Somewhere this week, an AI agent compared you to your competitors, formed a view, and passed a recommendation to a buyer who will never visit your website. You weren’t consulted. Neither was your brand book.
When an AI agent evaluates a company, it reads the raw text of your website, your structured data, and what everyone else on the internet says about you.
That should worry more brands than it currently does.
How to check
Take your homepage. Remove the hero image, the gradient, the logo, the animation. Read what’s left as plain text.
For most tech companies, what’s left is likely a sentence like “We help teams unlock their potential with intelligent solutions”. Which breaks down to not differentiate or mean much at all.
It contains no claim an agent can extract, no fact it can compare, no statement that distinguishes you from the four competitors it’s reading in parallel tabs.
Humans have always been generous readers. We fill vague copy with vibe, the confidence of the layout, the quality of the paper stock, the fact that someone we respect shared it.
Agents extend no such generosity. They are the first audience in the history of branding that takes you completely literally. If the words don’t say it, you didn’t say it.
The industry is busy debating schema markup and llms.txt files while the actual problem sits in plain sight: most brand copy, read literally, says nothing.
Where agents form their opinion of you
Yext analysed 6.8 million citations across AI-generated answers and found that around 86% come from sources brands either control or strongly influence: your own site, your listings, your reviews. Brand websites alone account for roughly 44% of citations.
However, Agents triangulate. Ahrefs and Surfer have both shown the same pattern: the more of an answer’s cited sources mention a brand, the more strongly that brand gets recommended.
Reddit sits near the top of the most-cited domains across every major model, which means a thread from 2023 where someone described your onboarding as “fine, I guess” is now part of your positioning.
Consensus is the new share of voice. An agent asked for the best tool in your category is looking for the name that keeps appearing when other people answer the question.
Distinctiveness was always about being describable, giving the market a specific thing to say about you. Agents have simply made the mechanism visible, and put a cost on failing it.
If nobody can articulate what makes you different, the model averages you into the category, and the category answer won’t include your name.
What to actually do
Limy analysed 500 million AI bot traffic events and found the major search crawlers almost never fetched llms.txt. They read your HTML directly. The real work is verbal.
Write one sentence about your company that is true only of your company. Not “we’re the leading platform for” — a sentence a competitor truly couldn’t say. Then check whether it appears anywhere on your website in plain, extractable text. Maybe it got rewritten in a review, buried in a case study PDF, or delegated entirely to a visual metaphor.
Put your facts in text. Pricing, integrations, who you’re for. Agents cannot read a claim that lives inside a visual. The brands doing this well read almost blunt — short declaratives, specifics where the category uses abstractions, the confident plainness of a company that knows what it is.
Make every surface tell the same story. Consistency matters more than volume: the same company description, the same category language, the same claims across your site, your listings, your founder’s LinkedIn, your G2 profile. Every inconsistency is noise in the model’s picture of you, and noisy entities get summarised cautiously.
Feed the consensus. A podcast appearance, a genuinely useful answer in a forum, a piece of original research other people cite. Agents triangulate from what everyone else says, and earned mentions are how you shape what everyone else says.
A caveat before anyone torches their homepage in pursuit of uniqueness. Some overlap with your competitors is necessary.
Category language is how an agent works out what you are; if you are a “project management tool” that refuses to call itself one, you’re opting out of every project management answer.
Shared vocabulary gets you into the room. The sentence only you could write is what gets you chosen once you’re in it. Keep both, and know which job each one is doing.
Pony Studio is the Emerging-Tech Brand Studio — a London-based branding and creative design agency specialising in strategic brand development for tech companies worldwide. If you’re building something bold and want a brand that moves at the same speed as your ambition, let’s talk.