OpenAI is building its own job board. It is due to arrive in mid-2026, will be led by Fidji Simo, and will go head-to-head with LinkedIn. The difference lies in the engine. Where LinkedIn revolves around CVs and job titles, this model matches based on skills and demonstrable AI knowledge.
A model will soon make that match. It chooses based on what it knows about your organisation. And that is where the shoe pinches, because to that model, most employers are virtually invisible.
This is not an isolated incident. The search behaviour has been moving in the same direction for months.
The change in search behaviour
In the first four months of 2026, 68% of all Google searches ended without a single click, according to research by SparkToro based on Similarweb clickstream data. AI Overviews appear at the top of more than 20% of searches and halve the number of clicks as soon as they appear. Google’s AI Mode surpassed the 1 billion monthly users mark at I/O 2026. ChatGPT has around one billion users per week.
Candidate Kim is no longer searching, she is asking. Which employer suits someone like me. Where can I genuinely progress. Which companies have a culture that fits. The model answers with three, maybe four names. The rest fall out of the picture.
Two types of employers
For a language model, there are now only two types of employers. The employer that the language model can read, understand and recommend. And the employer that the language model cannot find, and therefore does not mention either. Between those two there is no grey area. You are in the answer, or you do not exist there.
Think about what that does to your first impression. For years, your careers site was your calling card: that is where candidates landed, that is where they formed an impression. Now, the model forms that impression, often long before anyone even gets close to your site. The candidate reads a summary about you, written by a system you have never spoken to, based on sources you never chose. And you rarely get to correct a first impression.
For recruiters and employer branding teams, that changes the arithmetic. The budget that goes towards your careers site, your campaigns and your content now competes with an answer the model already had ready. You can have the best 'work-for-us' page in the Netherlands; if the model doesn't fetch you, you're talking to an empty room.
How does that work with OpenAI's job board, then?
On OpenAI's job board, that distinction becomes stark. A certificate via OpenAI Academy will soon make the candidate readable to the system; the goal is 10 million certified Americans by 2030. The underlying question is what makes you, as an employer, readable. A skills-matching model has to source from somewhere what your organisation stands for, who works there, and why someone would fit in with you.
The fact that OpenAI is building this while Microsoft, its biggest backer, owns LinkedIn, says everything about what is at stake. Two giants fighting for the same spot: that first moment when a candidate and an employer find each other. That moment is shifting from a search result to an answer. And within that answer, there is room for only a handful of names.
What can you do about this change?
What today still seems like a fringe phenomenon will be commonplace in six months. What today is still an experiment will be the primary point of contact in a year. And what today is still your choice will, in eighteen months, be the place where the candidate encounters you for the first time, or misses you entirely.
The good news is that this is measurable and can be managed. In one of our own assessments, an employer brand scored 7.5% for visibility in response to a niche exploratory query. After getting the basics right, that figure rose to over 90%. Same query, same model, a different result.
There is work to be done, and it is concrete.
First, make sure AI crawlers are allowed in. An empty robots.txt, slow pages, or content that only appears after a JavaScript layer: what the model cannot retrieve, does not count. This is the most boring step and the most decisive.
After that, write content that a model can cite. A clear question with a clear answer, paragraphs that also make sense independently of the page, figures with a recognisable source underneath them. Vague employer branding language will not get a single model out of your site; those words are precisely the kind of noise that a model filters out.
Finally, build onto your entity. What is said about you elsewhere, in trade media, on review sites, in interviews, and on the site of your own people, determines whether the model recognises and trusts you. A brand that only talks about itself on its own website remains a rumour to AI.
OpenAI’s careers site is not yet open. Your preparation can be.
Ask ChatGPT, Gemini and Perplexity today which employers they recommend in your sector, for your type of role. If your name isn't there, you know precisely how much time you have. Until mid-2026.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.