53.7 per cent is still outstanding
More than half of all categories in AI search engines do not have an owner yet. 53.7 percent, to be precise.
Semrush tracked over 50,000 brands across 1,094 categories in ChatGPT from January to June 2026. In more than half of those categories, not a single brand appears in even three out of the five answers. Nobody wins. The spot is wide open.
For employers, that is this year's most important figure.
Employer branding is pre-eminently such an open playing field. Most employers have built their story for Google and for the candidate who is reading it themselves. They have never factored in a model that quotes. As a result, the door remains wide open in almost every sector.
“The best employer in healthcare.” “A good IT company to work for.” “Which agency suits a strategy consultant who wants to leave the Big Four.” Those are precisely those kinds of categories. And candidates no longer ask Google those questions. They ask a language model, long before they ever reach your vacancies page.
Now for the figure that hurts.
Semrush investigated whether standard SEO signals predict who wins a category in AI. They don't. Brands that own a category only had higher branded search in 55.7 percent of cases. More organic traffic yielded a score of 48.4 percent. A higher Authority Score came out at 52.5 percent. Flipping a coin gives you almost the same prediction.
The biggest name on Google is not automatically the name that ChatGPT mentions. Anyone who spent years at the top of the search results can be completely missing from the AI answer. Your Glassdoor score, your advertising budget, your careers site with the nice photos: the language model sees none of it if it cannot cite you anywhere.
What does work then.
Semrush mentions three things. Broad substantive coverage of a subject, so that the model encounters you across the entire theme. Mentions in the sources the model draws from, so in places outside your own site. And clear entity signals: a consistent name, structured data, third-party citations confirming who you are.
In plain English: the model needs to know who you are, what you stand for, and it needs to see that confirmed in multiple independent places. One strong careers page won't cut it. It is about the sum total of what exists about you, and whether that whole picture adds up.
That is where the opportunity lies. If 53.7 percent of categories do not yet have an owner, then the space is still up for grabs. Not next year. Now.
And time is of the essence. From the same study: once brands own a category, they retain that top spot in 90.4 per cent of month-on-month comparisons. Whoever claims a subject continues to claim it. The lead grows naturally, because the model you mention today learns tomorrow that you are the answer. Whoever waits now will soon run into a competitor who can no longer be dislodged.
I see with clients what this means on the human side. A candidate no longer opens ten tabs. They get one answer, compiled by a model, and that answer is their first impression of you as an employer. If your story is not in there, you do not exist to that candidate.
Fifteen years in recruitment has taught me one thing. Candidates choose on emotion and justify with facts. That emotion now arises at the very first answer a language model gives, even before a recruiter or an advert comes into the picture. A conversation with a machine that may or may not drop your name.
The nasty thing is that you do not have complete control over that story. The model builds it from everything that can be found about you. Old reviews, expired job postings, a press release from three years ago. What is missing is filled in. What is unclear is simplified. And what is nowhere to be found does not exist.
What can you do?
Start measuring what the model is saying about you today. Literally, in ChatGPT, Perplexity, Gemini and Google's AI answers. Ask it for the best employers in your sector and read who it mentions. There is a strong chance you won't see yourself, even if you have a strong brand.
Afterwards, choose one category you want to own, one that your organisation is truly good at, and build the content and sources so that the model encounters you there time and time again. Ensure external sources confirm your story, because the model is more likely to believe the outside world than your own careers site. And repeat that, because the advantage you are building now is the advantage you won't have to defend next year.
What is still open today will be forgotten in six months. What is still cheap today will be expensive in a year. And what is still an advantage today will simply be the baseline in eighteen months.
Half the categories are wide open. The only question that matters is whether you are the answer to that, or someone else.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.