled as sent by the recruiter's Hiring Assistant, and each one counts against your InMail credits like any manual message. Many teams let the agent draft but keep a human approving each batch for the first weeks of use, then relax that control for roles where the messaging has proven itself. That approach costs a little speed and saves a lot of employer-brand risk.

Pre-screening, feedback and handoff

The final stage is where version 2 adds the most new capability. Interested candidates can be pre-screened by voice or text, with answers attached to their card, and LinkedIn's product page lists interview scheduling and ATS stage syncing as part of the version 2 workflow. Practically, the handoff to a human happens later in the funnel than it did with version 1: the recruiter's first live conversation is increasingly with a candidate who has already been sourced, evaluated, contacted and pre-screened by the agent.

That later handoff is the point of the product, and it is also where judgment matters most. A recruiter who inherits a pre-screened candidate should still review the evidence behind the fit signals, listen to or read the pre-screen, and decide independently whether the candidate deserves an interview. The agent's job is to make that decision faster and better informed, not to make it, and the teams that keep that distinction clear get the time savings without the compliance exposure covered in section 7.

5. What Hiring Assistant 2 Costs in 2026

The cost of Hiring Assistant 2 has two completely different answers depending on who you are. For existing Hiring Assistant customers, version 2 costs nothing extra: LinkedIn is rolling it out automatically and for free from November 2026 - LinkedIn Newsroom. For everyone else, the price of version 2 is the price of getting Hiring Assistant at all, which means a LinkedIn Recruiter contract (Recruiter Corporate, or Recruiter Professional Services Plus for staffing firms) plus the Hiring Assistant add-on on top, negotiated with LinkedIn's sales team.

LinkedIn does not publish either price on its website, and its product page routes every pricing question to sales. That opacity is deliberate, and it is not unusual for enterprise software, but it makes budgeting hard and benchmarks valuable. Fortunately there is one place where LinkedIn has published a real rate card, and it is more useful than any rumor: the price list LinkedIn files to sell to the UK public sector.

The one public rate card: UK G-Cloud

Suppliers selling through the UK government's G-Cloud framework must upload a pricing document, and LinkedIn's G-Cloud 14 document (uploaded in October 2025) lists both Recruiter seats and the Hiring Assistant add-on - UK Digital Marketplace. It is the only official LinkedIn price list for these products that anyone outside a sales process can read, and it is specific: prices per licence per year, in pounds, excluding VAT, by volume band.

The document prices a LinkedIn Recruiter seat at £8,925 a year for one or two licences, falling through five more bands to £6,350 for 251 or more, with at least 150 InMails per month included per licence. The Hiring Assistant add-on appears as a separate row under a launch promotion that ran from October 1, 2025 to June 30, 2026, priced from £2,079 per licence per year for one or two licences down to £1,575 for 101 to 250. During the promotion, buyers received "Tier 2" capacity (800 sourced candidates and 9,000 evaluated candidates) at the Tier 1 price.

The chart makes the structure of the bill obvious: the seat is the expensive part, and the agent is a comparatively small add-on on top of it. At the listed rates, a team of five pays about £42,900 a year for Recruiter seats and £10,000 for Hiring Assistant, so the agent adds roughly a quarter to the cost of the seats it rides on. Converted at recent exchange rates, the promotional add-on works out to very roughly $2,000 to $2,800 per licence per year, while a seat for a small team costs well over $10,000.

Three cautions apply before you treat these as your price. The promotion ended on June 30, 2026, and the document does not say what the add-on costs afterward, so a quote for the same capacity in late 2026 may be higher. UK public-sector rates are not US commercial rates, and they exclude VAT. And note one widely repeated misreading: the £6,350 figure in that document is the 251-plus band of the Recruiter seat ladder, not the price of Hiring Assistant.

What US buyers actually pay

US buyers rarely see a list price, so the best evidence comes from aggregated purchase data. Vendr's LinkedIn buyer guide, updated in February 2026, puts the median annual spend at about $38,500 across about 1,700 purchases, with real contracts ranging from roughly $8,400 to $166,000, and estimates Recruiter list prices at $8,000 to $12,000 per seat per year - Vendr. That median covers every LinkedIn product a company buys, not just Recruiter seats, so treat it as a sense of scale rather than a quote.

The same data shows how negotiable the number is. Vendr reports that discounts of 15% to 30% off list are typical, and that annual price increases of 3% to 7% are common at renewal. For Hiring Assistant specifically, no US price has ever been published: LinkedIn did not announce one at launch, and its product page still sends buyers to sales. Treat any precise "per seat per month" figure you find on third-party sites with suspicion; several widely copied numbers have no primary source, and one popular claim that the agent can be bought with Recruiter Lite is contradicted by LinkedIn's own plan comparison, which offers the add-on only with Recruiter and Recruiter Professional Services Plus - LinkedIn Help.

What the capacity tiers mean

The price of a Hiring Assistant licence is tied to its capacity tier, which caps how many candidates the agent may source and evaluate each month. LinkedIn's help center publishes the tiers, and they are the most useful planning numbers LinkedIn offers, because they translate a licence into workload - LinkedIn Help. A sourcing credit is used only when a candidate meets your qualifications, passes the agent's evaluation and is saved to your pipeline; evaluation credits are used as the agent assesses applicants. The limits apply per licence, per month, and are not pooled across users.

Two operating rules make the tier choice important. When a recruiter hits the cap, the agent simply stops sourcing or evaluating until the next month. And unused credits are forfeited at the end of each month, so buying a tier sized for your busiest month means paying for idle capacity in quiet ones.

Tier	Recruiter Corporate (sourced candidates / evaluated applicants per month)	RPS+ for agencies (sourced candidates / evaluated applicants per month)
Tier 1	200 / 3,500	150 / 4,500
Tier 2	800 / 9,000	600 / 15,000
Tier 3	1,500 / 25,000	1,500 / 50,000

Read against the UK rate card, the tiers explain the 2025-26 promotion: buyers paid the Tier 1 price and received Tier 2 capacity, four times the sourcing allowance. For a corporate recruiter carrying eight to ten open roles, Tier 1's 200 sourced candidates a month can run out quickly, while Tier 2's 800 is comfortable for most. Agencies get a different shape, with fewer sourced candidates and many more evaluations, reflecting their heavier applicant flow. Ask which tier your quote includes, what the next tier costs, and whether the promotional upgrade can be carried into your renewal.

