OpenAI is building its own recruitment platform. It is set to go live midway through this year.
A language model matches candidates with employers, based on demonstrable skills. Fidji Simo, head of the applications division at OpenAI, announced it last autumn. Large companies will soon be on it. Small local employers too. And governments.
One matching layer in between. That layer assesses you based on what the web finds out about you, far beyond your own job descriptions.
I understand that this feels a long way off for many employers. A launch, somewhere in America, sometime this year.
But the behaviour that platform plays upon is already here at the kitchen table.
Between June 2025 and May 2026, the major AI platforms received a combined 9.5 billion visits per month. Seventy per cent more than the previous year. Kim is among them.
Kim no longer asks Google which employer suits her. She asks ChatGPT. And the model replies with names.
A model only knows you through your sources
For a language model, employers come in two types.
One has consistent, verifiable information in places the model trusts. That employer is mentioned.
The other has outdated texts, scattered stories, a brand that says something slightly different in every place. That employer doesn't get a mention.
That is rarely down to the quality of the company. It is down to what the model can cite.
What is missing will be filled in. What is unclear will be simplified. And what is nowhere stated does not exist.
The story has been told, but the model doesn't read anything into it
Employer branding teams build beautiful careers sites. Films, employee stories, and a campaign that took months of work.
And then that same site blocks the AI crawler in the robots.txt. Or the strongest story is in a PDF that the model doesn't read out.
The budget is spent. The story told. The model reads nothing of it.
And the model looks wider than your own site. It weighs reviews, trade media, a profile on an external job board, an old interview with your director. Where Glassdoor returns a list of opinions, ChatGPT gives one answer. That answer determines whether Kim clicks through or asks about the next employer.
For a recruiter, that changes the game at the front end. Soon, you will be competing for one thing: being mentioned in a sentence. If your name isn't there, there is no second place to drop back to. You simply do not exist.
Employer branding provides the sources, a model reads them differently
Over recent years, many employers have invested heavily in employer branding and their visibility on Google. Good work. Precisely that builds the sources that a model draws from.
However, a model reads differently than a search engine. SEO aims for a blue link that someone clicks on. A model reads, weighs, and summarises into a single answer, without ever clicking. Your visibility then depends on one thing: whether your facts are consistent and traceable everywhere.
That is more concrete than it sounds. The same core message on your site, in trade media and in interviews. Figures on locations, roles and culture that never contradict each other anywhere. Structured data so that a machine accepts your facts without doubt.
A recruiter can already feel this. The candidate walks in with an image that a model has shaped, even before you've exchanged a single word.
The head start you never catch up with later
This is the work we do at AI Rebels. Generative Engine Optimization. We ensure that an employer is correct in the sources the model draws from, so that their name surfaces when a candidate asks who is a good fit for them.
Fooling the model doesn't work. A model that suggests a wrong match penalises itself and you along with it. The job is to make sure it's right: clear, verifiable, aligned.
What is still an experiment today will be a channel in six months. What is still a channel today will be a habit in a year. And what is still a habit today will be the place where the choice is made in eighteen months.
An OpenAI jobs platform only accelerates that. Put a model between candidate and employer, and the question shifts from “do I find your vacancy” to “does the model know you well enough to put you forward”.
The candidate of 2035 asks, compares and has a model advise them. The employers that already feature in those answers are building an advantage. Authority in a language model is accumulated over time; you cannot simply buy it off with a campaign in the quarter it comes out.
Start small. Ask ChatGPT today who a good employer is in your sector, in your region, for your type of role. See if you are on the list. There is a strong chance the answer will surprise you, and that the conversation is already happening without you.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.