The reference check has been reversed
82 per cent of jobseekers who use AI have seen their perception of an employer change because of a chatbot.
That figure comes from new research by PerceptionX, May 2026, among active jobseekers in seven countries. 96 per cent of them use AI to research employers. 74 per cent do so systematically. And 82 per cent say that an AI conversation has changed their opinion of a company as a workplace.
In the past, a candidate would call a former colleague. Now, she asks ChatGPT.
For ten years you checked the candidate: the CV, the gaps, the story behind the leap. Today the candidate checks you, via a language model you have never spoken to. Ten years in recruitment and employer branding revolved for me around one thing: a grip on the story. What the vacancy says, what the hiring manager says, what is on the careers site. That story had an owner. Now it no longer has one.
Kim asks the machine. And the machine is often wrong.
Let's call the candidate Kim. She types her question into ChatGPT. The tool used by 89 per cent of jobseekers who use AI, and even more often by the under-25s. She wants a verdict: is this a good employer for someone like me?.
And this is where it gets awkward. 58 per cent of those same jobseekers have caught AI providing incorrect information about an employer. More than half have heard the machine say something about a company that wasn't right. A salary from three years ago. A culture that existed two reorganisations ago. A growth path that was scrapped long ago.
And Kim usually doesn't spot that mistake. 72 per cent say that AI sounds convincing when it's wrong. 76 per cent don't even know that AI demonstrably gives biased career and salary advice.
The machine never hesitates. You do, if you're being honest. But Kim only hears the machine.
What is she asking about? Precisely the things you would rather answer yourself.
Salary (53%), career progression (52%), how the interview goes (51%), and whether she’ll really get ahead with you (49%). Exactly the questions you’d prefer to answer yourself, in a conversation, with the relevant context. Except that this conversation is taking place without you.
And it goes beyond research. 81 per cent of Americans would hand over at least one job-seeking task to AI. 67 per cent let AI search for vacancies. 67 per cent let AI do the preliminary research on the employer. Increasingly, it's not even Kim who reads your page, but her agent.
Your employer brand is the sum total of everything that is lying around about you.
A language model builds its answer out of what it finds about you. Your careers site, your job vacancies, your reviews, an old press release, a forum post from 2022. It gives equal weight to everything. It fills in the gaps.
The model has no ill intent. It chooses the most likely sentence based on what can be found about you. If your salary scale is not clearly stated anywhere, it grabs an old job vacancy archive or an estimate from a comparison site. That becomes your answer, purely because it was findable.
What is missing will be filled in. What is unclear will be simplified. And what is nowhere stated does not exist.
And this is growing. 65 per cent expect to lean more on AI for employer research over the coming year. 77 per cent would hand over the search for work partly or completely to an AI agent. 72 per cent already use AI before applying, so before you even know Kim exists.
The first impression of your brand is made in an answer that you cannot see and cannot correct. Unlike a review on Glassdoor, you cannot report or refute that answer. There is no button, no appeals procedure, no right of reply. There is only what the model knows about you, and whether that is correct.
Three things you can do this week
Write the answers to the real questions on your own domain. Salary indication, career progression, how your selection process works.
Clear your old tracks. That press release from 2022, the job advert with a job title you no longer use, the culture promise that no longer rings true. A model reads them all with equal seriousness.
See what AI is saying about you today. Ask ChatGPT, Gemini and Perplexity the question that Kim is asking, and read the answer as she reads it.
The latter is confronting and precisely why it is valuable. Most employers have never read what a language model says about them to a candidate. Kim reads it every day.
The candidate is no longer asking you. She is asking the machine. The only question that matters is whether the machine knows your answer.
Bryan Peereboom is the founder of AI Rebels. AI Rebels makes employers visible in language models.
Sources
- PerceptionX Research 2026, AI’s Impact on Employer Brand Perceptionperceptionx.ai/research/ai-candidate-usage
- 4 Corner Resources, AI in the Job Search 20264cornerresources.com/job-market-news/ai-job-search-trust-gap-2026