The table below brings the components of a Hiring Assistant 2 budget together. Use it as a checklist for your own quote rather than as a price list, because almost every line is negotiable and several depend on contract terms LinkedIn does not publish.

Cost component	What it covers	Benchmark
Recruiter seat	Recruiter Corporate (or RPS+ for agencies), required per agent licence	£6,350 to £8,925 per seat per year (UK rate card)
Hiring Assistant add-on	The agent licence, one per Recruiter seat at most	£1,575 to £2,079 per licence per year (2025-26 promo)
Version 2 upgrade	Memory, fit signals, verification, voice pre-screening, ATS pool	No extra charge for existing customers
Capacity tier	Monthly sourced and evaluated candidates per licence	Tier 1 to Tier 3; unused credits forfeited monthly
InMail usage	Every agent message counts against InMail credits	150 InMails per Recruiter seat per month, pooled
ATS integration	RSC+ for Hiring Assistant	Requires a Hiring Assistant licence; supported ATS needed

Two lines in that table are where budgets go wrong. The first is the one-seat-per-agent rule: you cannot hold more Hiring Assistant licences than Recruiter licences, so giving the agent to a sourcer, coordinator or regional recruiter means buying them a full Recruiter seat first. The second is InMail consumption: because every message the agent sends, including each follow-up, draws on the same InMail pool your recruiters use, an agent that triples outreach volume can exhaust allowances that were sized for human sending. Credits come back when a candidate accepts, declines or replies within 90 days, so well-targeted outreach is far cheaper than spray-and-pray, and extra credits beyond the pool cost money.

Worked example: a five-recruiter team

Abstract price ranges are hard to plan with, so it helps to walk through one realistic case using the UK list rates, which are the only official numbers available. A corporate talent team with five recruiters, all on Recruiter, wants Hiring Assistant for every seat. At the 3-10 band, five Recruiter seats cost 5 x £8,575 = £42,875 a year, and five Hiring Assistant licences at the promotional 3-10 rate cost 5 x £2,000 = £10,000, for a total of £52,875 a year before VAT.

Now adjust for reality. If the post-promotion add-on is priced higher, or your volume needs Tier 2 or Tier 3 capacity at full price, the agent line grows. If the agent's outreach pushes you past your InMail allowance in busy months, you pay for extra credits. If your hiring managers need Recruiter seats to work in the agent's projects, the seat line grows too (though Teams-based feedback avoids that for simple approvals). On the other side of the ledger, enterprise buyers routinely negotiate meaningful discounts off list, especially on multi-year terms. The honest planning range for this team is therefore "around £50,000 to £60,000 a year, before negotiation," and the only way to narrow it is to get a written quote that answers the questions in section 6.

How to negotiate

The opacity that makes LinkedIn's pricing frustrating also makes it negotiable, and a few levers reliably move the number. Timing matters most: negotiate against your renewal date, not after it, because leverage disappears once auto-renewal kicks in. Volume bands matter next: the rate card shows prices stepping down at 3, 11, 31, 51 and 101 licences, so buying just over a band edge can lower the price of every licence. Multi-year terms unlock larger discounts but lock you in while the product, and the market, are changing fast.

The strongest lever is a credible alternative. A quote or a completed pilot with an open-web agent establishes your walk-away price, and it tells LinkedIn's sales team that the agent add-on has to earn its place. Ask for the capacity tier, the post-promotion add-on price, the price of the next tier up and InMail top-up rates in writing, and ask whether the promotional Tier 2 upgrade can be carried into your renewal. Teams that arrive with data and an alternative negotiate from strength; teams that have already told their recruiters the agent is coming negotiate from weakness.

6. The Limits: What Hiring Assistant 2 Still Cannot Do

Hiring Assistant 2's biggest limit is the same as version 1's: its talent pool is LinkedIn's. Every improvement in this release (better reasoning, memory, fit signals, verification) makes the agent smarter about LinkedIn members and about the applicants already in your ATS. None of it extends the agent's reach to people who are not on LinkedIn, who keep a thin or outdated profile, or whose best evidence of skill lives somewhere else entirely. For many commercial and corporate roles that is a small gap. For engineering, research, design, healthcare, skilled trades and most markets outside North America and Western Europe, it is the gap that decides whether the agent finds your next hire at all.

The second category of limits is commercial rather than technical. Version 2 is not a product you can buy on its own; it is an upgrade for teams that already pay for Hiring Assistant, which is itself an add-on to an enterprise Recruiter contract. The rollout is staged and language-bound, autonomy is deliberately capped, and several details that matter for budgeting (such as what each capacity tier costs once the launch promotion has ended) are not publicly documented. None of these are reasons to dismiss the product. They are reasons to read the fine print before you plan around it.

Limit 1: The walled garden

The agent's talent pool is LinkedIn's member base, and its evidence is what members choose to publish. LinkedIn's network is enormous, but self-reported profiles skew toward roles and regions where LinkedIn is culturally central, and they systematically under-represent people who do great work without curating a profile. Many senior engineers keep minimal LinkedIn pages and live on GitHub; researchers are visible through publications; designers through portfolios; frontline and trades workers often not at all.

LinkedIn has started to soften this. Since May 2026, "professional sources" bring trusted candidate data from GitHub, patents and similar places into Hiring Assistant's evaluations - LinkedIn Talent Solutions. Version 2 adds more connected apps and lets the agent evaluate people already sitting in your ATS. But these additions enrich the evaluation of candidates the agent can already reach; nothing LinkedIn has published says the agent can discover people who are neither LinkedIn members nor already in your own system. If your best hires historically came from referrals, communities or open-source contributors who never kept a LinkedIn profile, the agent will not reproduce that pipeline, and you will need a second channel for it.

Limit 2: Eligibility, language and rollout

Version 2 arrives first for existing Hiring Assistant customers working in English, starting in November 2026 - LinkedIn Newsroom. Version 1 already works in seven languages (Dutch, English, French, German, Italian, Portuguese and Spanish), but LinkedIn's product FAQ warns that version 2 language availability will vary by feature - LinkedIn Talent Solutions. That has three practical consequences. Teams that do not already have the Hiring Assistant add-on must buy version 1 access first, through a Recruiter Corporate or Recruiter Professional Services Plus contract, before version 2 means anything to them. Global teams recruiting in other languages should expect a staggered experience, with English-language requisitions getting the new memory, fit signals and voice pre-screening before the rest. And because rollout is staged, a feature demonstrated at Talent Connect may reach your account weeks or months after the announcement.

