Refolk
11 min readMarkdown

LinkedIn's Hiring Agent Flags Mismatches. Audit Your Last 90 Days

LinkedIn Hiring Assistant 2 cross-checks your profile, screening answers, and ATS history for mismatches. Here is the 90-day activity audit to run this week.

If your LinkedIn profile, your pre-screen answers, and the resume you sent last year tell different stories, LinkedIn's recruiting agent can now flag the gap. Hiring Assistant 2 reaches existing recruiter customers from November, and LinkedIn says it cross-checks your verification status, the employer's ATS records, and your pre-screening answers. A LinkedIn spokesperson also told one outlet it scans your posts and comments. If the last 90 days of your activity don't back up your resume, fix that this month, before a recruiter's shortlist gets built without you on it.

What does LinkedIn Hiring Assistant 2 actually read?

LinkedIn officially lists five inputs: your work experience, your verification status, your pre-screening answers, the employer's ATS records, and skill evidence from connected apps. The claim that it reads posts and comments comes from a single interview. It does not appear in LinkedIn's announcement.

LinkedIn Hiring Assistant is LinkedIn's AI recruiting agent. It evaluates and pre-screens candidates for a recruiter. LinkedIn announced version 2 on September 29, 2026, at its Talent Connect event in New York. Existing English-language customers get it automatically, at no extra cost, starting in November. Aurecon and MediaNews Group are two of the customers LinkedIn names on its product page.

Here's where each input comes from:

Input the agent usesWhere it's statedStatus
Profile work experienceLinkedIn announcementConfirmed
Verification statusLinkedIn announcementConfirmed
Pre-screening answers (text or voice)LinkedIn announcementConfirmed
ATS records (Greenhouse, Lever, SmartRecruiters)LinkedIn announcement, EREConfirmed
Connected apps (Base44, Fiverr, Duolingo, HubSpot)LinkedIn announcementConfirmed
Posts and commentsCalcalist interview with a LinkedIn spokespersonReported, one source

In the Calcalist interview, the spokesperson said the AI "scans not only what is written in the profile, but also posts and comments." So should you assume LinkedIn's AI reads your posts and comments? Treat it as likely, not settled. The audit below is quick, and it's worth doing whether the agent reads comments or not.

Why does the agent's shortlist decide who gets seen?

Recruiters using the agent start from a filtered list, not the whole pool. LinkedIn says the first version cut the number of profiles recruiters review by 83%.

83%
Fewer profiles reviewed by recruiters using Hiring Assistant v1

LinkedIn's own figure from its announcement. Its product page says 81%. The 80% figure that often gets repeated is wrong.

That changes how you get found. A recruiter used to scroll through a long list and form a judgment. Now they open a short list the agent has already filtered. A strong profile that doesn't make that list may never be read.

Version 2 also adds memory. Dan Reid, VP of Product at LinkedIn Talent Solutions, says the agent "understands more about how you hire, not how everybody hires." Read that as: you're matched against what that particular recruiter looks for, not a generic standard. That raises the bar on specifics. A profile that clearly says what you did, at what scale, with what result, gives the agent something to match.

Who is on the other end of the agent?

In the US, recruiters outnumber people who publicly say they're looking by about five to one. In Refolk's index of professional profiles, 123,514 US profiles carry Recruiter, Technical Recruiter, Talent Acquisition Partner, or Sourcer titles. Only 24,398 US profiles have "open to work" in the headline.

Segment (Refolk's index, Nov 2026)Figure
US recruiter, TA, and sourcer profiles123,514
UK recruiter, TA, and sourcer profiles9,444
US vs UK multiple~13.1x
US profiles with "open to work" in headline24,398
US recruiters per "open to work" headline~5.1
Top title in sampled recruiter profiles (n=25 each)"Technical Recruiter" 36% (US) vs "Talent Acquisition Partner" 52% (UK)

A few caveats:

  • The UK count probably misses people titled "Recruitment Consultant," so read the 13x gap as a lower bound.
  • The "open to work" figure only counts people who typed the phrase into their headline. It leaves out the green-banner setting, so the real number is higher.

Even so, the ratio matters. With about five recruiters for every self-declared job seeker, a headline that says you're looking isn't what's scarce. Evidence the agent can verify is.

