91 percent of TA leaders say that AI determines how candidates research their next employer. 33 percent dare to say that the picture painted by AI is also accurate.
That gap, 58 points, is the whole story.
The figures come from Built In's 2026 report on employer reputation, based on 162 TA and HR leaders at mid-sized to large organisations. Almost everyone sees that search behaviour has shifted. One in three trusts that their own employer will come out of it well. And 73 percent are concerned that candidates are getting outdated or incorrect information about them via AI.
Three in four employers know that AI says something about them, and suspect it is incorrect.
It is no longer a niche where this happens either. ChatGPT sits at around 900 million weekly users, and together with Claude, Gemini and Perplexity, that group now captures the vast majority of all AI referrals. That is the place where the first impression is made.
This week, that realisation became commercial. On August 5th, Symphony Talent, one of the big names in employer-branding technology, launched a product called AI Search Visibility Solutions. It measures how an employer appears in ChatGPT, Gemini, Claude and Perplexity, and helps adjust that image. Four components, ranging from a one-off benchmark to ongoing monitoring.
Michael Drayer, their SVP AdTech & Media, captured the essence in a single sentence: “AI search changes the equation, because a candidate can form an image of an employer without ever visiting the careers site.”
I know that sentence. I've been saying it on stages for a year.
What happens when an organisation builds something like that is simple. The category is no longer speculation. When I started talking about Generative Engine Optimization for employers, the standard reaction was “it's not happening all that fast”. A major agency building a dashboard for it now shuts down that conversation.
And yet a warning is in order.
A Visibility Index measures the problem. Measuring is the easy part. You can know precisely that you are at 12 per cent visibility and still not know why, let alone how to get to 90. The dashboard tells you that you are invisible; it does not make you citable.
Becoming quotable is the job. That's technology the crawler can access, structure a model can summarise, and content that's accurate without anyone needing to click through. Anyone who skips that and just stares at the meter is buying insight into their own invisibility.
Look at what a candidate is actually asking. Not “software engineer vacancies Amsterdam”, but “which tech employer suits someone looking for peace and autonomy”. The model then reads what it can find about you, weighs it, and mentions two or three names. If your story isn't clear and verifiable, your name simply won't come up.
That is where the real shift lies. In the past, a candidate looked at your vacancy, read your careers page, and then formed an opinion. Now, she asks a model, gets a single answer back, and the shortlist is finalised before you even know you were in the running.
What is still a dashboard metric today, will be a selection criterion in six months' time. What still feels optional today, will be the baseline in a year's time. And what can still be your edge today, will be the minimum everyone has in eighteen months' time.
Every employer will soon be measuring; another ten tools that give you a score are arriving this year. The distinction lies one step further: either you become the answer, or you read the report.
We saw that difference. On ProRail’s jobs site, the average AI visibility grew to 71.8 percent across five platforms, with less than 1 percent negative sentiment. At Independent Recruiters, visibility reached 4.3 times that of the nearest competitor. Those are not figures that stay in an index; that is ground that has been won.
The crazy thing about those 58 points is that they are a gift. As long as the majority does measure and does not act, the answer remains open. The model will soon name two or three employers per question. There is room. For now.
Symphony Talent are right that this counts. I just wouldn't wait for your own supplier to build an extra tab. By the time the insight is standard, the visibility will already have been divided up.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.