The licensing rule that carries over from version 1 matters too: you cannot hold more Hiring Assistant licences than Recruiter licences, because the agent is strictly an add-on to a seat. Teams that want to give the agent to a coordinator, a sourcer in another region, or a hiring manager therefore need to buy a full Recruiter seat for that person first. That single rule is often the line item that turns an attractive per-agent price into a much larger contract.

Limit 3: Autonomy is capped by design

LinkedIn positions Hiring Assistant as an assistant that keeps the recruiter in control, and version 2 keeps that stance. The agent can reason, plan, source, screen and coordinate, but key actions remain subject to recruiter review, and the recruiter stays accountable for decisions about candidates. That is the right design for compliance, and it is also a real limit on the time you will save. An agent that needs approval at each gate saves sourcing hours but still consumes review hours.

The documented operating caps reinforce this. Hiring Assistant can send initial and follow-up messages to up to 25 candidates at a time, a project can carry up to 35 qualifications, and each licence has monthly sourcing and evaluation credit limits on top of your normal InMail allowance. In the EU, LinkedIn goes further and disables instant outreach entirely for any project whose location includes an EU region, so European teams approve every message by hand - LinkedIn Help. Those caps are sensible guardrails against spam, but they also mean a recruiter running a high-volume search still works in batches, approving and releasing outreach in chunks rather than letting the agent run unattended.

The practical effect is that Hiring Assistant 2 is not a "set it and forget it" recruiter. Teams that expected to reduce headcount by switching on an agent will be disappointed; teams that expected to let each recruiter carry more requisitions at the same quality will be closer to the mark. Plan capacity on the second assumption, and measure review time explicitly during your pilot, because it is the cost that vendor case studies tend to leave out.

Limit 4: Undocumented operating details

The fourth limit is about transparency rather than capability. LinkedIn publishes a great deal about what Hiring Assistant can do and much less about what each tier costs, what the agent stores and what you can take with you. For a tool that will sit inside your hiring process for years and learn from every decision you make, those operating details are as important as the feature list, and they are exactly the details that tend to surface only after a contract is signed. LinkedIn's help center has started to fill some of these gaps (the capacity tiers and InMail rules covered in section 5 are now documented), but several practical questions still have no public answer.

Several details that matter for operations are not publicly documented and should be confirmed in writing before you sign:

• Tier pricing: what each capacity tier costs once the launch promotion has ended
• Pre-screen data: how voice recordings and answers are stored, and for how long
• Memory controls: what the agent remembers, and how to view or reset it
• Exports: whether fit assessments and agent notes can be exported for audits
• ATS write-back: which fields sync to your system of record, and how often

One question that used to belong on this list now has a documented answer, and it is the one most buyers get wrong: every message Hiring Assistant sends counts toward your InMail credits and toward your Hiring Assistant credit limits, exactly like manual outreach. An agent that sends more messages therefore burns the same allowance your human recruiters depend on, and because credits are only refunded when candidates respond, a poorly targeted campaign costs real money. Budget InMail for the agent explicitly, and watch the first month's consumption closely before you let it send without review.

The remaining questions are not edge cases either. Tier pricing decides whether a busy quarter triggers an expensive upgrade, since the agent simply stops working when a licence hits its monthly cap. Pre-screen data handling decides your exposure under privacy law. Memory controls decide whether you can correct a bad pattern the agent has learned or must live with it. Exports and ATS write-back decide whether the agent's work survives in your own systems if you leave LinkedIn, or only inside LinkedIn's. Ask for each answer in the order form or a written side letter, not just in a sales call, because the public record does not settle any of them.

Limit 5: Self-reported results

Every performance number attached to Hiring Assistant, including the new version 2 metrics, comes from LinkedIn. That does not make the numbers false, but it does make them marketing outputs rather than audited benchmarks, measured on LinkedIn's chosen baseline (usually "traditional sourcing" on LinkedIn) with LinkedIn's chosen customers. The metrics have also shifted between releases, as earlier figures were replaced by different ones measuring slightly different things.

The fix is in your control: measure the agent against your own baseline, on your own roles, using the pilot structure in section 10. Until independent audits exist, your own data is the only performance evidence that should drive a renewal decision of this size.

7. Compliance, Fairness and Data Risks

Hiring Assistant 2 moves LinkedIn's agent closer to the decisions hiring laws regulate. Version 1 mostly found and messaged people, which is lower-risk territory in most jurisdictions. Version 2 evaluates applicants, assigns fit and interest signals, and runs automated voice or text pre-screens, which are exactly the activities that AI-in-hiring rules in the EU, New York City, Colorado, Illinois and California were written to cover. LinkedIn states that Hiring Assistant does not screen out applicants or make hiring decisions on its own - LinkedIn AI Transparency. Under most of these rules, though, the employer using the tool carries much of the responsibility for how it is used.

None of this makes the product unusable; it makes configuration and documentation part of the purchase. LinkedIn's design (recruiter approval at key steps, explanations on every recommendation, the ability to switch individual actions off) supports compliant use, and LinkedIn already disables instant outreach entirely for projects whose location includes an EU region. But a design that supports compliance is not the same as compliance, and the gap is filled by the policies, notices and audits your team puts around the tool.

The rules that apply

The regulatory picture differs by jurisdiction, and it changed substantially during 2026, so many summaries written last year are out of date. Five regimes matter most for a typical Hiring Assistant 2 customer, and each attaches to a different part of the workflow. The common thread is that rules focus less on whether you use AI and more on whether you can show a human reviewed its output, whether candidates were told, and whether you checked the results for bias.

The EU AI Act classifies AI used to target job ads, filter applications and evaluate candidates as high-risk - AI Act Annex III. The EU's Digital Omnibus, adopted in July 2026, moved the start of those high-risk obligations from August 2, 2026 to December 2, 2027. When they apply, employers using such systems must ensure human oversight by competent people, inform workers' representatives and inform candidates. One prohibition already applies today: emotion recognition in the workplace has been banned since February 2025, so voice pre-screens must never be configured or used to infer a candidate's emotional state.

New York City's Local Law 144 requires an annual independent bias audit and at least ten business days' notice to candidates before an automated employment decision tool is used to substantially assist hiring decisions about NYC residents, with penalties from $500 per violation - NYC DCWP. Enforcement was weak in the law's first years, but a December 2025 state audit criticized that weakness and raised the risk for employers who assumed nobody was checking.

