A results page gives you ten doors. An answer engine walks through them, reads what is behind, and tells you what it found. That difference is the whole story.
What an answer engine does
When someone asks ChatGPT, Gemini, Perplexity or Copilot a commercial question, roughly this happens:
- Query understanding. The prompt is rewritten into one or more searchable queries.
- Retrieval. The system fetches candidate sources, often from a live web index.
- Grounding. The model reads those sources and composes an answer from them.
- Citation. It names or links the sources it leaned on.
Two consequences matter more than any tactic.
First, you cannot rank for a prompt. There is no position one. There is only whether you were retrieved, whether you were readable, and whether the model chose to use you.
Second, the answer is synthesised. Even when you are cited, the reader may never visit. The citation is the impression.
Retrieval, grounding, citation
Think of it as a funnel with three gates.
Retrieval. Can the system find your page at all? This depends on classic crawling and indexing, plus how well your page matches the rewritten query. Pages that answer a narrow question in plain language retrieve better than pages that circle it.
Grounding. Once retrieved, can the model extract a clean, factual claim? Dense marketing prose is hard to ground. A short, direct sentence with a number, a definition or a named entity is easy.
Citation. Will the model name you? This is where brand clarity and corroboration matter. If three independent sources describe your business the same way, the model has something to be confident about. If your site says one thing and your LinkedIn says another, it has noise.
AEO, GEO, SEO
The acronyms are new. The work is not entirely.
| Term | Stands for | What it targets | | --- | --- | --- | | SEO | Search engine optimisation | Classic results pages | | AEO | Answer engine optimisation | Being the direct answer to a question | | GEO | Generative engine optimisation | Being included in a synthesised response |
In practice they overlap. AEO is mostly question-shaped content and clean structure. GEO adds brand entity work and citation-building. SEO is still the substrate: no crawl, no index, no citation.
What makes content quotable
Answer engines lift sentences. Write sentences worth lifting.
- Answer first, then explain. Open a section with the one-sentence answer. Expand afterwards.
- Use the question as the heading. "How much does X cost?" beats "Pricing considerations".
- Be specific. "From £1,200, typically 4 to 6 weeks" is quotable. "Competitive pricing" is not.
- Define your terms. If you use a term your reader might not know, define it in the same paragraph.
- Name things. Products, services, places and people with consistent names are easier to cite than descriptions of them.
Entity clarity
A model does not see "a leading provider of solutions". It sees a named entity with attributes.
Make yours easy to assemble:
- One canonical description of the business, used consistently on your site, your social profiles and any directory you control.
- A clear Organization schema with a name, URL, logo and sameAs links.
- A real About page that says who does the work, where, and for whom.
- Consistent service names across pages.
Measuring it
AI visibility is measurable, but not with a rank tracker. You build a prompt set, run it repeatedly, and record whether you were mentioned and whether you were recommended. The measurement guide covers the method in detail.
The honest summary: treat AI visibility as a distribution, not a position. You are looking for a trend across many prompts, not a moment of certainty.