How to Build Topical Authority That AI Recognises

Topical authority means covering one subject so completely and so credibly that both search engines and AI models treat you as a go-to source on it. Here is how to build topical authority: pick a narrow subject you can genuinely own, publish deep connected pages on it, and get talked about off your own site. Coverage alone is no longer enough.

This is an old idea, and Google never called it a ranking factor

Topical authority has been in SEO conversation for years. Instead of one page about a big subject, you write twenty covering every corner of it and link them together. The bet is that a search engine notices the pattern.

Google has never confirmed this as a lever you can pull. In 2023, John Mueller of Google was asked about it on Reddit and agreed with a user who said it was just a new name for relevance and authority. His advice was to not worry about it.

The idea spread anyway, because it is a useful way to plan content. That is different from being a dial in an algorithm. Keep the difference in mind. It matters for what comes next.

What actually changed

An old search engine ranked pages. An answer engine picks a source, then writes a sentence about it. Different job.

Jason Barnard of Kalicube put the split well in Search Engine Land this year. Topical authority describes what you have built, he says, while topical ownership describes whether the system picks you. He argues the entity, meaning the company or person behind the content, is now what the other signals hang off.

Andrew Holland argued in Search Engine Land in May 2026 that brand authority beats topical authority in AI search, and that authority comes from becoming a recognised source elsewhere, not from what you publish on your own site.

Both sound right. Both are arguments, not data. Treat them as a way of thinking, and read the studies before you spend money.

The gap that should change your plan

Here is the number that reframes the whole subject.

Amit Prakash Sharma published a study on arXiv in January 2026 covering 112 startups from the Product Hunt top 500, running 2,240 queries through ChatGPT and Perplexity. Asked about a product by name, ChatGPT knew it 99.4% of the time. Asked an open question, like which tools launched this year are best, ChatGPT named those same products just 3.32% of the time.

The model knows you. It just does not bring you up.

Two more findings are worth your attention. Products that had done visible AI optimisation work did no better than products that had done none. And on Perplexity, what did correlate with being surfaced was ordinary: referring domains, Product Hunt ranking, and a community presence on Reddit.

Sharma's own conclusion is blunt. Do not optimise for AI discovery directly. Build the search foundation and AI visibility follows.

Entity SEO, in plain words

Entity SEO sounds technical. It is not. An entity is something the machine keeps a record for: your company, your founder, your product. Entity work means making every record of you agree with the others.

Same company name everywhere, spelled the same way. Same one-sentence description on your site, your LinkedIn page, your listings, your press coverage. Same category. Same founder name attached to the same company. If three sources describe you three different ways, the machine has three weak records instead of one strong one.

That is most of it, and you can do it in an afternoon.

Now the honest part. A critical survey of generative engine optimisation research published on arXiv in July 2026 by Olivier Martinez looked across the published tests and found only two factors held up repeatedly: whether the page genuinely answers the question, and where it sits in the set of documents the engine retrieved. An authoritative tone showed weak and unstable effects. So the practitioners saying entity work is the new foundation and the replication evidence saying relevance is what holds up are not saying the same thing. Nobody has reconciled that. The evidence on backlinks and brand mentions, next, shows a similar split.

Backlinks vs brand mentions

One stat shows up in almost every post on this: brand mentions correlate with AI visibility at 0.664, backlinks at only 0.218. It is a real number. It is also applied far more broadly than the study supports.

It comes from Ahrefs. They analysed 75,000 brands (Domain Rating above 40) and ran Spearman rank correlations between several metrics and AI visibility. Branded web mentions came out at 0.664, backlinks at 0.218, URL rating lower still at 0.18 (Ahrefs). Large sample, open method, worth taking seriously.

The detail that gets dropped: the study measured Google's AI Overviews, not ChatGPT. A related Ahrefs analysis broke it down by platform. Branded mention correlation was strong for AI Overviews (about 0.65), moderate for Perplexity (about 0.30) and much weaker for ChatGPT (about 0.15). So the popular "three times" framing is mostly a Google AI Overviews finding applied to ChatGPT by extension. Ahrefs adds the caveat itself: correlation isn't causation, and improving these metrics won't automatically boost your AI visibility.

Why might mentions beat links at all? Nobody inside OpenAI or Anthropic has published the mechanism. Anthropic's own engineering writeup on retrieval describes embeddings, word matching and reranking, with nothing about how often a brand is mentioned across the web. The most interesting theory comes from John Lovett at Seer Interactive in March 2026: the model first recalls which brand to recommend from what it already knows, then finds a source to back it up. His supporting figure: when a brand is named in an answer, its citation rate runs around 53%; when it isn't, roughly 11%. If that holds, links, which shape a link graph rather than a model's familiarity with your name, would count for less here than in classic search.

Review sites follow the same pattern. Kevin Indig, writing on G2's own Learn Hub, analysed 30,000 AI citations across 500 G2 software categories. A 10% rise in review count went with roughly a 2% rise in citation share, explaining less than 2% of the variance. Separately, a DerivateX study of 40 B2B software categories and 188 cited sources, reported by AiThority in June 2026, found G2 and Capterra got zero direct citations in every category tested. A small, unbranded consulting blog in that data out-cited Forbes, Reuters and Gartner combined. Review platforms look more like a gate you clear than a dial you turn.