Colorado repealed and re-enacted its AI Act in May 2026 as an automated-decision law covering employment, operative from January 1, 2027 - Colorado General Assembly. Illinois amended its Human Rights Act from January 1, 2026 to prohibit AI use that discriminates (including through zip-code proxies) and to require notice to candidates. California's Civil Rights Council rules on automated decision systems took effect on October 1, 2025, require four years of record retention, and allow a vendor to be treated as the employer's agent - California Civil Rights Department.

Litigation is moving in parallel with regulation. In Mobley v. Workday, a federal court allowed an age-discrimination claim over AI screening to proceed as a nationwide collective action in May 2025, and a further class-certification motion is set for hearing in March 2027 - Holland & Knight. A separate 2026 class action accuses Eightfold of building hidden candidate scores. No lawsuit has targeted Hiring Assistant itself, but these cases establish that both employers and vendors can be pulled into disputes about automated screening.

The practical checklist for a Hiring Assistant 2 rollout follows directly from these rules:

• Notify candidates that AI is used in sourcing, evaluation and pre-screening
• Keep a human reviewer on every decision that advances or rejects a candidate
• Commission a bias audit if you hire in New York City, and keep the results
• Retain records of agent outputs and recruiter decisions for at least four years
• Disable emotion inference and avoid scoring voice pre-screens on tone or delivery

Most of that list costs very little if you do it before rollout and a great deal if you do it after a complaint. The notice and human-review items are largely a matter of process design and outreach templates. The bias audit and record-retention items need cooperation from LinkedIn, so ask during procurement what audit support, data exports and logs LinkedIn provides for Hiring Assistant evaluations and pre-screens. If the answers are vague, treat that as a risk factor in the buying decision, because you cannot audit what you cannot export.

Fairness: what memory and interest signals can get wrong

The fairness risk in version 2 is specific and predictable: memory learns from your past, and interest signals learn from platform behavior. If your past hires over-represent particular schools, employers or demographics, a system designed to find candidates "like the ones you've hired before" will tend to reproduce that pattern. If interest signals reward platform activity and InMail responsiveness, people who use LinkedIn less (often caregivers, some older workers and many senior professionals) may score lower without being any less open to a move.

These are not accusations against LinkedIn's models; LinkedIn says it runs fairness and bias reviews on each model, though it does not publish the results. They are structural properties of any system that learns from history and engagement, and independent researchers have found ranking skews in LinkedIn's older search products. An academic audit of Recruiter Lite search, published in November 2025, reported under-representation of minority groups in the top-ranked results - arXiv. That study did not test Hiring Assistant, but it shows why your own monitoring matters.

The mitigation is in your hands. Review fit signals for proxies you would not defend, use interest to order outreach rather than to exclude candidates, write explicit instructions when you want a search to break from past patterns, and compare the make-up of agent-sourced slates with your human-sourced baseline during the pilot. Document what you find, because that record is what a bias audit, a regulator or a court will ask for. Candidate trust is low enough already: in one 2025 survey, only 8% of job seekers said AI makes hiring fairer - Greenhouse.

Data: what you give the agent, and what LinkedIn keeps

The third risk area is data. LinkedIn's help center says Hiring Assistant uses OpenAI models through Microsoft Azure for intake and search alongside LinkedIn's own models, that LinkedIn data will not be used to train third-party AI models, and that the conversation inputs and outputs are treated as customer personal data under LinkedIn's data processing agreement - LinkedIn Help. It also says that information is retained beyond the session (your master administrator can request access, export or deletion) and that LinkedIn may evaluate conversations to improve the product, which is worth knowing before recruiters paste sensitive intake notes into the chat.

Version 2 raises the stakes because its memory accumulates a detailed model of how your company hires, and the ATS integration brings your applicant records into LinkedIn's evaluation flow. Before switching those on, confirm how that data is processed, where it is stored, how long the memory persists and what happens to it if you end the contract. For EU and UK candidates in particular, your data protection impact assessment should cover the agent's evaluations and the pre-screen recordings, not just the Recruiter seat.

It is also worth knowing how LinkedIn treats member data more broadly. Since November 3, 2025, LinkedIn has used member data from the EU, EEA, Switzerland, the UK, Canada and Hong Kong to train its content-generating AI models unless members opt out through a "Generative AI Model Training" setting, excluding private messages - LinkedIn Help. That policy concerns members rather than your company's recruiting data, but it shapes candidate trust in the platform, and candidates who object to AI processing may increasingly prefer a human conversation. Offering one, and saying so in your outreach, is both good practice and good recruiting.

8. Where Hiring Assistant 2 Works Best, and Where It Fails

Hiring Assistant 2 works best for repeatable, LinkedIn-native roles at companies that already live in LinkedIn Recruiter. That sentence sounds narrow, but it covers a large share of corporate hiring: sales, marketing, customer success, operations, finance, HR and most mid-level business roles in North America, the UK and Western Europe. In those roles, the candidates keep active profiles, respond to InMail, and look enough like each other that a memory of past hires is genuinely predictive. The agent's version 2 upgrades (memory, candidate interest signals, applicant evaluation and pre-screening) all compound in exactly that setting.

It struggles where any of those conditions break. Roles where the best people are not on LinkedIn, roles you hire once, roles where the brief changes during the search, and markets outside the English-language rollout all weaken the agent's advantage. It also struggles in organizations whose real bottleneck is not sourcing at all but hiring-manager responsiveness, interview capacity or offer competitiveness. An agent can fill a pipeline faster; it cannot make a hiring manager review it.

Where it shines

The clearest wins come from high-repeat corporate hiring. A company hiring account executives every month gives the agent a steady stream of feedback, a clear pattern of successful hires, and a candidate pool that is overwhelmingly on LinkedIn. Memory turns "find me an account executive" into "find me an account executive like the last six we hired and kept," which is the kind of context a human sourcer takes months to absorb and a new recruiter never fully gets.

The second win is applicant review at volume. With applications arriving at a rate of around 11,000 per minute across the platform, according to LinkedIn's own figures reported in 2025, many recruiters spend more time filtering inbound applicants than sourcing passive ones - eWeek. Version 2's applicant evaluation, which checks every applicant against the qualifications you set and now extends to applicants who came in outside LinkedIn, targets exactly that pain. Teams buried in inbound volume may get more value from the screening features than from the sourcing features the product is famous for.

