How LinkedIn's AI Recruiter Reads Your Profile
If you apply or get sourced through LinkedIn, there is a decent chance software reads your profile before a person does. In September 2026, LinkedIn announced Hiring Assistant 2, the next version of its recruiting agent, rolling out to English-language customers from November.
LinkedIn does not explain how its fit scores are calculated, so nobody outside the company can tell you the formula. But the announcement and ERE’s coverage of it describe the inputs, and those are enough to act on.
What it says it does
Hiring Assistant uses memory from a recruiter’s past hires to infer what they mean rather than just what they typed, covering things like location, industry and seniority. It produces a “fit” assessment and an estimate of candidate interest.
According to LinkedIn, interest is estimated from platform engagement, Open to Work details and InMail responsiveness. Other inputs include verification status, ATS insights, pre-screening responses and relevant work experience. When a company connects its applicant tracking system, the agent can also evaluate people who applied outside LinkedIn.
Version 2 extends into screening. ERE reports a median pre-screening time of six minutes, after which standouts are flagged for a recruiter and borderline candidates are sent on for a screen. LinkedIn’s own figures say recruiters are four times more likely to contact candidates sourced by the tool than by traditional methods. Those are vendor claims, from LinkedIn’s announcement, not independent measurements.
LinkedIn also reports pilot results from two customers: Cisco saw roughly a 60-fold increase in candidate review volume, and Zillow said sourcing efficiency improved by roughly a third. For candidates, the first number matters most. More profiles reviewed per recruiter means yours is read faster, and judged on less.
Five things to do
- Verify your identity. LinkedIn says 115 million members have verified. It also says 39% of recruiters name not knowing whether candidates or their skills are real as a top challenge, and that verification status now shows in several ATS products. An unverified profile gives a reader one more reason to hesitate.
- Fill in Open to Work details. They are one of the stated inputs for estimating interest. If you are looking, say what roles and locations you want, and who can see it.
- Answer InMail, even to decline. Responsiveness is a named signal. A short, polite reply costs nothing.
- Connect apps that show real work. LinkedIn says members with a connected app receive up to 2.9x more recruiter messages and lists partners such as Behance and Replit for work highlights. Treat the multiple as marketing, but the underlying logic is sound: evidence beats assertion.
- Write for meaning, not keywords. If the system infers intent, specific beats generic. “Built and evaluated a retrieval pipeline for a support assistant” tells it far more than “AI enthusiast.” Name the work, the tools and the outcome.
What to be skeptical about
ERE reports that LinkedIn removed the company pages of two AI sourcing companies, Pin and Gem, in September, and that a LinkedIn spokesperson said the company acts to prevent members’ information from being scraped. Whatever the dispute’s merits, it is a reminder that these systems sit on a platform with its own commercial interests.
Also remember that the tool serves recruiters. Fit signals can flag mismatches as well as matches, and you cannot see them. Do not tailor your whole career story to a score you cannot inspect. Tailor it to what a human hiring manager will want to read after the agent hands you over.
The practical takeaway
Make your profile easy to verify, easy to read and easy to say yes to. That helps with an AI screen, and it helps just as much with the person who reads your profile afterwards.
If you are targeting AI roles, apply the same standard to the postings: use the AIfirst.jobs feed to find roles where AI is the daily work, then make your profile show you doing that work.