← All insights
GEO Fundamentals4 November 20257 min read

How AI engines decide who to cite (and how to be that brand)

AI engines are not magical. They are systems with rules — and you can read those rules in their behaviour. Here is what they reward, what they ignore, and how Aether shapes content to land on the right side of that line.

MH
Mo Hassan
Managing Director, Aether

The first time you watch a large language model answer a question, it feels like sorcery. It read something, somewhere, and decided to mention this brand and not that one. Why?

The honest answer is that nobody outside the labs knows the full picture. But after a year of tracking citations across six engines and 12,000 buyer-intent queries, the behaviour is far less mysterious than it looks. Engines do not roll dice; they are quietly applying a consistent set of heuristics. Once you learn to read them, you can shape content that lines up with what they reward.

The first heuristic is repetition across high-authority domains. If an engine sees your brand consistently mentioned, with the same description, across well-trusted sources, it will treat you as the canonical answer in that space. Repetition across the open web is the closest thing AI has to social proof.

The second is entity clarity. Engines do not retrieve raw text; they reason over entities. They need to know that you are a security company, that you serve prime central London, that you hold these accreditations, that you handle these property types. If your site is vague about who you are, the engine cannot place you in the question.

The third is freshness. Where Google can wait six months to recrawl, AI engines refresh their internal models far more often. A piece of content published last week, citing this year's data, with clear entity markup, is meaningfully more likely to be cited than a five-year-old guide ranking on page two.

The fourth — and most under-appreciated — is what we call the citation gap. AI engines often describe your category in detail without naming any specific brand. If you study where that happens, you find some of the highest-leverage opportunities in marketing. The reader is interested; the engine is open; nobody has filled the silence. That's the sentence your content has to earn.

Generative engine optimisation, in practice, is the discipline of working backwards from these heuristics. Brief content from real citation gaps, not from keyword volume. Make entity signals unmistakable. Refresh continuously. Earn external mentions, not just on-site pages.

The brands that operate this way show up in AI answers within a single quarter. The ones still optimising for ten blue links are running last year's race.

Want to know where AI cites your competitors instead of you?

Free 60-second audit. We’ll send a personalised visibility report within 24 hours.