The conditions under which the agent's design should pay off most share a few traits:

• Repeat roles with stable, well-understood requirements
• LinkedIn-active talent in commercial, business and corporate functions
• Engaged hiring managers who give fast, specific candidate feedback
• Existing Recruiter Corporate contracts, so the add-on is incremental
• English-language hiring in markets where LinkedIn is the default network

When most of those traits are present, the agent's economics look good because the base seat is already paid for and the learning loop runs fast. When only one or two are present, the same add-on price buys far less. That is why two companies can buy the identical product and reach opposite conclusions about it: the product did not change, the conditions did. Score your own hiring against these five traits before you trust any vendor case study, because the case studies are drawn from customers where the traits line up.

Where it fails

The failure modes are as predictable as the successes. The most common is reach failure: the agent cannot find people who are not there. For a machine learning researcher, a senior backend engineer who never updates LinkedIn, a nurse, an electrician or a candidate in a market where LinkedIn penetration is low, the agent searches an incomplete map. It will still return candidates, which is the dangerous part, because a confident shortlist drawn from a partial pool looks like a complete answer.

The second failure is pattern lock-in. Memory learns what you have hired before, which is excellent when your past hires were good and representative, and harmful when they were not. A company trying to change its talent profile (new markets, new seniority levels, more diverse slates, career switchers) is asking the agent to do the opposite of what memory optimizes for. In those searches, recruiters should write explicit instructions that override past patterns, and review fit signals closely, rather than letting the agent default to history.

The third failure is outreach fatigue. LinkedIn members already receive a heavy volume of recruiter messages, and an agent that makes outreach cheaper also makes it easier to send more of it. If every competitor is running the same agent against the same candidate pool, response rates for generic outreach will erode over time. The teams that keep high acceptance rates will be those who use the agent's receptiveness signals to message fewer, better-matched people, and who still personalize the message that actually goes out.

The warning signs are already visible in adjacent channels. Gem's analysis of 6.2 million recruiting email sequences found reply rates falling from 22.6% in 2024 to 16.9% in 2025 as automated outreach scaled up - Gem. LinkedIn polices the same risk on InMail directly: a Recruiter seat that sends 100 or more InMails in a 14-day period must keep its response rate at or above 13%, or bulk InMail is switched off for two weeks - LinkedIn Help. An agent that sends carelessly on your behalf can therefore take your recruiters' outreach offline, which is one more reason to keep message volume tied to genuine fit.

9. Hiring Assistant 2 vs the Alternatives

The alternatives to Hiring Assistant 2 split cleanly along one question: do you need candidates LinkedIn cannot see? If the answer is no, Hiring Assistant 2 is hard to beat for teams already on Recruiter Corporate, because nobody else has LinkedIn's engagement data, InMail channel and verification layer. If the answer is yes, or if the entry price is the problem, a different class of tool fits better: open-web sourcing agents that search far beyond one network, publish their prices and can be piloted in an afternoon.

The market has moved fast around LinkedIn's launch. In the same week as Talent Connect, Metaview raised $60 million and made its autonomous recruiter generally available, and Juicebox opened a London office; a few weeks earlier hireEZ relaunched its whole platform around agents. The table below compares the options most teams actually shortlist, using each vendor's own published pricing where it exists and buyer-reported contract data where it does not.

Tool	What it is	Talent pool	Published entry price
LinkedIn Hiring Assistant 2	Agent inside LinkedIn Recruiter	LinkedIn members plus your ATS	Not published; add-on to a Recruiter seat
Indeed Talent Scout	Sourcing agent inside Indeed	Indeed job-seeker profiles (US)	Requires Smart Sourcing, from $520/mo
Juicebox	Open-web search with always-on agents	800M+ profiles, 30+ sources	$99/seat/mo annual; agents $199/mo each
hireEZ	Agentic sourcing and engagement stack	1B+ open-web profiles plus your ATS	$494/mo paid annually upfront
SeekOut	AI search, outreach, ATS rediscovery	1B+ open-web profiles	$149/mo annual or $179 monthly
Gem	ATS plus CRM with sourcing agents	800M+ profiles plus your history	Startups from $130/mo; others quoted
Metaview fillmore	Autonomous recruiter run from Slack	Not disclosed	Pay on hire; fee not published
HeroHunt.ai	AI Recruiter: open-web sourcing and outreach	Up to 1B profiles across platforms	From $99/mo, unlimited positions

A table flattens important differences, so read it with three caveats. First, entry prices are not comparable totals: LinkedIn's figure must be added to a Recruiter seat, Indeed's agent needs a sourcing subscription, and several "per seat" prices climb once you add contact data or extra agents. Second, talent-pool sizes are vendor claims measured differently, and a smaller, fresher, better-matched pool can beat a larger stale one. Third, the right comparison is cost per qualified candidate for your roles, which only a pilot reveals. Each tool gets a closer look below.

Indeed Talent Scout: the other walled garden

Indeed's employer agent is the closest structural match to LinkedIn's: a conversational sourcing agent that works inside one platform's own candidate data. Indeed introduced Talent Scout in September 2025 alongside a job-seeker counterpart, Career Scout, describing a pool of more than 615 million job-seeker profiles - Indeed. Access runs through Indeed's Smart Sourcing subscription, where the Professional tier is listed at $520 a month or $4,992 a year - Indeed.

The two walled gardens complement each other more than they compete. LinkedIn's graph is strongest for professional, corporate and white-collar roles; Indeed's is strongest for hourly, frontline, healthcare and skilled-trade roles where candidates actively apply rather than curate profiles. A company hiring both kinds of roles at volume may reasonably run both platform agents, each on its home turf, rather than forcing one to do the other's job.

Juicebox: open-web search with always-on agents

Juicebox is the fastest-growing independent alternative and the simplest to try. Its plain-English search covers more than 800 million profiles from over 30 data sources, and its Agents add-on runs continuous background sourcing that can auto-shortlist or auto-email candidates - Juicebox. Pricing is published: Starter at $99 per seat per month billed yearly ($119 monthly), Growth at $179, and agents at $199 per agent per month.

Where Juicebox beats Hiring Assistant 2 is reach and access: it finds people who keep thin LinkedIn profiles, and a recruiter can start without a sales cycle. Where it loses is engagement data and channel. It cannot see who is Open to Work or who answers InMail, and its outreach depends on finding contact details, which consumes credits and works better in some markets than others. It raised an $80 million Series B at an $850 million valuation in March 2026, so the product is moving fast.

hireEZ: the agentic sourcing stack

hireEZ is the most complete open-web alternative for teams that want sourcing, engagement and scheduling in one place, running on top of their existing ATS. It searches more than a billion profiles and rebuilt its platform around agents in August 2026, under what it calls guided autonomy: the recruiter directs and approves while agents source, engage and schedule. The solo plan is listed at $494 a month - hireEZ, which is the annual plan paid upfront (about $5,929 a year); paying month by month on an annual term costs more.

