Answer Engine Optimisation Is a Strategy, Not a Tactic

The short version: AEO and GEO are not checklist problems. Every guide that hands you "10 tactics for answer engine optimisation" is describing the vehicle, not the destination. The destination is becoming a cited source — and that's an authority problem. You don't get cited by ChatGPT because you used question-format H2s. You get cited because the engine has enough confidence that you're the best answer to stake its response on.

The tactics matter. They just don't matter in the way most guides imply.

Why the tactic framing is backwards

Search engine optimisation has always tolerated a certain amount of gaming. You can buy links, cluster keywords, optimise page speed until you've squeezed an extra score point — and rank better than content that genuinely deserves to. The ranking signal set is large enough that indirect proxies work.

Citation signals are structurally different. When a language model generates an answer and decides which sources to surface, it's not running a PageRank variant. It's making a retrieval decision: is this source a reliable, specific, citable answer to this question? The model has no mechanism to reward keyword density or heading frequency — it's reading your content and assessing whether it earns citation.

This changes the game in one important way: you can't buy your way in. You can structure your way in, and you can earn your way in, but you cannot optimise your way in without the underlying authority. The tactics people publish — FAQ schema, plain language, question-format headings — are correct as expressions of authority. They fail when applied as a substitute for it.

The three elements of citability

I've been building content systems intended to earn AI citations, and the pattern I keep returning to has three elements.

1. Entity clarity

Before an AI system will confidently cite you, it needs to know who you are. Not in the abstract — in the structured, verifiable sense. Is your brand name unambiguous? Does your entity appear consistently across the web (your own site, Wikidata, Wikipedia if applicable, third-party directories, press coverage)? Does the AI have enough signal to resolve "this source" to "this entity with this area of expertise"?

Entity confusion is underrated as a citation blocker. If your brand name is shared with several other things, or if your site mentions your name in three different forms, or if you have no structured entity data anchoring who you are — the model's confidence in citing you drops. It's not a ranking decision; it's a disambiguation decision. Resolving that means consistent name-address-phone data where relevant, Wikidata entity creation or maintenance, and deliberate cross-linking of your canonical identity across authoritative surfaces.

2. Factual authority

The second element is whether your claims are verifiable and your content earns trust. AI systems are essentially doing source quality assessment at inference time. Hedged, vague, or unsubstantiated claims get deprioritised. Specific, sourced, demonstrably correct claims get pulled.

This is the part most brands find uncomfortable, because it requires actually having something to say. "Comprehensive guide to X" with 2,000 words of general-purpose content doesn't earn citation. A page that says something specific, defensible, and ideally original — with the sourcing to back it — does. The AI needs to believe that if it cites you and a user checks, you'll hold up.

There's no shortcut here. The path is original research, cited data, primary sources, and specificity. The guides that tell you to "write authoritatively" are correct; they're just not explaining that authority is demonstrated through what you actually claim, not through tone of voice.

3. Topical specificity

The third element is being the best answer to a specific question rather than a reasonable answer to a broad category. Generalist content gets cited less than specialist content, because the retrieval problem the AI is solving is "which source is most likely to be correct about this?" not "which source seems generally credible?"

This has practical implications for content architecture. A single page that thoroughly answers one question — with enough depth that it's genuinely the best treatment of that question on the open web — is worth more for citation purposes than five pages that each address the question partially. The strategy that follows from this is narrowing, not broadening: identify the specific questions your brand is best-positioned to answer, and be definitively, specifically right about those.

Where structured data actually fits

FAQ schema, Article schema, structured headers — these are real, and they matter. But the reason they matter is misunderstood.

FAQPage schema tells crawlers and AI systems: "This page is structured as explicit Q&A pairs. The question is X and the answer is Y." That signal reduces the ambiguity about what the page is trying to say. It makes the citability decision easier. But it only works if the underlying content earns citation — the schema is surfacing the answer, not manufacturing the authority.

Article schema with datePublished, author, and about fields gives AI systems explicit context about what this content is, who wrote it, and when. That context feeds directly into the trust calculation. A fresh, explicitly-authored, topically-labelled piece of content is easier to cite than an undated, anonymously-written blob — even if the content itself is equivalent.

Structured data is the vehicle. If the vehicle is carrying something the model wants to cite, the structured data helps it get there. If the vehicle is carrying generic content, schema just makes the generic content more legible.

The GEO stack I actually use

When I build for AI citation, the layer structure looks like this:

llms.txt — A plain text file at your domain root that gives AI crawlers a direct signal about what's on your site, which pages are canonical, and what you want indexed. It's a proposed standard, not yet widely enforced — but it's a low-cost signal that costs nothing to add and is likely to matter more as AI crawling matures.

FAQPage + Article schema — On every page with a clear question-answer structure, implement FAQPage. On editorial content, implement Article with datePublished and author. Both. Not as an afterthought in a plugin but as a deliberate part of the page design.

Entity consolidation — Wikidata entry (or maintenance of an existing one), consistent entity name across all web presence, citations pointing back to canonical entity pages. If you have a Wikipedia page, the structured data there feeds directly into AI system knowledge. If you don't, Wikidata is the next best lever.

Timestamped, updated content — AI systems and their retrieval pipelines weight recency differently, but freshness is uniformly positive as a signal. Publishing dates, explicit update dates, and actually keeping content current all contribute.

Authoritative external citations in your content — Citing credible primary sources is not just good journalistic practice. It signals that your content exists in the same epistemic ecosystem as trusted material. A page that cites Nature, government sources, or well-established institutional research sends different signals than one that doesn't link out.

None of these is individually sufficient. Together, they form the infrastructure that makes the underlying authority legible to AI systems.

The one question

Every brand asking about AEO should start here: If someone asked an AI about your category right now, would your brand be cited?

If you don't know — run it. Ask ChatGPT, Claude, and Google AI Overviews a specific question your ideal customer would ask. See what comes back. If your brand isn't there, ask the next question: why not?

In most cases the answer is one of three things: the AI doesn't know who you are clearly enough (entity problem), the AI doesn't have enough confidence that your content is right (authority problem), or the AI can find a more specific answer elsewhere (specificity problem). The tactics all address one of these three. Starting with the tactics before diagnosing which problem you have is why most AEO effort underperforms.

The checklist content isn't wrong. It's just solving the wrong problem first.

---

I've written about what AI actually cites based on my own measurement — 3,050 answers, 11,700 citations, with a breakdown by engine and source type. If you want the data behind the strategy, start there.