In 85 out of 100 subjects, not a single brand has control over the AI answer yet.
That comes from a study by Semrush and Kevin Indig. From January to June 2026, they tracked 1,094 categories in ChatGPT every month. 50,000 brands, 220,000 domains, 600,000 citations. One underlying question: who owns the answer when someone puts a category to ChatGPT?
Hardly anyone.
In 15.2 percent of the categories there is a clear owner. In 31.2 percent an emerging leader. The rest, 53.7 percent, is wide open. No brand steering the conversation, no name that rolls off the tongue naturally.
What can you do as an employer?
For you as an employer, that is an open goal. “Which healthcare employer suits someone like me” is precisely that kind of category. A few prompts, a handful of names that keep coming back, the rest that fall away. And in most sectors, that top row still belongs to no one.
Kim is already asking. She no longer types a job title into a search bar; she submits her career question to a model and reads the answer that comes back. That answer names two, three employers. You hope one of those names is yours.
Here lies the part that most executive boards miss. Being mentioned and being recommended are two different things.
The difference between being mentioned and being recommended.
Semrush found that only 21 percent of the most cited domains coincided with the most mentioned brand. The two signals even diverged slightly. You can therefore crop up everywhere as a source and still never be the name that the model recommends. A link in the footnote is not a place on the shortlist.
For recruitment, that's the whole difference. A candidate doesn't read a reference list. She reads a few names with a reason attached. If you're in the references but not among those names, you don't exist to her.
A second study, by BuzzStream, makes that crystal clear for your exact situation. With list questions, the type “which companies suit X”, only 7.2 percent of the mentioned brands also received a citation. And around 80 percent of all citations were on third-party sites, not on your own domain. Your story is therefore largely shaped in places you do not manage yourself: reviews, trade media, an old interview with your director. Your own careers site is just one voice in that choir, and by no means the loudest.
And you don't buy that place with the tricks that used to work.
The researchers compared their category owners against classic SEO metrics. More organic traffic, a higher authority score, more brand searches: none of these are hardly predictive of who owns the AI answer. Kevin Indig dryly summarised it: the old SEO yardstick does not explain who owns a subject.
Your head start over Google doesn't count here.
That means your head start over Google won't get you far here. The employer with the largest careers site and the biggest advertising budget is on the same footing as the smaller competitor who actually has their story consistent and quotable.
There is a second finding in the data, and that works in your favour.
The very largest, busiest subjects least often have an owner. 11.3 percent with high demand, compared to 19 percent with lower demand. The categories where the most candidates search, “best tech employers”, “where do you enjoy working in logistics”, are therefore most often still owned by no-one. The larger the playing field, the emptier the front row.
And whoever takes that place once, keeps it.
Category owners with a solid lead remained in first place in 90.4 percent of the monthly measurements. Whoever holds the answer rarely loses it to the next one who comes along. The first to get there builds a lead that becomes harder to catch up with month by month.
What is still an empty response today will be a habit in six months. What is a habit in six months will be a reputation in a year. And once something is a reputation in a language model, you can no longer catch up with a campaign.
What does that mean in practical terms.
Choose the subject on which you want to be found and make it small enough to own. “Good employer” is too big to claim. You can, however, own “Employer for data engineers seeking autonomy in the North of the Netherlands”. A narrow category that you win beats a broad one in which you drown.
Make sure your story sounds the same everywhere. On your own site, in trade media, in interviews, on the review sites the model draws from. A brand that says three different things in three places won't be chosen. What contradicts itself is skipped over. What is consistent nowhere, doesn't exist.
And see if you win that spot. Ask your candidates the same questions every month, on every platform they use, and see if your name comes up or sinks. Next year's owner is the employer who starts now, while the queue is still empty.
Most categories do not belong to anyone yet today. That will not stay that way.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.