For a mid-sized team, the comparison with Hiring Assistant 2 is less about features than about architecture. hireEZ lives next to your ATS and reaches across the open web; Hiring Assistant 2 lives inside LinkedIn and reaches into your ATS. Teams whose recruiters work mostly in the ATS and whose roles need open-web reach tend to prefer the former, while teams whose recruiters live in LinkedIn Recruiter tend to prefer the latter.

SeekOut: deep search and ATS rediscovery

SeekOut combines AI search over a billion-plus open-web profiles with multi-step outreach and rediscovery of candidates already in your ATS, and its MCP server lets AI assistants such as ChatGPT, Claude, Copilot and Gemini query its data directly. Its entry Sourcing Core plan is published at $149 a month paid annually or $179 month to month - SeekOut, with higher tiers quoted.

SeekOut's traditional strength is specialist and technical search, with filters for skills, publications and code that LinkedIn's profile-based model handles less well. The trade-off is that, like every open-web tool, it relies on periodic data refreshes rather than live engagement signals, and its outreach runs over email rather than InMail. It is a strong second tool for teams that keep LinkedIn for commercial roles.

Gem: CRM, ATS and agents in one

Gem approaches the problem from the candidate-relationship side. It combines an ATS, a recruiting CRM and sourcing agents that search more than 800 million profiles alongside your own history, and in 2026 it added a fraud-detection agent and an official connector for Claude and ChatGPT. Published pricing covers only startups (from $130 a month for small companies on an annual term); everyone else is quoted, and Vendr's buyer data puts the median contract around $25,700 a year - Vendr.

Gem fits teams that want one system of record for every candidate interaction, including nurture campaigns and pipeline analytics, more than teams that want a pure sourcing agent. If you already run Gem as your CRM, Hiring Assistant 2 and Gem's agents overlap substantially; if you run Greenhouse or Lever, LinkedIn's new ATS evaluation feature narrows the gap that Gem's rediscovery used to fill.

Metaview fillmore: the pay-on-hire agent

Metaview, best known for its interview notetaker, made fillmore generally available on September 30, 2026 alongside a $60 million Series C - Metaview. fillmore sources, writes outreach, follows up and books screening calls, and it runs entirely inside Slack rather than in a web app. Its pricing model is the most radical in the category: you pay only when you hire, though the fee is not published.

Pay-on-hire pricing changes the risk equation completely, since the vendor rather than the buyer carries the cost of a failed search. It also changes the incentives: an agent paid per hire is optimized to close, which is good for speed and requires close attention to quality and fit. For teams that hire occasionally and cannot justify a five-figure annual commitment, it is a genuinely different option from LinkedIn's seat-based add-on.

HeroHunt.ai: open-web AI recruiter with published pricing

HeroHunt.ai takes the open-web side of the split. Its AI Recruiter searches across platforms (up to a billion profiles depending on plan), screens each profile with large language models against the recruiter's written brief rather than against keyword matches, finds candidate emails and runs personalized outreach automatically. Pricing is published and self-serve: plans start at $99 a month with unlimited positions and searches, metered on AI-screened profiles delivered and candidate emails found.

The honest comparison with Hiring Assistant 2 cuts both ways. HeroHunt.ai reaches people LinkedIn's agent cannot, needs no Recruiter seat and can be tested on a single hard role in an afternoon. It does not have LinkedIn's engagement and Open to Work signals, does not send InMail, and is not an ATS, so it sits beside your system of record rather than replacing it. Teams typically use it for the roles where LinkedIn's pool runs thin, and keep LinkedIn for the roles where it does not.

How to choose

The cleanest way to choose is to start from your hardest roles rather than from the vendors. If your hardest roles are commercial and corporate roles in English-speaking markets, and you already pay for Recruiter Corporate, Hiring Assistant 2 is the natural default and the version 2 upgrade costs you nothing. If your hardest roles are technical, specialist or in markets where LinkedIn is thin, an open-web agent will find candidates LinkedIn's agent never will. If your problem is inbound volume rather than sourcing, applicant-screening tools (including LinkedIn's own applicant evaluation) matter more than any sourcing agent.

For most mid-sized and large teams, the answer is not either-or. The most common pattern in 2026 is one platform agent for the network where most of your candidates live, plus one open-web agent for everything else, with both writing into the same ATS. That combination costs less than most teams expect, because the open-web tools are priced per user or per outcome rather than as enterprise add-ons, and it removes the single point of failure that comes with depending on one network for every hire.

10. How to Evaluate, Pilot and Roll Out Hiring Assistant 2

The single most important thing to know about rolling out Hiring Assistant 2 is that the agent gets better with use, and so does your lock-in. Memory is the headline feature of version 2, which means the first 60 to 90 days of usage are not just a trial. They are the period in which the agent learns your hiring patterns, and the period in which you decide whether you want those patterns to live inside LinkedIn. A good pilot is therefore designed to answer two questions at once: does the agent produce better pipelines for your roles, and are you comfortable with what it will know about you a year from now?

Adoption takes longer than most teams plan for, even among LinkedIn's showcase customers: TCS, an early pilot user, described itself in September 2026 as "only halfway through adoption, as recruiters continue to adjust to new workflows" - Business Today. Most teams get the rollout wrong in a predictable way. They switch the agent on for every open requisition at once, let it run on default settings, and then judge it on gut feel after a month. The result is noisy data, frustrated hiring managers and a renewal decision made on anecdotes. The better approach treats Hiring Assistant 2 like any other expensive hire: give it a clear job description, a defined set of roles, a baseline to beat and a review date. That structure costs almost nothing to set up, and it is the only way to know whether the add-on is earning its price.

Step 1: Pick the right pilot roles

The pilot roles decide the verdict, so choose them deliberately. Hiring Assistant 2 is strongest where the ideal candidate keeps an active, accurate LinkedIn profile and where you hire the same kind of role repeatedly, because repetition is what feeds the memory layer. It is weakest on one-off executive searches, on roles where talent lives off LinkedIn (many engineering, research, trades and healthcare roles), and on markets outside the English-language rollout. A fair pilot includes some of each so you learn where the boundary sits for your company, instead of discovering it six months into a contract.