None of this makes backlinks pointless. A page has to be crawled and indexed before it can be cited, and none of these correlation studies account for pages that were never discovered. Backlinks still help with that.

What it means for you: brand mentions seem to matter more than backlinks, but by how much depends on the engine. Being talked about consistently across the web beats only being linked to. And because these studies disagree, the only way to know what moves your category is to track your own citations, not borrow someone else's 75,000-brand average.

How to get a Wikipedia page for your business

You probably cannot, and that is the real answer.

Wikipedia's rules for companies ask for significant coverage in reliable secondary sources independent of you, and say a single independent source is almost never enough. Then they list what does not count. Press releases and anything based on them. Anything you published. Your own website. Advertising. Quotes from your staff. Routine notices about hiring, funding, opening a location or changing a product line. Directory listings.

Hold that list against your own press coverage. For most small businesses almost all of it falls in the excluded pile, funding announcements especially. The guideline says plainly that self-promotion and product placement are not routes to qualifying for an article. You can check it yourself at Wikipedia's notability guideline for organisations.

You are also not meant to write it. Wikipedia asks paid editors to disclose and discourages writing about your own company. Agencies selling you a page are selling you something that often gets deleted.

There is also a trap in the research that nobody seems to be flagging. A June 2026 arXiv study by Pratyush Kumar, who works at the AI visibility vendor Ranqo, tracked over 100,000 prompt responses and found global brands in 72.9% of relevant answers, mid-market brands 43.6%, small brands 11.4%. People quote that as proof you need a Wikipedia page. But read the method. Kumar defined his top two tiers partly by whether the brand already had a Wikipedia article. The study cannot show that a Wikipedia page causes visibility, because Wikipedia presence is baked into how the tiers were built.

In that same vendor study, Wikipedia came fourth among non-corporate cited sources, behind YouTube, Reddit and editorial media. Around 78% of citations went to company websites. Meanwhile Wikimedia reported human pageviews down roughly 8% in 2025 against 2024, which it linked to AI and search answering questions directly.

So the honest answer on how to get a Wikipedia page for your business: qualify first by being genuinely covered, and expect less from it than the pitch suggests.

How to build brand authority when you are not Stripe

If you want to know how to build brand authority in a world where size wins, first check whether your sector is actually locked.

Semrush released an expanded AI visibility index in June 2026 covering 126 million US prompts from January to April across ChatGPT, Gemini, Google AI Mode and AI Overviews. Semrush sells software in this market, so treat the framing carefully, but the sector split is useful. In news and media the top three brands held 82.9% of visibility. In finance it was 41.4%, in industrial 42.2%. Some categories are closed. Others are wide open.

The same study found ChatGPT cites around 15 sources per answer while Gemini cites around 3. If your sector is crowded and your buyers use Gemini, you are fighting for three slots.

What to do this week

Pick a subject narrow enough that you could plausibly be the best source on it. Not marketing. Not even email marketing. Something like email deliverability for Shopify stores.

Write pages that actually answer the questions in that subject, including the awkward ones your competitors avoid. Relevance is the one thing the research keeps confirming.

Clean up your records. One name, one description, one category, everywhere you appear.

Then be present where the citations come from: forums, review roundups, best-of lists, the podcasts and videos in your niche. That part cannot be done on your own website, and it is where the data keeps pointing. If you want help mapping which of those surfaces your category actually uses, that is what we do at Aquinel (then GeoCited).

Sources

• Google On Topical Authority: Don't Worry About It (reporting John Mueller), Search Engine Journal, November 2023.
• Why topical authority isn't enough for AI search (Jason Barnard), Search Engine Land, 2026.
• Why brand authority beats topical authority in AI search (Andrew Holland), Search Engine Land, 4 May 2026.
• The Discovery Gap: How Product Hunt Startups Vanish in LLM Organic Discovery Queries (Amit Prakash Sharma), arXiv 2601.00912, 1 January 2026.
• Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization 2023 to 2026 (Olivier Martinez), arXiv 2607.14035, 15 July 2026.
• An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied), Ahrefs. https://ahrefs.com/blog/ai-overview-brand-correlation/
• Google Seems More Biased Towards Big Brands Than ChatGPT and Perplexity, Ahrefs. https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search/
• Contextual Retrieval, Anthropic. https://www.anthropic.com/engineering/contextual-retrieval
• John Lovett on brand mentions and citation rates, Seer Interactive, 24 March 2026. https://www.seerinteractive.com/
• Do More G2 Reviews Mean More AI Visibility? (Kevin Indig), G2 Learn Hub. https://learn.g2.com/do-more-g2-reviews-mean-more-ai-visibility
• Unknown Blogs Beat Gartner and G2 in ChatGPT's Software Recommendations (DerivateX study), AiThority, 19 June 2026. https://aithority.com/saas/unknown-blogs-beat-gartner-and-g2-in-chatgpts-software-recommendations/
• Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines (Pratyush Kumar, Ranqo, vendor), arXiv 2606.20065, 18 June 2026.
• Semrush Releases Expanded 2026 AI Visibility Index, Analyzing 126 Million AI Search Prompts, Semrush (vendor), 26 June 2026.
• Notability (organizations and companies), English Wikipedia guideline, read September 2026.
• New User Trends on Wikipedia, Wikimedia Foundation, Diff, 17 October 2025.