The headline is also the most exposed part of a job search. Your current manager can see "open to work." Former colleagues can see it. If you're still employed, you can't announce it at all. So you're left applying through forms and waiting, and the silence tells you nothing about why.

Why does consistency matter more than activity?

The biggest risk isn't looking quiet. It's looking contradictory. ERE reports that the agent combines verification checks with ATS data, pre-screening answers, and work history to "flag inconsistencies."

A vague profile gives the agent little to work with. A contradiction is worse: two sources the system treats as reliable disagree, and nothing in the record says whether that's an honest typo or an inflated claim. LinkedIn is also adding more sources to cross-check against:

  • 115 million members are now verified.
  • Employers will be able to flag profiles that falsely claim to have worked there.
  • Colleagues can verify each other's employment.
A quiet profile reads as thin. A contradictory one reads as a risk.

These are the contradictions to look for first:

  • Title drift. Your profile says "Senior Product Manager" and the resume you sent says "Product Manager."
  • Date drift. An end date on LinkedIn is off by a few months from the one on your resume.
  • Scope drift. A comment mentions "my team of four" while your resume says you led twelve people.
  • Answer drift. A pre-screen answer on location or work authorization doesn't match your profile.

Fix these before you post anything new. Adding activity on top of a contradiction just gives the agent more chances to find it.

The 90-day audit: what should you delete?

Start with your comments. Delete anything templated or AI-sounding, and anything that contradicts your resume on title, dates, or scope. LinkedIn already limits the reach of generic automated comments, so they don't help you anyway.

LinkedIn's crackdown has been public and steady:

  • Since May 2026, LinkedIn limits the reach of automated and AI-generated comments.
  • On July 30, it added a "Seems like AI slop" button that lets members flag them. Chief product officer Hari Srinivasan announced the automation crackdown.
  • LinkedIn says it blocks hundreds of thousands of automated comment attempts every day.
~30%
Share of LinkedIn comments that were fully AI-generated, April to June 2026

Pangram Labs analysis of about 57,000 public posts. Third-party research, not LinkedIn data.

If nearly a third of comments are machine-written and LinkedIn is limiting their reach, your old "Great insights!" replies aren't neutral filler. They look like the content LinkedIn is trying to suppress. Deleting them is cheap insurance, whatever the agent ends up reading.

Open your own activity history and scroll back 90 days through your comments. Delete:

  1. One-line reactions: "Great insights!", "Thanks for sharing," "So true."
  2. Comments you generated with a tool and posted without editing.
  3. Any comment that states a title, date, team size, or result that doesn't match your resume.

Then do the same pass on your posts. Most people have fewer posts than comments, so this part goes faster.

What should you post and comment instead?

Pick three claims from your resume and make every new post and comment support one of them. LinkedIn's Global Editorial VP put the rule plainly: "It's ok to use AI to help you write, but your posts and comments need to represent your voice."

A useful comment shows working knowledge in your own words. A good one has three parts:

  1. The claim. Something you actually did that the post relates to.
  2. The specific. The tool, the constraint, or the tradeoff you ran into.
  3. The outcome. The same number or result that appears on your resume.

An example shape: "I hit this with [system] at [company]. Moving [step] earlier is what cut [metric] for my team." Every bracket should match a line on your resume. If the agent reads comments, it finds corroboration. If it doesn't, the hiring manager who checks your activity before an interview finds the same thing.

Which three claims? Let the postings you're targeting decide. That's the exact work Refolk takes off you. Paste a posting and Refolk scores how well your history actually fits it, then rewrites your resume around the claims that posting rewards. Those claims are the three you repeat everywhere else.

Leave volume for last. A few substantive comments in your field each week will do more for you than daily ones that say nothing.

How do you optimize your LinkedIn profile for AI recruiters?

Put the evidence inside each role entry, not just in your headline. Work experience is one of the agent's confirmed inputs, and a headline is one line while your experience section can carry the whole case.

This matters most if you're between jobs. In a sample of "open to work" profiles from Refolk's index, the most common current titles were placeholders like Co-Founder, Independent Consultant, and Semi Retired (n=25, a small sample). Those titles may be honest, but they give the agent nothing to match. If your current entry is a placeholder, the past roles under it have to do the matching.

