GEO · 12 min read
Brand is the only moat left in AI search
Brand building now decides who AI assistants name. Castle's view, backed by Seer, 6sense, Ahrefs and Ofcom data, plus three things to do this quarter.

On this page — 12 sections
- TL;DR
- Table of Contents
- Why do AI assistants shrink every category to three names?
- What does buyer research say about brand and the shortlist?
- What does ‘brand’ mean to an AI model?
- Where do most businesses go wrong?
- Three things to do this quarter
- Our perspective
- Is your brand on the shortlist before anyone asks?
- FAQ
- Sources
- Recommended
Brand building is now the most reliable way to win in AI search. When ChatGPT or Google’s AI Overviews compress a whole category into three names, they pick businesses that people and machines already recognise. That’s Castle’s position, and the data backs it: cited brands get far more clicks, and buyers mostly choose names they knew before they started looking.
Ask an assistant for “the best commercial roofer in Manchester” and you won’t get ten blue links. You’ll get a paragraph and a handful of names. If you’re not one of them, you’re not in the conversation, however tidy your schema is.
Below is why we think brand has become a performance channel, what “brand” means to an AI model, where most businesses go wrong, and what to do before the quarter’s out.
If your business is invisible when AI assistants name the shortlist, that’s fixable. Our generative engine optimisation (GEO) team works on exactly that.
TL;DR
- AI answers shrink a category to a few names, so being recognised matters more than ranking tenth.
- In Seer Interactive’s 2026 CTR study, brands cited in AI Overviews got about 120% more organic clicks per impression than brands left out.
- In 6sense’s 2026 buyer report, the eventual winner was on the day-one shortlist 92% of the time.
- To an AI model, “brand” means consistent facts, mentions on sites you don’t own, reviews and a point of view people repeat.
- Treating AI visibility as a technical fix is the most common mistake we see. Google says there’s no special markup required.
Table of Contents
- Why do AI assistants shrink every category to three names?
- What does buyer research say about brand and the shortlist?
- What does ‘brand’ mean to an AI model?
- Where do most businesses go wrong?
- Three things to do this quarter
- Our perspective
- Is your brand on the shortlist before anyone asks?
- FAQ
- Sources
- Recommended
Why do AI assistants shrink every category to three names?
AI assistants shrink categories because their job is to answer, and an answer is short. A results page could show ten businesses and let you choose. A generated paragraph names a few, explains why, and moves on.
That matters because the audience is no longer niche. Ofcom’s Adults’ Media Use and Attitudes research, published in April 2026, found that 54% of UK adults now use AI tools such as ChatGPT, Copilot or Gemini, rising to 79% of 16 to 24-year-olds. On the local side, BrightLocal’s 2026 survey of 1,002 US consumers found 45% had used AI tools for local business recommendations, up from 6% the year before.
Seer Interactive has put a number on being named. It tracked 53 brands, 5.47 million queries and 2.43 billion organic impressions from January 2025 to February 2026. In its 2026 update on AI Overview click-through rates, brands cited in an AI Overview earned about 120% more organic clicks per impression than brands that weren’t cited on the same kind of results page.
Organic clicks per 1 million impressions, informational queries
Scale: 0 to 35,000 clicks
No AI Overview shown: about 33,500
AI Overview shown, brand cited: about 20,743
AI Overview shown, brand not cited: about 9,445
Two honest caveats. Being cited still leaves you about 38% short of a results page with no AI Overview at all. And Seer says plainly that it can’t claim causation, because higher-authority brands are more likely to be cited anyway. Frankly, we’d read that caveat as support for the argument: the brands that were already strong are the ones getting named.
What does buyer research say about brand and the shortlist?
Buyer research says the shortlist is mostly decided before anyone searches. 6sense’s 2026 B2B Buyer Experience Report findings, drawn from 3,596 completed purchases and 1,011 evaluations that ended without one, found the eventual winner was on the day-one shortlist 92% of the time. The vendor buyers favoured before speaking to a seller went on to win 82% of deals.
Here’s the detail that gets skipped in most LinkedIn summaries. The same report found 81% of buyers use large language models at some point during an evaluation, but only 26% use them to find new vendors. In its companion piece, The Never-ending Journey, 6sense goes further: LLMs were the least-used channel for staying current on a category, at 15%.