Five roles is the right size for most teams. Fewer than that and one unusual requisition can swing the verdict; many more and the pilot becomes a rollout with no control group. Keep the rest of your open roles on your current process so you have a live comparison running in parallel, under the same market conditions, rather than comparing the agent against last year's numbers. A practical pilot mix for a mid-sized team looks like this:

• Two repeatable roles you hire every quarter (for example account executives or customer success managers)
• One hard-to-fill technical role where your best hires historically came from referrals or GitHub
• One high-volume applicant role to test applicant review and screening
• One hiring manager known to give fast, specific feedback

That mix is not arbitrary. The repeatable roles test the memory feature where it should shine, the technical role tests the walled-garden limit where it should struggle, and the high-volume role tests the applicant evaluation and pre-screening features that version 2 expands. The engaged hiring manager matters more than most teams realize, because the agent's personalization depends on feedback, and a manager who never rates candidates gives the memory nothing to learn from. If the agent wins on all four role types, you have a strong case for broad rollout. If it wins on two and loses on two, you have learned exactly where to deploy it and where to keep another tool.

Step 2: Set a baseline before you switch it on

Without a baseline, every pilot result looks good. Before the agent touches a role, record what your team achieved on the last comparable requisition: how many profiles were reviewed per InMail accepted, the InMail acceptance rate, recruiter hours spent on sourcing and applicant review, days from intake to first qualified slate, and the hiring manager's slate acceptance rate. LinkedIn's own customer metrics are measured against "traditional sourcing" on LinkedIn, not against your team, so your own baseline is the only benchmark that reflects your market, your brand and your roles.

Track the same numbers for the agent-assisted roles during the pilot, and add two that LinkedIn will not report for you. The first is qualified-slate yield: of the candidates the agent surfaces, how many does a human recruiter agree are genuinely qualified? The second is candidate experience: reply sentiment, unsubscribe or "not interested" rates, and any complaints. An agent that doubles outreach volume while doubling negative replies is damaging your employer brand on LinkedIn, the one network where every candidate can see your company page.

Step 3: Feed the memory deliberately

Version 2's memory only helps if it learns the right lessons. Treat the intake conversation as the most important input of the whole process: paste the full job description, the hiring manager's real priorities (not the boilerplate requirements), the deal-breakers, and two or three examples of past hires who succeeded in the role. Then rate candidates consistently. Every thumbs-up and thumbs-down teaches the agent what "good" means for your company, and inconsistent feedback teaches it noise.

It is also worth being careful about what you teach it. If your past hiring skewed toward a narrow set of schools, employers or demographics, a memory system that learns "what you have hired before" will happily reproduce that skew at scale. Review the fit signals the agent shows for its recommendations, and if they lean on proxies you would not defend in a bias audit, correct the brief rather than letting the pattern harden. This is not a theoretical concern; it is exactly the failure mode that regulators in New York, Colorado, Illinois and the EU are writing rules about.

Step 4: Decide your autonomy level per stage

Hiring Assistant 2 can take on more of the process than version 1, from intake through screening to interview coordination, but nothing forces you to hand over every stage at once. The sensible pattern is graduated autonomy: let the agent run sourcing and shortlisting with recruiter review, let it draft outreach that a human approves for the first few weeks, and only then allow it to send within agreed limits. Automated pre-screening deserves the most caution, because that is the stage where an automated decision most directly affects a candidate and where most AI hiring laws apply.

A simple way to document this is a one-page autonomy map per role family: for each stage, who decides, who reviews, and what the agent is allowed to do without asking. Write it down, share it with legal and the hiring managers, and revisit it at the 30-day review. That page doubles as evidence of human oversight if a regulator, auditor or candidate ever asks how a decision was made, which is worth far more than the hour it takes to write.

Step 5: Run the 30, 60 and 90 day reviews

LinkedIn's Hiring Assistant Monthly Report gives admins team-level performance data and recruiters personalized guidance, which is a useful input for these reviews but not a substitute for your own baseline - LinkedIn Talent Solutions. Hold three short reviews. At 30 days, check quality: are the surfaced candidates genuinely qualified, and is the hiring manager accepting the slates? At 60 days, check efficiency: compare recruiter hours and time-to-slate against your baseline. At 90 days, check economics: divide the add-on cost by the hires and qualified interviews it produced, and compare that cost per outcome to the alternatives you could have used for the same roles.

The 90-day number is the one to take into your renewal negotiation. If the agent delivered, you have proof of value and can negotiate volume pricing with confidence. If it underdelivered on certain role types, you have the data to buy fewer licences and cover those roles with a cheaper open-web tool. Either way, you negotiate from evidence instead of from a vendor's case study, and that is the difference between a tool you control and a tool that controls your budget.

11. The Future: Where Recruiting Agents Go After Version 2

Hiring Assistant 2 tells you where the whole category is heading: from agents that do tasks to agents that run processes. Version 1 was a sourcing assistant that happened to be conversational. Version 2 reasons about which tools to use, remembers your history, pre-screens candidates by voice or text and helps coordinate interviews. The next step, which every serious vendor is now building toward, is an agent that owns a requisition end to end and reports back on outcomes, with the recruiter acting as manager, editor and closer rather than operator.

That shift changes what recruiters are paid to do. When sourcing, first-touch outreach and initial screening are automated, the human value moves to the parts no agent does well yet: understanding what a hiring manager actually needs, persuading a passive candidate to take a risk, negotiating an offer, and judging the evidence an agent cannot see. Teams that treat the agent as a cheaper sourcer will capture some savings. Teams that redesign the recruiter role around supervision and closing will capture most of them.

Agents on both sides of the table

The most underrated trend is that candidates now have agents too. Job seekers use AI to rewrite profiles, tailor applications and apply at volumes no human could manage, and recruiters on every platform report applicant floods of polished, near-identical submissions. This is why version 2 leans so heavily on verification and trust signals: in LinkedIn's own survey, 64% of US talent acquisition professionals said it has become harder to know what is real about candidates - LinkedIn Newsroom. When a profile can be generated in seconds, the value moves from what a profile says to what can be confirmed about it. Expect verification, skills evidence and fraud detection to become a core battleground for every recruiting agent through 2027.

The logical end state is agent-to-agent recruiting, in which a company's hiring agent and a candidate's career agent exchange structured information before any human is involved. Early versions already exist in the form of AI interviewers and chat-based screening assistants. The open questions are about consent, accuracy and fairness: who is accountable when two automated systems decide a candidate is not a fit, and how a rejected candidate can contest a decision that no person made. Regulators are moving toward requiring a human in the loop for exactly these moments, which is why the "assistant" framing in LinkedIn's product name is a legal stance as much as a marketing one.

