LinkedIn is telling job seekers that a richer, more active profile will help its new AI recruiting agent, Hiring Assistant 2, find them. In practice, that turns the profile into an AI resume, and it carries every weakness of the paper version: it measures how well people describe their work, not how well they do it. Now the same platform also offers an AI tool that writes applications for candidates. Research shows AI screeners favor AI-written text, and employers already say they don't trust resumes. A better-ranked profile won't fix a weak signal. Assessments, structured interviews and work samples will.

Update your LinkedIn profile and the machines will find you.

That was the message from LinkedIn's Talent Connect conference this fall. LinkedIn Chief Business Officer Mark Lobosco told Forbes that the richer a member's profile, "the more signals can be discovered" by Hiring Assistant 2, the company's next-generation recruiting agent. He urged professionals to spell out past responsibilities, accomplishments and leadership in more detail, and to keep it current.

It's sensible advice for an individual. For the hiring process, it's a warning sign. What LinkedIn is describing is an AI resume: a self-reported document written for a ranking system rather than a reader. That's the same flawed resume problem that has plagued recruiters for years, running faster and at greater scale.

What LinkedIn's Hiring Assistant 2 actually reads

LinkedIn announced Hiring Assistant 2 at the end of September. The company says more than 20,000 companies already use its agentic hiring products, and the new version adds stronger reasoning, memory of how each recruiter has hired before, and more indicators of why a candidate matches.

The memory feature is the headline. LinkedIn's own example: a recruiter asks for a product manager, and the agent now returns one near the office, from retail, at mid-level, because that's who the recruiter hired in the past.

The agent also estimates how receptive a candidate is likely to be, using platform engagement, Open to Work details and InMail response history. LinkedIn is adding verification badges as trust signals, along with voice pre-screening and the ability to evaluate candidates already sitting in an employer's applicant tracking system.

Lobosco told Forbes the inputs go further still. A member's activity, interests, the companies they follow and even the videos they watch can help surface matches.

A resume used to be something a candidate wrote once per application. The AI resume is a behavioral record that updates every time they log in.

LinkedIn now sells both sides of the application

Here's the part of the story that didn't come up at Talent Connect. In June, LinkedIn began rolling out a Premium Apply Assistant for paying job seekers. It flags roles where a member is a top match, pre-fills the application and drafts an introductory letter. It works on roles posted outside LinkedIn, too.

LinkedIn's own help documentation says draft creation and AI assistance aren't shown to recruiters. The employer sees a finished application. It doesn't see who, or what, wrote it.

So one platform now offers an AI that helps candidates write the application, and another AI that helps employers rank it. Both feed on the same profile. Social Media Today's Andrew Hutchinson pointed out the obvious tension at launch: the more an application is assembled by software, the less it tells a recruiter about the person who submitted it.

That's the resume problem in its purest form. The document gets better at looking like a fit. The signal underneath it gets weaker.

The volume numbers show where this leads. LinkedIn now processes about 11,000 applications a minute, and Greenhouse data shows the average open role drew 244 applications in 2025, up from 116 in 2022, as we recently reported. Criteria's 2026 Candidate Experience Report found 34% of job seekers have already used AI to apply.

Why an AI resume is still a resume

The resume was always a proxy. It rewards people who write confidently, match keywords and have networks that coached them on what to say. Ranking a profile with AI keeps all of that and adds three new problems.

First, it scores visibility. If posting, engaging and answering InMails feed the ranking, the edge goes to people with time to perform on LinkedIn. That cuts against the heads-down employee, the caregiver returning from a gap and anyone in a field where LinkedIn isn't part of the workday.

Second, it learns from the last hire. "Find me another one like before" is a convenience. It's also how yesterday's hiring patterns get copied into tomorrow's shortlist, gaps included.

Third, it only sees its own pool. HeroHunt notes the agent finds only LinkedIn members and people already in an employer's ATS, and that most of its published performance figures are self-reported.

Employers already know the underlying signal is weak. In a 2026 study by Lighthouse Research & Advisory and Criteria of 998 employers, only a third said they were very confident resumes reflect a candidate's true skills. Two in three still use a resume screen as their first step. Companies that rely on resumes as their primary decision driver were 35% more likely to report a bad hire, and 92% of recruiting leaders said AI-generated resumes are now common in their applicant pools.

Criteria's 2026-2027 Hiring Benchmark Report found about 4 in 10 are struggling with candidates who use AI to misrepresent themselves, and more than a third say they can't reliably detect it, as Talent Signal News covered this month.

Candidates aren't attached to the resume, either. Criteria's Candidate Experience Report found 68% of job seekers would skip it for a process built on assessments and structured interviews. "The resume was built for a different era of work," Criteria CEO Josh Millet said.

Moving that document onto a platform and ranking it with AI doesn't bring it into the new era. It just makes the old one faster.

Do AI screeners favor AI-written profiles?

The research says they can, and that should worry anyone pairing AI-written applications with AI ranking.