So AI isn’t conjuring up shortlists from nothing. Mostly, it’s checking, comparing and confirming the names buyers already had in mind. That’s why we think “how do we get into ChatGPT?” is slightly the wrong question. The better one is “why would anyone, human or model, think of us first?”

What does ‘brand’ mean to an AI model?
To an AI model, a brand is a cluster of consistent facts and repeated associations across the web. It has no feelings about your logo. It sees whether independent sources agree on who you are, what you do, and whether you’re any good. In practice that comes down to four things.
Consistent entity facts
Your company is an entity: a named thing with attributes like location, services, founders and sector. If your website says “Liverpool”, Companies House says “Wirral” and an old directory listing still carries a previous trading name, a model has three versions of you to reconcile. Some will reconcile it badly. We’ve covered the mechanics in our guide to entity SEO and making ChatGPT know who your company is, and it’s the least glamorous, most fixable part of this list.
Being discussed on sites you don’t own
Mentions on other people’s websites appear to matter more than links. In an Ahrefs study of 75,000 brands (published May 2025, updated April 2026), branded web mentions had the strongest correlation with AI Overview visibility at 0.664, against 0.218 for backlink count. Brands in the top quartile for mentions averaged 169 AI Overview mentions; the next quartile averaged 14. Ahrefs is clear that correlation isn’t causation, and so are we.
Where those mentions live matters too. The citation leaderboards shuffle every month, but the same few names keep turning up:
- Ahrefs’ November 2026 ranking of sites ChatGPT cites (US, all topics) puts Wikipedia first at 10.1% mention share and Reddit second at 6.2%, with YouTube eighth.
- Obsurfable’s October 2026 summary of Ahrefs Brand Radar data shows Reddit top on ChatGPT, Perplexity, AI Mode and Gemini in its September snapshot, and YouTube top on AI Overviews.
- Ahrefs’ live tracker of user-generated content in AI answers showed YouTube in nearly half of tracked AI Overview responses in mid-November 2026, Reddit in roughly one in eight.
- For B2B, Obsurfable’s top 50 most-cited domains across 5.7 million citations (September 2025 data) placed LinkedIn sixth.
None of these are places you can buy your way onto the front page. You get there by being worth talking about, which is the oldest definition of brand there is.
Reviews
Reviews are third-party evidence with a star rating attached, and AI tools read them in bulk. BrightLocal found AI tools are now the third most-used source of local business recommendations, behind Google and Facebook. It also found 88% of AI users fact-check the reviews an assistant cites, by checking the source or whether the review looks genuine. So a thin or stale review profile costs you twice: once when the model summarises it, and again when the customer goes to check.
A point of view people repeat
This one’s harder to measure, and it’s the one we’d bet on. Models repeat what the web says about you. If the only thing anyone can say about your firm is “a full-service provider of solutions”, there’s nothing to repeat. Imagine a Leeds accountancy practice that argues, loudly and often, that small firms shouldn’t pay for bookkeeping software until they’re near the VAT threshold. Agree or not, journalists, Reddit threads and podcast hosts now have something to quote, and the firm gets attached to that idea.
Where do most businesses go wrong?
Most businesses treat AI visibility as a technical fix. They add an llms.txt file, bolt FAQ schema onto every page, buy a tracking dashboard and wait. We get the appeal: it’s a ticket with a definition of done.
Google has been unusually blunt about this. Its guidance on AI features and your website says there are no additional requirements to appear in AI Overviews or AI Mode, no need for new machine-readable files or AI text files, and no special schema to add. Crawlability and clean structure still matter, and you shouldn’t block the bots you want reading you. But they’re the entry fee. They decide whether you can be considered, and they say almost nothing about whether you’ll be chosen.
Think of it like a shop window in a market town. Washing the glass helps. If nobody in town has ever mentioned your shop, a cleaner window won’t change who gets recommended at the school gate. We go into which tactics deserve the budget in the GEO tactics most agencies miss (and the ones they oversell).

Three things to do this quarter
You can’t build a brand in twelve weeks, but you can stop leaking it and start earning mentions. These are the three moves we’d make first for an established UK business.
1. Run a facts audit and fix the contradictions
List every place your business is described: your site, Google Business Profile, Companies House, LinkedIn, Wikipedia or Wikidata if you have entries, the main trade directories and the review sites. Check name, address, founding date, services and leadership. Fix every mismatch you control and request corrections for the rest. Then ask ChatGPT, Gemini and Perplexity to describe your company and note what they get wrong. Our 40-point AI visibility audit checklist covers this in more depth.