The platform war: walled gardens versus open agents

The second trend is a structural split between platform agents and open agents. LinkedIn, Indeed and the large HR suites are building agents that run inside their own data and their own workflows, which gives them unmatched data on their own turf and a natural ceiling at the edge of it. Independent agents take the opposite bet: search across the open web, plug into whatever ATS you already use, and compete on reach and price rather than on owning the network.

LinkedIn is defending its side of that line actively. In September 2026 it took down the LinkedIn company pages of two sourcing vendors, Pin and Gem, stating that it "takes action when necessary to prevent our members' information from being scraped and used without their consent" - ERE. Pin says it remains operational and is resolving the matter. Whatever the merits of that dispute, the direction is clear: LinkedIn is pulling more data into its own agent while making it harder for outside tools to pull data out.

For buyers, the practical implication is that the stack is unlikely to consolidate into one agent. Most teams will run a platform agent where their candidates already live (LinkedIn for commercial and corporate roles, Indeed for hourly and frontline roles) and an open-web agent for everything the platforms cannot see. The integration layer between those agents and the ATS, increasingly built on open standards such as the Model Context Protocol that lets AI assistants call external tools, will matter more than any single agent's feature list.

What to watch through 2027

Predictions about AI agents age badly, and most forecasts made in 2024 about fully autonomous recruiters by 2026 did not come true. So instead of forecasting a date for fully autonomous recruiting, it is more useful to watch for concrete signals. Each of the signals below is observable, most will be announced publicly, and each one changes a specific part of the buying decision for Hiring Assistant 2 or its alternatives.

None of them requires insider knowledge to follow. Product release notes, earnings calls, regulatory calendars and your own LinkedIn account team will surface all five over the next eighteen months, and a quarterly check against this list takes less time than a single vendor demo. A handful of signals will tell you how fast this future arrives:

• Non-English rollout of version 2 and the regions it reaches next
• Published pricing or usage-based plans replacing negotiated add-ons
• Third-party audits of agent screening accuracy and bias
• EU AI Act high-risk obligations for hiring tools taking effect
• Open data access, or the lack of it, for agents outside LinkedIn

Each item maps to a decision. Wider language support decides whether a global team can standardize on one agent. Transparent pricing decides whether smaller teams can buy at all, since today the add-on is effectively an enterprise product. Independent audits decide whether the self-reported metrics can be trusted. The EU deadline, now December 2027, decides how much documentation every vendor must produce, which in practice raises costs and slows launches in Europe. And data access decides whether LinkedIn's walled garden stays walled, which is the single biggest factor in whether an open-web agent remains a necessary second tool. If you only revisit your agent strategy once a year, tie that review to these five signals rather than to the vendor's release calendar.

12. The Verdict: Should You Buy Hiring Assistant 2?

Hiring Assistant 2 is the best AI agent for searching LinkedIn, and that is both its recommendation and its limit. If you already pay for the Hiring Assistant add-on, the decision is easy: version 2 arrives at no extra cost, the memory and fit signals should make your projects better with use, and the voice pre-screen and ATS evaluation features are worth switching on carefully. Your only real job is to configure it well, measure it honestly and keep control of what it learns.

If you do not have Hiring Assistant yet, version 2 does not change the size of the decision, only its quality. You are still buying an enterprise Recruiter seat plus an agent add-on, negotiated rather than published, renewing annually, with InMail consumption and monthly credit limits you need to model. That is a sensible purchase for some teams and an expensive mistake for others, and the difference comes down to where your candidates live and how repeatable your hiring is.

Staffing agencies face a slightly different calculation. They buy through Recruiter Professional Services Plus, where the capacity tiers favor evaluating large applicant volumes, and their economics depend on placements per recruiter rather than cost per hire. For an agency already on RPS+, version 2's ATS evaluation and pre-screening can lift recruiter throughput meaningfully; for an agency that has not committed to RPS+, the add-on is one more fixed cost in a business with variable revenue, and pay-per-use or open-web tools may fit the model better.

A simple decision framework

The framework below condenses the guide into four questions. Answer them honestly for the roles that matter most to your business, not for your easiest roles, because the hardest roles are where an agent earns or wastes its price.

It also helps to answer them with the people who will live with the decision in the room: the recruiters who will supervise the agent, a hiring manager who will review its slates, and whoever owns the LinkedIn contract and the renewal date. Agents fail most often not because the technology is weak but because the people around them were never asked whether the workflow fits how they actually hire. Thirty minutes spent on these questions with that group will save months of arguing about a tool after it has been bought.

• Already paying for Hiring Assistant? If yes, turn on version 2 and run the pilot in section 10.
• Candidates active on LinkedIn? If mostly no, prioritize an open-web agent.
• Hiring repeatable? Memory pays off on repeat roles, much less on one-off searches.
• Entry ticket affordable? If a Recruiter Corporate seat is out of budget, start with a self-serve tool.

If you answered yes to the first three, check one more thing before you sign: whether your ATS is on LinkedIn's integration list for Hiring Assistant, because without it the talent-pool feature does not exist for you. With that box ticked, Hiring Assistant 2 is very likely the right primary agent, and the remaining question is how to negotiate it. If you answered no to the second or fourth question, you will get more hires per dollar from an open-web agent with published pricing, used alone or alongside a basic LinkedIn presence. And if your answers are mixed, which is the most common case, the two-agent pattern from section 9 is usually the best value: Hiring Assistant 2 for LinkedIn-native roles, an open-web agent for the rest, both writing into one ATS.

Whatever you choose, judge it on your own numbers. LinkedIn's metrics are directional, every vendor's case studies are curated, and the only benchmark that matters is cost per qualified candidate on your roles, against your own baseline, over a 90-day pilot. Teams that measure that way rarely overpay for an agent, and they negotiate renewals from evidence rather than from hope.

One final point is easy to lose in a feature-by-feature comparison. The agent is only as good as the recruiter supervising it, and the biggest gains tend to come from teams that redesign the recruiter's job around reviewing, persuading and closing, not from teams that simply buy the most capable tool. Choose the agent that fits your candidates, then invest just as seriously in the people who will run it.

This guide reflects LinkedIn Hiring Assistant 2 as announced in late September 2026, ahead of its November 2026 rollout. Features, availability and pricing for AI recruiting agents change quickly, and LinkedIn's rollout is staged by language and feature, so verify current details with LinkedIn and with each vendor before you buy.