Researchers Jiannan Xu, Gujie Li and Jane Yi Jiang tested large language models as resume screeners in a study presented at AIES. Holding each candidate's qualifications constant, models chose resumes they had written themselves over human-written versions roughly 67% to 82% of the time, according to the University of Maryland. Criteria broke down the findings: GPT-4o picked its own version more than 80% of the time, which means candidates may be ranked partly on which chatbot they used.

The older bias problems haven't gone away, either. University of Washington researchers Kyra Wilson and Aylin Caliskan found that language-model resume screening favored names associated with white and male candidates.

And the human in the loop isn't the safeguard it's assumed to be. In an experiment with 528 participants across 1,526 screening scenarios, reviewers largely adopted an AI's biased recommendations, even when they said they didn't trust it.

The trust gap shows on the candidate side. Only 8% of job seekers believe AI makes hiring fairer, while 70% of hiring managers trust AI to make faster, better decisions, according to Greenhouse's 2025 AI in Hiring Report. Greenhouse CEO Daniel Chait calls the result an "AI doom loop," with candidates using AI to apply at scale and employers using AI to filter it. Among U.S. job seekers in that survey, 41% admitted to hiding text in applications to slip past AI filters.

An AI resume doesn't break that loop. It gives it a new front door.

AI capability is outrunning accountability

Hiring is one corner of a bigger pattern. Stanford HAI's 2026 AI Index, released in April, describes AI capability racing ahead of the systems meant to govern it.

Agents improved fastest. Success rates on Terminal-Bench, a test of real-world task completion, rose from 20% to 77.3% in a year, and average scores on the Foundation Model Transparency Index, which tracks how much AI developers disclose, fell from 58 to 40. That's the class of technology now sorting candidates, and it's getting less transparent as it gets more capable. Documented AI incidents also rose to 362 in 2025 from 233 the year before.

The public isn't convinced. The Index found 73% of U.S. experts expect AI to help the job market, compared with 23% of the general public.

Courts are starting to test who's responsible when software screens people out. In Mobley v. Workday, a job seeker alleges Workday's AI screening rejected him from more than 100 jobs because of age, race and disability. The court let the case proceed on the theory that Workday could act as an employer's "agent" when customers hand it the work of rejecting and advancing candidates. In June, the court kept certain disability and California claims alive. Workday denies the claims, and the case is in discovery.

That "agent" question is worth watching as LinkedIn markets Hiring Assistant as exactly that. Regulators are moving too: Ontario now requires employers to say in public job postings when AI is used to screen or select applicants.

What to screen on instead

The fix isn't a better profile. It's a different first signal entirely.

Treat a LinkedIn profile, AI-ranked or not, the way you'd treat a resume: background for an interview, not the filter that decides who gets one. Put something candidates have to demonstrate at the front of the process instead. Short, validated assessments are far harder to fake with AI than a resume, and structured interviews predict performance about twice as well as unstructured ones. In the Lighthouse study, 98% of employers already rated those methods as more reliable than resumes. Criteria builds assessments on that premise.

If you do use an AI sourcing agent, audit what it learns. A memory built on your last ten hires copies their blind spots. And given how readily reviewers follow AI picks, someone should check the candidates it passed over, not just the ones it surfaced. Talent Signal News has a list of questions to ask AI hiring vendors before you buy.

Lobosco closed his Forbes interview by urging people to look past their next move to the one after it. Hiring needs the same advice. The next move is a better profile. The move after that is to stop treating any self-reported document, paper or AI-ranked, as the strongest signal of what someone can do.

FAQ

What is LinkedIn Hiring Assistant 2?

Hiring Assistant 2 is LinkedIn's updated AI recruiting agent, announced in September 2026. It searches LinkedIn members and an employer's applicant tracking system for candidates, remembers what a recruiter has hired for before, estimates how receptive candidates are likely to be, and adds verification signals and voice pre-screening.

What is an AI resume?

An AI resume is a candidate profile written to be read and ranked by AI rather than by a person. On LinkedIn, that includes a member's profile details as well as activity signals like engagement, follows and InMail responses that the platform's recruiting agent can use to surface matches.

Does LinkedIn tell recruiters when an application was written with AI?

No. LinkedIn's Premium Apply Assistant can pre-fill applications and draft introductory letters, and LinkedIn says AI drafting isn't shown to recruiters. Employers see only the submitted application.

Do AI resume screeners favor AI-written resumes?

Research suggests they can. A study by Jiannan Xu, Gujie Li and Jane Yi Jiang found large language models chose resumes they wrote over human-written versions of the same candidate roughly 67% to 82% of the time, with GPT-4o above 80%.

Do employers still trust resumes?

Increasingly, no. In a 2026 Lighthouse Research & Advisory and Criteria study of 998 employers, only a third were very confident resumes reflect true skills, yet two-thirds still screen resumes first. Ninety-eight percent rated assessments, structured interviews and work samples as more reliable.

What should employers use instead of resume or profile screening?

Start with evidence a candidate has to demonstrate: short, validated assessments, structured interviews and work samples. Use résumés and profiles as background for interviews rather than as the first filter, audit any AI screening tool for bias, and disclose where AI is used.