LinkedIn just turned your coworkers into its verification layer

The prompt is small, innocuous, and polite. A face you haven't thought about in years from a company you both left, and a question: did you work with this person? It was some time ago, but you still remember the desk by the window, maybe the fancy coffee machine. You tap yes. Nothing in the exchange asks whether they were any good at the job.

That tap is LinkedIn's newest answer to a problem AI made cheap. A convincing professional profile used to take effort: a believable career arc, a photo, a few years of posts.

Now one model writes the arc and another makes the face, and Gartner forecasts that by 2028, "1 in 4 candidate profiles worldwide could be fake." In a Gartner survey of three thousand job candidates, 6% said they'd taken part in interview fraud, posing as someone else or having someone pose as them. The 1-in-4 is only a forecast.

LinkedIn announced its version on 23 September, and the company is careful to call it a test: "we’re testing a new way to verify the workplace and educational experience of the people you've actually worked or studied with." It prompts verified members to confirm connections they worked or studied with, and once someone collects enough confirmations, a badge appears on their profile. LinkedIn doesn't say how many is enough. It does say more than 115 million members have already verified their identity or workplace, so the pool is large, though only members verified at the same workplace can confirm a colleague, and none of them is on the payroll.

Your coworkers are the infrastructure now

Peer verification moves the check from documents to people. An ID proves a person exists, and a work email proves someone can open an inbox. A vouch asks a third party with a verified account of their own to say the two of you shared an office or a classroom. LinkedIn's FAQ says a verification can come "without independent review by LinkedIn or confirmation by the relevant employer or educational institution."

The proof lives in the network's memory of who worked where.

The design is smarter than it looks from the tap. LinkedIn's help pages set the rules: the voucher must be a 1st-degree connection, verified at the same workplace, with overlapping dates there and two-factor authentication on, and must clear unspecified "trust-based" checks.

That changes the economics of faking. In 2022, Stanford researchers found more than a thousand LinkedIn accounts that appeared to use computer-generated faces, and a 2025 study found existing detectors missed 42-52% of GPT-written profiles. A voucher's own verification at that employer is harder to fake than a profile.

TechCrunch reports one more rule, a connection of at least a year. LinkedIn's own pages don't list it, though the FAQ adds that "Additional eligibility requirements may apply." If the year is real, a fake network has to stay alive for twelve months before any of it can vouch for anything. Lucy Hall, whose SocialDay piece carries the same rule, puts the trade neatly: "That time threshold makes vouching expensive for attackers but cheap for genuine colleagues."

It also aims at a real gap. Email verification needs an employer with a working mail server, and plenty of jobs on resumes outlive the company that offered them. When the agency closed or the startup was bought and its domain switched off, a former colleague may be the only record left that you worked there.

I have a job like that; I'm sure we all do. From 2004 to 2005 I worked at BKV (Bennett Kuhn Varner) in Atlanta. In 2018 it merged with three other agencies to form the DRUM Agency, with offices in Atlanta, Chicago and New York, and in April 2020 DRUM shut its doors. I'm still in touch professionally with people I worked with there.

ERE reports that peer verification is meant for members who can't verify through the usual routes. The catch is in LinkedIn's rules: whoever vouches needs their own verification at that same workplace. If my old colleagues never verified before the shutdown, they probably can't get one now, and a job like mine may still fall through.

What the rules miss is collusion. Nothing in the published rules stops a few real employees, verified at one company, from confirming a friend who never worked there, and no time threshold stops a ring that already passed it. Malwarebytes warns of a related risk, "carefully constructed networks of fraudulent profiles."

LinkedIn's FAQ says it "may use detection systems" to flag suspicious activity, and that confirmed abuse can cost a member their verifying privileges and the confirmations they already gave. It doesn't say how that detection works, or what its "trust-based" rules actually check.

A reference call with one question

A vouch confirms where you worked. Malwarebytes draws the line plainly: a vouch "confirms an affiliation; it does not rate the person's ability or recommend them." That sets it apart from an endorsement, the older LinkedIn feature that recommends a person's skills. The badge says you were there.

The vouch is the first question of a reference call, asked of thousands of people at once, with the rest of the call cut off. Did this person work here? Yes.

The questions a hiring manager picks up the phone to ask never come up: would you work with them again, what were they like when the deadline moved, who cleaned up after them? Those answers were always the useful part of a reference and the expensive part, because someone had to know the candidate and say something true. The vouch keeps the cheap question and scales it beautifully.

That leaves the question hiring needs answered sitting where it was. LinkedIn's own chief executive, Dan Shapero, said in September that employers are struggling to tell which applicants can do the work. A confirmed shared past has nothing to say about that.

Neither does the person confirming it, who gives the vouch cheaply. LinkedIn sends the prompt and a tap answers it. Their name and photo may appear on the profile, and misuse can cost them their verifying privileges, but nothing follows if a real colleague turns out to be weak at the job. What a signal that cheap is worth to the people relying on it is a separate argument, and a real one.

November puts the badge in front of the agent

The badge matters more this month because of where it's going. LinkedIn is rolling out Hiring Assistant 2, its recruiting agent, to its existing English-language customers in November, and HR Brew reports that the update brings "new signals for recruiters in the hiring process related to candidate quality," with member verification among them. LinkedIn's own words are more careful.

Its post says verification lets recruiters "know if the person behind the profile is real," and it pitches connected apps, which show what members have done in the tools they use, at "who's best qualified for the job." The badge still lands in one view beside the signals that are supposed to say whether someone's good, and it's going into recruiters' applicant tracking systems too.

That placement is the risk. A recruiter facing a pool of thousands leans on the agent to narrow it, and the agent leans on whatever signals it's given.

The badge was honest about what it proves when your former colleague tapped yes. Somewhere between that tap and a ranked shortlist, a fact about presence can start being read as a fact about quality, by software in a hurry and by a person trusting the software. How much weight Hiring Assistant 2 gives verification is something LinkedIn hasn't said.

The badge itself may not help whoever's reading it. Jackie Singh pointed out in July that one grey shield can stand for an ID check or a workplace, and you have to click to tell which. Whether a peer-confirmed job looks any different at a glance, I can't tell from LinkedIn's pages.

Reading the badge

Read the badge as what it is: several colleagues' word that someone worked where they say, from people verified there themselves, with no employer checking. That's worth having on a network where Gartner expects a quarter of candidate profiles could be fake by 2028, and the rules make it harder to counterfeit at scale. When you're the one getting the prompt, the question in front of you is whether they worked there, and you can answer it honestly in a second, knowing your name may show up beside it.

When you're the one hiring, the badge has answered the easy question for you, and the hard one still needs a call to somebody who watched them work.