82 percent of jobseekers who use AI changed their minds about an employer because of what a language model said. A chatbot talked about your company, the candidate listened, and you were not there.
That figure comes from research by PerceptionX, published in May 2026, among 306 job seekers in seven countries. 96 percent of AI users deploy those tools to research employers. 72 percent do so before deciding to apply.
So the judgement is made before the application even exists.
Then vs now
In the past, a candidate looked at your careers site and formed an impression there. Now, she asks ChatGPT a question and gets back an answer in which you haven't written a single letter. 89 per cent use ChatGPT, 65 per cent Gemini. Those are the places where the conversation about you is taking place.
And that conversation is about something. Candidates ask about salary (53 per cent), about career progression opportunities (52 per cent), about how the interview process works (51 per cent). Precisely the topics that the average careers page remains silent on.
This is where it gets awkward. 58 percent of job seekers have caught AI providing incorrect information about an employer.
The model talks about you. And it often talks wrongly about you. The candidate who doesn't spot the mistake takes it as truth, and you never get the chance to set it straight.
A language model doesn't just make things up out of nowhere. It builds its answer from what it finds about you: your site, your reviews, your old job postings, a forum thread from 2023, a press release nobody reads anymore. What's missing is filled in. What's unclear is simplified. And what isn't anywhere simply doesn't exist.
In doing so, the model chooses the source that is easiest to understand and repeat. A scathing Glassdoor review from three years ago thus beats a vague careers page from today. Clarity beats timeliness, even if the clear source is long out of date.
So if your story about salary, culture and career progression is nowhere to be found clearly, the model takes the first source that fills that gap. Rarely is that the source you would choose.
The shift
For employer branding, what you manage shifts accordingly. For years you managed the feeling: the campaign, the tone of voice, the career story. Now you also manage the facts that a machine can retrieve and repeat. Salary range, career progression paths, what an application process is really like, what employees actually do in a day. Vague still sounds quite nice on a billboard. To a language model, vague is the same as absent.
As far as AI is concerned, there are only two types of employers left. The employer with clear, citable information about working for them. And the employer that leaves the model to guess.
The first one appears in the answer. The second one does not appear in it.
And this is going to become a bigger problem, not a smaller one. Seventy-seven percent of jobseekers want to let an AI agent search for work. An agent that compiles a shortlist itself and filters out the rest before a human sees anything. If you are not in that answer, you are not on the list. You are not rejected; you are simply not mentioned.
And it doesn’t stop at the application stage. A further 33 per cent use AI after receiving an offer, to decide whether to accept it. The model is involved from the initial exploration right through to signing the contract, and at every stage it can make or break your story.
It is invisibility that you cannot measure. The candidates you never considered appear in no dashboard, in no funnel, in no conversation with your hiring managers.
Recruiters know tightness. Too few applicants for a difficult role, you see that and you steer by it. This is a different tightness. The candidate who compared you to three others and chose someone else based on a wrong answer doesn't call you to explain. You can't read that silence anywhere.
The easy reaction now is to shout that it's all not going to happen that fast. That candidates will still come to your site anyway. Some will. But 74 percent regularly use AI for employer research, and that share is growing every year. You can wait until it overtakes you, or you can determine today what the model thinks of you.
What can you do now?
Start simple. Now ask ChatGPT, Gemini and Perplexity what they say about you as an employer. Read the answer as a candidate, not as a marketer. Is it accurate. Is it complete. Would you apply based on this answer.
Where the answer is shaky, there lies your work. Fancier wording won't fix that. The model needs clear, verifiable information in the places where it finds and dares to quote it.
The candidate isn't asking you anymore. They're asking AI. The only question that matters is whether AI gives the right answer.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.