This week:

  • Rewrite your last two or three role entries so each one states the same scope and results as the matching resume bullets.
  • Match titles and dates exactly to the resume you send most often.
  • Complete verification. With 115 million members verified, an unverified profile is now the odd one out.
  • Connect a skills app if you use one of the named partners (Base44, Fiverr, Duolingo, HubSpot). That puts outside proof of a skill on your profile.
  • Make the headline a short version of your three claims, not a status update.

Why do your pre-screen answers matter so much now?

Your pre-screen answers end up next to your verified record, and a pre-screen with a median length of 6 minutes decides whether you go to the recruiter or to another screen. Before, a recruiter typically took 3.5 hours just to respond for the first time.

The agent can run the pre-screen by text or by voice. Stand-out candidates go to the recruiter. Borderline ones get sent to a screen. That's where LinkedIn's AI screening now decides who moves on, and it's easy to answer on a phone, in two lines, without thinking about it.

Treat each answer like a resume bullet:

  • Use the same numbers. If your resume gives a budget or team size, use that exact figure, not "a large budget."
  • Use the same title. Don't upgrade or downgrade it to suit the question.
  • Answer the question that was asked first, then add one line of evidence.
  • Keep yes/no answers literal. If you meet the requirement, say yes. If you partly meet it, say yes and add one line explaining how.

Can your old applications follow you to the same company?

Yes. Through ATS integrations with Greenhouse, Lever, and SmartRecruiters, the agent can evaluate an employer's whole talent pool, including people who applied outside LinkedIn. An ATS (applicant tracking system) is the software employers use to store applications and track candidates through hiring.

This is the least obvious risk in the whole rollout. Suppose you applied to a company last year with a resume that listed a different title, or rounded your dates differently. That application is still in their ATS. When you apply again, or when the agent surfaces you for a new role, the old record sits beside the new one, and any difference looks like a contradiction.

The fix is a single source of truth:

  1. Keep one master record of every title, employer, and start and end month.
  2. Write every tailored version from that record. Change the emphasis, never the facts.
  3. Before you reapply somewhere, check what you sent them last time if you still have it.

Refolk works this way by default. Every tailored resume is written from your one history, so the bullets change for each posting while titles and dates stay the same in every version.

FAQ

Does LinkedIn AI read my posts and comments?

Possibly. A LinkedIn spokesperson told Calcalist that Hiring Assistant 2 "scans not only what is written in the profile, but also posts and comments." LinkedIn's official announcement doesn't mention posts or comments. It lists verification, ATS insights, pre-screening responses, work experience, and connected apps. Treat activity as a likely input and clean it up, but put the most effort into the confirmed ones.

Should I delete all my LinkedIn activity before a job search?

No. An empty profile gives the agent and the hiring manager nothing to corroborate. Delete the templated comments, the AI-generated filler, and anything that contradicts your resume from the last 90 days. Keep substantive comments that show real knowledge of your field, especially ones that support the three claims on your resume.

Is it okay to use AI to help write LinkedIn comments?

LinkedIn says yes, as long as the result represents your voice. The problem is unedited output. LinkedIn has limited the reach of automated comments since May 2026 and added a "Seems like AI slop" flag on July 30. Pangram Labs estimates about 30% of comments from April to June 2026 were fully AI-generated, which is why generic comments now read as noise.

Does "open to work" help with LinkedIn Hiring Assistant?

Not much, by itself. In Refolk's index there are about 5.1 US recruiter profiles for every profile with "open to work" in the headline, so saying you're looking isn't what's scarce. What helps is evidence the agent can verify: matching titles and dates, verified employment, role entries with real scope, and pre-screen answers that use the same numbers as your resume.

The data behind this

Put this to work

Paste your career in once. Every application after that is written for you.

Drop a resume or a LinkedIn URL. I rank the live openings against it, rewrite the resume and write a cover letter for the best of them, and fill in the employer's form when you press the button. You read, you decide what goes out.

  1. 01Drop your resume

    A PDF or a LinkedIn URL. About a minute, once.

  2. 02I rank the openings

    Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.

  3. 03Each one is written up

    Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.

  • New matches ranked and written before you are up.
  • Every bullet stays inside what your history supports. Nothing invented.
  • Queued, submitted, interviewing, offer: one screen, not a spreadsheet.

Free, no card. Up to 25 applications sent a week, and nothing is sent until you say so.

Keep reading