GEO starts with credibility, not format

Over the past few months, many communications professionals have been busy with FAQ blocks, schema mark-up and structured data. The reasoning is understandable: if AI search engines like ChatGPT, Perplexity and Google AI Overviews select sources based on recognisable structures, then you optimise for those structures. That's true, but it's only half the story. The other half determines whether you make the shortlist at all.

FAQ schema solves the wrong problem

The GEO market is mass-prescribing FAQ schema as the citation recipe. Logical, because Google itself recommends it. But AccuraCast analysed 9,000 cited sources in AI answers and found that FAQPage schema appeared in just 1.8 per cent of those citations. At the same time, an analysis of 804,000 AI responses shows that brands with active external mentions and editorial citations are cited in 75 per cent of AI answers, compared with 1 per cent for brands without those signals.

That gap is not marginal. It's the difference between being in the running and being absent.

Technical structure raises your odds at the margin. External credibility determines whether you're in the pool at all. Anyone who only tackles the first is optimising for a rule that isn't decisive.

AI search engines copy human judgement

The mechanism behind it is not new. ChatGPT, Perplexity and Google AI Overviews were trained on the web as it was, including the authority signals that people and search engines were already using: backlinks, editorial mentions, author profile, dateable publications. Those signals tell an AI system who is credible long before the system reads a page.

Ahrefs studied 75,000 brands and found that brand mentions correlate three times more strongly with presence in AI Overviews than traditional backlinks do. And 85 per cent of the brand mentions AI systems use to weigh authority come from external pages, not from your own domain. Content on your own site alone won't get you there.

Google formally codified this in December 2022 by adding Experience to E-A-T, turning E-E-A-T into the documented basis for quality assessment. In the September 2025 update, explicit criteria for AI Overviews were added to the same guidelines. The thinking behind it is consistent: trustworthiness is the most important element, and you don't build trustworthiness with an FAQ block.

What actually makes a page citable

Three things work structurally better than format optimisation alone.

First, a declarative opening paragraph. No question-and-answer construction, no definition that could apply to any agency, but a statement that stands for something. AI systems cite sources that take a position, because a position answers the user's question. A page that hedges cautiously and weighs every side gives an AI nothing to work with.

Second, an explicit author. Schema.org Article mark-up with a populated author field, datePublished and about is a signal AI indexers use to weigh sources. Not as a trick, but as evidence that a human stands behind it and vouches for it. Pages without those fields are skipped more often, even when the content itself is strong.

Third, dateable claims. An observation you can trace back to a specific moment, from a recognisable source, with a year attached. That's the difference between an opinion and a testimony. AI systems need testimonies in order to cite, because otherwise they have nothing to verify.

These are not GEO tricks. They're the hallmarks of well-written specialist content.

Branded searches are the exception

There's an interesting countercurrent running through the data. While informational searches have seen a hefty drop in click traffic because of AI Overviews (from 15 to 8 per cent click-through rate when an AI Overviews block appears), branded searches actually saw an 18 per cent rise in clicks after the rollout.

The explanation is simple: an AI can't replace a brand if the user already knows who they're looking for. Anyone already searching for you clicks through. Anyone searching for a category gets an AI summary and may disappear without ever seeing you.

That's the citation equivalent of brand awareness. A brand that doesn't stand for anything won't be asked about either. Travel blog The Planet D lost 90 per cent of its traffic after the AI Overviews rollout, not because the pages were technically poor, but because the information was generic enough that an AI could replace it. An opinion piece with a recognisable author and a dated position is harder to replace, because it stands for something.

Structure without foundations won't hold

The problem with treating GEO as a separate discipline is that it reverses the order. If you start with structure and only then think about what should go into it, you're building a house without foundations. The structure checks out, the content is empty, and an AI picks up on that.

A growing number of communications professionals report that AI-optimised pages without substantive depth are picked up temporarily, but quickly replaced by sources with more authority as soon as the AI model refreshes its index. The reason is structural: AI systems learn continuously. Anyone scoring on structure without substance today will lose ground tomorrow to those who have substance too.

What's more, a page that has been rewritten for AI but takes no position feels empty to a human reader. And a page that feels empty to people doesn't build external authority. People don't mention it, link to it or cite it. Which closes the circle: no external authority, no AI citations.

GEO and brand strategy are the same job

This is where it all comes together. The signals that convince an AI to cite you are precisely the signals that convince a human to believe you: you stand for something, you sign your name to it, you can be traced.

A brand that knows what it stands for, an author who signs off on it and a page that makes a sharp observation. That's both good brand work and good GEO. Not two separate projects with a shared dashboard, but one and the same job viewed from a different angle.

Treating GEO as a technical chore for the developer misses the point. Treating GEO as a reason to write more sharply, be more specific and take a position wins on both fronts: with the human who reads and the AI that cites.

The practical consequence is concrete. Give every page an author with a name and a profile. Take a position you can defend. Use dateable claims with a source. Write for the person asking the question, not for the machine processing it. And make sure others write about you, because 85 per cent of the authority AI systems weigh sits on other people's pages.

That isn't a new insight. It's what good communications work has always demanded. GEO just makes it more urgent.

---

Further reading

Online marketing: SEO, GEO and content optimisation Branding and brand strategy with Schwung