2. Get discussed somewhere you don’t control
Pick two third-party venues that matter in your category and commit to showing up properly. That might be a trade publication you can pitch original data to, a subreddit where your customers ask for advice, a YouTube channel that reviews kit in your sector, or a LinkedIn presence from named people rather than a logo page. One genuinely useful piece of research or a strong opinion, placed where your buyers already read, will do more than twenty blog posts on your own site. Our guide to how LLM citations work and how to earn them explains why.
3. Make reviews a routine
Ask every happy customer, at the moment they’re happiest. Reply to every review, grumpy ones included, because a measured reply is also something a model reads. Track a monthly review target next to your sales figures; on this evidence, it is one.
Our perspective
Castle’s view is simple: brand is now a performance channel, and the businesses that keep treating it as a “nice to have” will spend the next two years wondering why their SEO reports look fine while enquiries dry up.
We don’t say that because tactics have stopped mattering. They haven’t. We still fix crawl issues, entity data and page structure for every client, and our AI SEO work starts there. But those fixes only pay off when there’s something worth retrieving. A technically perfect site for a business nobody talks about is a beautifully catalogued library that nobody visits.
The uncomfortable part is that brand building is slow and doesn’t fit neatly into a monthly report. We’d still rather say so than sell anyone a schema plug-in and a dashboard. The 6sense numbers point the same way as the Seer numbers and the Ahrefs numbers: the names that win were usually known before the question was asked.
If you want the wider picture first, our complete guide to AI SEO and GEO sets out how the pieces fit, and our read of whether AI really builds the B2B shortlist goes deeper on the buyer studies.
Is your brand on the shortlist before anyone asks?
If you’re not sure what ChatGPT, Gemini or Google say about your business today, that’s the place to start. We’ll show you, and tell you honestly what it’ll take to change it.
Require AI SEO, GEO, or digital marketing and AI automation within your business? Book a free appointment with a senior consultant.
FAQ
Does brand building really affect AI search visibility?
The evidence points that way, though nobody can prove cause. Ahrefs found branded web mentions correlated with AI Overview visibility at 0.664, far stronger than backlinks at 0.218. Seer found cited brands earned about 120% more organic clicks per impression. Both note that stronger brands are simply more likely to be cited.
Which websites do AI assistants cite most often?
Reddit, YouTube and Wikipedia dominate most leaderboards, with the order changing by assistant and by month. Ahrefs’ November 2026 ChatGPT ranking put Wikipedia first and Reddit second. YouTube leads on Google AI Overviews. LinkedIn ranked sixth in Obsurfable’s cross-industry top 50 from September 2025 data.
Do I need llms.txt or special schema to appear in AI Overviews?
No. Google’s guidance says there are no additional requirements, no AI text files and no special schema needed to appear in AI Overviews or AI Mode. Good crawlability and clear structure still help, but they make you eligible rather than chosen. Recognition and third-party evidence do the choosing.
How long does it take to build a brand AI assistants recognise?
Expect months, not weeks. Fixing contradictory facts can show up quickly once models and search indexes refresh. Earning mentions, reviews and a reputation for a clear point of view takes longer, typically two or more quarters of consistent effort. That’s why we suggest starting this quarter rather than next.
Sources
- Seer Interactive: AIO Impact on Google CTR: 2026 Update (April 2026)
- 6sense: 9 key findings from the 2026 B2B Buyer Experience Report (November 2026)
- 6sense: The Never-ending Journey (November 2026)
- Ofcom: Passive social media use, AI companionship and online side hustles: UK adults’ media and online lives revealed (April 2026)
- BrightLocal: Nearly half of consumers are asking AI for business recommendations (March 2026)
- Ahrefs: An analysis of AI Overview brand visibility factors, 75K brands studied (May 2025, updated April 2026)
- Ahrefs: The 50 most-cited websites in ChatGPT (November 2026)
- Ahrefs: UGC AI visibility trends (live tracker, accessed November 2026)
- Obsurfable: Top domains cited by LLMs, October 2026 (October 2026)
- Obsurfable: AI’s most cited domains, top 50 (February 2026)
- Google Search Central: AI features and your website (updated December 2025)


