The AI Recruiter at Zero Cost? What LinkedIn's Hiring Assistant 2 Means for TA Teams
At the end of September, LinkedIn unveiled Hiring Assistant 2 at Talent Connect. The headline that sticks: the new version comes at no extra charge from November. ERE reports that it will roll out free to English-language customers. An AI sourcing agent as standard equipment sounds like a gift to every TA team.
On a second look, the fine print holds two caveats. Behind them sits a strategic question that matters more than the price: what remains a competitive advantage when everyone points the same agent at the same talent pool?
What is actually free here
The upgrade is free, not the agent. According to the available information, the new version goes automatically to customers who already use Hiring Assistant. That in turn is an add-on to LinkedIn Recruiter, and LinkedIn does not publish prices for it. Anyone without a Recruiter contract that includes this add-on does not get an AI recruiter at zero cost, only the same entry price as before.
The price list LinkedIn filed for the UK public sector offers a reference point (G-Cloud 14, October 2025). A Recruiter seat costs between £6,350 and £8,925 per year there, depending on volume. During the promotional period, the Hiring Assistant add-on was listed at £1,575 to £2,079 per licence. The promotion ran until the end of June 2026, and what applies afterwards is not public. At the time of research, the list was only available through an analysis by the vendor HeroHunt, which competes with LinkedIn. The figures are therefore a rough guide, not a quote. They do show the pattern: the seat costs a multiple of the agent, which is exactly why “free” means nothing to new customers.
Then there is language. Hiring Assistant has been available in German since June 2026. For version 2, LinkedIn names November and English-language customers. According to OnlineMarketing.de (30 September 2026), it is still open when the German version will follow. For teams in the DACH region, that means: ask, wait, and do not plan around it yet.
What the agent can do now
In terms of content, LinkedIn is adding five things:
Memory: The agent remembers whom a recruiter and the company have hired before and searches accordingly. LinkedIn's example: not just a product manager, but one near the office, from retail, at mid-level.
Fit and interest signals: Besides the match, there is an assessment of how open someone is to outreach, derived from platform activity, “Open to Work” status and InMail responses.
Verification: Identity and employer appear as a trust signal on the candidate card.
Pre-screening by voice or text: According to LinkedIn, it takes six minutes at the median.
Applicant tracking system integration: The agent also evaluates people in your own system, initially in Greenhouse, Lever, SmartRecruiters, Zoho Recruit and Tracker. iCIMS and Workday are to follow.
The figures come from LinkedIn itself. At the 2025 launch, the claim was 62 percent fewer profiles reviewed. The German product page now says 81 percent, and the coverage of the version 2 launch says 83 percent. The direction is plausible, but these are vendor claims measured against LinkedIn's own baseline. Part of the German-language coverage, for example at Personalwirtschaft, is also labelled as sponsored content.
The commodity thesis
Until now, sourcing quality could still be improved to some extent through the choice of tools. When the most capable agent on the largest professional network becomes standard equipment, the advantage shifts from the tool to how it is used. The thesis: tooling alone no longer differentiates. Three effects are foreseeable.
First, everyone draws on the same pool.
As things stand, the agent only finds LinkedIn members and people who are already in your own applicant tracking system. Anyone looking for specialists whose work is mostly visible on GitHub, in publications or in portfolios gets a shortlist from an incomplete map that still looks convincing.
Second, the same candidates land in many senders' inboxes.
Where the fit is good, everyone shows up at the same time. Data from Gem covering 6.2 million recruiting email sequences hints at where this leads: reply rates fell from 22.6 percent (2024) to 16.9 percent (2025) as automated outreach grew. The figures are taken from the HeroHunt analysis; the original report was not verified.
Third, the interest signal reflects platform behaviour.
People who are less active on LinkedIn, often senior profiles or the very busy, drop down the ranking without being any less open to a move. That is an assessment based on how the signal is built, not a finding from a study. It argues for using the signal to order outreach rather than to decide who gets contacted at all.
Free as a strategic lever for the platform
For LinkedIn, the free jump to version 2 is not a loss but a form of lock-in. Memory makes the agent more useful every month, and switching more expensive every month. What the agent has learned about a company's hiring behaviour cannot, as far as is known, be taken along. Whether and how this memory can be viewed, corrected or exported is not publicly documented. The same applies to prices after the promotion ends and to which fields are written back to the applicant tracking system.
Then there is data sovereignty. According to the LinkedIn help centre, as HeroHunt reports it, the agent uses OpenAI models via Microsoft Azure. Inputs and outputs count as customer data and are stored beyond the session, and LinkedIn may evaluate conversations to improve the product. With the applicant tracking system integration, applicant data flows into this process.
Anyone recruiting in Switzerland or the EU should reflect this in the data protection impact assessment before anyone flips the switch. For projects with an EU location, LinkedIn turns off automatic message sending, and every message is approved by hand.
The legal framework is also moving closer, because with voice pre-screening the agent moves nearer to the decision. Under the Digital Omnibus, the high-risk obligations of the EU AI Act for recruiting AI apply only from 2 December 2027, no longer from August 2026. Emotion recognition in the workplace is already prohibited today. A voice pre-screen should therefore never be scored on tone or “impression”. This post is not legal advice.
What remains as differentiation
When search and first outreach converge, what counts is what an agent does not deliver. Three areas suggest themselves.
The quality of outreach.
The agent produces volume. Anyone who writes to fewer people, but with a reason that cannot be read off the profile alone, stands out. This is exactly where a signal like interest helps: to set the order, not to exclude anyone.
Your own talent pools.
What sits in your own system is the cheapest channel: silver medallists from earlier processes, alumni, referrals, communities. The agent can evaluate these pools better than before, but only those that exist. Anyone who never built relationships outside LinkedIn will not find them there either.
Employer brand and relationships.
Verified profiles and strong fit signals bring candidates to the door. Whether they reply is decided by the employer's image and the person who writes.
Five steps for TA teams
Clarify the licence
Without a Recruiter contract that includes the Hiring Assistant add-on, the upgrade is not available. Get a quote before the 2027 budget is set.
Ask for the timeline
When does version 2 arrive in German, and which features are available in which language? Get the answer in writing.
Pilot with a baseline
Five roles, measured beforehand: share of candidates a human rates as qualified, reply rate, review time per slate and feedback from those contacted.
Steer the memory
Use consistent rating standards across the team and, for searches that change course, explicitly override past patterns. Check fit signals for proxy characteristics that could not be defended.
Keep a second channel alive
Keep your own pools, referrals and communities outside LinkedIn active.
Conclusion
The AI recruiter does not come at zero cost, but as a bonus on top of an existing contract. That is exactly the point. When the agent becomes standard equipment, what matters is no longer who has it, but what a team does with the rest of the work: who stays approachable, who maintains pools and who has a brand candidates respond to. Until version 2 speaks German, there is time to build exactly that.
Sources
Transparency Notice on the Use of AI
AI-based tools were used in the creation of this post to provide support. They are used in particular for researching and structuring information, linguistic and grammatical revision, translation, and, in part, creating or editing illustrations and visual material.
Conception, content statements, assessments and conclusions come from the author and reflect his personal thoughts, professional experience and convictions. All content is editorially reviewed and the responsibility of the author before publication.
Posts on this website are not created or published automatically by AI. AI is used as a supportive tool; editorial and substantive responsibility remains entirely with the author.
Comments