82 per cent of jobseekers who use AI have already changed their minds about an employer because of what ChatGPT or Gemini told them, even before opening a job vacancy.
That figure comes from research by PerceptionX, May 2026, among jobseekers in seven countries. And it affects every employer, including yours.
The order has been reversed
In the past, a candidate would open Google, type in your company name and scroll through your vacancies. Now, they open a chat window and ask a question. Which employer suits someone like me. Where can I actually progress. Is this a good company to work for.
Almost all AI users among job seekers research companies, roles, and applications in this way. Two out of three do so before deciding whether to apply at all. So the conversation about you as an employer has already started. You just aren't part of it.
The answer mentions two or three
It isn't one candidate who searches like that. It's becoming the standard way. Google is shifting its searchers step by step this year from ten blue links to an answer it writes itself. The major update for that rolled out this summer in the United States. The search engine is becoming an answer engine. And an answer engine doesn't name ten employers. It names two or three.
85% from your story appears in places you don’t control
Here is where it gets uncomfortable. 85 percent of what a language model thinks it knows about you as an employer comes from sources you don't manage yourself. Glassdoor. Indeed. Reddit. Old news articles. A forum discussion from 2021.
What is missing will be filled in. What is unclear will be simplified. And what is nowhere stated does not exist.
You have no control over those sources. You cannot reverse a review on Glassdoor, edit a thread on Reddit, or remove a 2022 article from the index. What you can do: ensure that the accurate story is so clear and so consistent in reliable places that a model cannot ignore it.
The error is already in there
58 per cent of those same jobseekers have caught AI providing incorrect information about an employer. The wrong salary. A culture that is long gone. A restructuring from three years ago, presented as the present. The model does not hesitate. It answers with the same calm as it does with a fact.
Silence is the most dangerous sign
I spent fifteen years in recruitment. The candidates you want most are precisely the ones who do their research beforehand. Who ask questions before investing their time. Exactly those people are now getting advice from a model. If that model says something outdated or wrong about you, they drop out without you ever seeing their name. No rejection, no application, no signal. Just silence.
That is the venomous part about it. An angry review, you see. A falling stream of job applications, you eventually see too. But a language model that politely and with conviction summarises you incorrectly remains invisible until you specifically ask for it.
That has one consequence for every employer reading this. The first impression of your company is now formed in a summary created by a model of what others say about you. Your recruiter, your campaign and your careers page only come after that. And that model increasingly determines which names are mentioned and which are not.
Two types of employers
For a language model, there are only two types of employers left. The employer with consistent, verifiable, quotable information about who they are and what it is like to work there. And the employer with an outdated careers site and otherwise a void that AI fills in as it sees fit. The first gets mentioned. The second never comes up.
The reflex is a new campaign. More content, a better-looking video, a fresh EVP. Understandable, and usually a waste of money. A language model relies on consistency and sources that corroborate each other. An expensive video changes nothing about that.
Make sure the story about you is present in the places where AI reads, in language that a model can adopt. Ensure your employees confirm the same image that you project outwardly, because their voice carries more weight than yours. And measure what is being said about you today in ChatGPT, Gemini and Perplexity, so that you know where you stand before a candidate tells you.
The only thing you can still control
It is still early. Most employers do not even know that this conversation is taking place, let alone what is being said in it. Whoever starts now is building a lead that will be hard to catch up with later.
What is still new today will be normal in six months. What is still an advantage today will be the minimum requirement in a year's time. And what still feels optional today will be the baseline in eighteen months.
Candidates are already asking AI about you. Today, as you read this. The only thing you can still determine is whether the answer is correct.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.