A growing share of searches now end in an answer rather than a click. Being the source that answer is built from is a different job from ranking first, though the two overlap more than the acronym-sellers admit.
AI assistants build answers from content they can parse and verify: direct answers stated early, structured data, clear headings, specific numbers, and identifiable authors. Most of what works for AEO and GEO is ordinary good SEO done more explicitly, with the difference that you optimise for being quoted rather than clicked.
What the acronyms mean
| Term | Stands for | In practice |
|---|---|---|
| SEO | Search engine optimisation | Ranking in a list of links |
| AEO | Answer engine optimisation | Being the source an answer is built from |
| GEO | Generative engine optimisation | Much the same, framed around generative models |
AEO and GEO substantially overlap and both overlap with SEO. Treat them as an emphasis rather than a separate discipline, and be sceptical of anyone selling them as an entirely new service.
How assistants choose sources
- They need to find you. Most assistants either use a search index or crawl the live web, so ordinary discoverability still applies.
- They prefer parseable structure. Clear headings, lists, tables and structured data are easier to extract from than flowing prose.
- They favour specificity. A concrete number with context is quotable. It depends is not.
- They weigh identifiable authorship. Named authors with verifiable profiles are cited more readily than anonymous corporate pages.
- They check for corroboration. Claims repeated across independent sources are safer to quote.
What to change on a page
- Answer in the first 60 words. Put a direct, self-contained answer near the top, before the context and the caveats.
- Use question-form headings that match how people actually ask.
- Put comparisons in tables. Tables are the easiest structure to extract accurately.
- Give real numbers. Prices, timelines, thresholds. Ranges are fine; vagueness is not.
- Add FAQ schema for genuine visible questions.
- Name the author with Person schema and sameAs to a real profile.
- Cite your sources so claims can be verified.
- Keep each section self-contained, since a quoted passage will be read without the paragraphs around it.
Directories and third-party sources matter more than people expect
For questions of the form best X in Y, assistants overwhelmingly draw on listicles, directories and review platforms rather than on vendor websites.
Which means being present on the sources your industry's answers are built from is often more effective than anything you do on your own site. For agencies that means Clutch and similar; for local businesses it means a complete Google Business Profile and real reviews.
llms.txt and the honest state of it
llms.txt is a proposed convention: a plain-text file describing your site for language models. It costs almost nothing to add.
It is also not confirmed as used by any major assistant, and there is no public evidence it affects citation. Add it if you like, and do not let anyone sell you a strategy built on it. The things with demonstrated effect are structure, specificity, authorship and third-party presence.
Measuring it
This remains genuinely hard. Referral traffic from assistants is small and inconsistently attributed, and there is no equivalent of Search Console for citations.
- Check your analytics for referrals from assistant domains.
- Periodically ask the major assistants questions in your category and note who gets cited.
- Watch for branded search rising without a matching rise in clicks, which can indicate people seeing you in answers.
- Treat any tool claiming precise AI visibility metrics with caution.
Frequently asked questions
What is AEO?
Answer engine optimisation: making your content the source that AI assistants and answer boxes build their answers from, rather than optimising purely for a position in a list of links. It overlaps heavily with ordinary SEO.
Is GEO different from SEO?
Not fundamentally. Generative engine optimisation emphasises being quoted rather than clicked, which favours direct answers, structured data, specific numbers and identifiable authors. The underlying requirement, that your content is discoverable and credible, is unchanged.
How do AI assistants decide what to cite?
They favour content they can find, parse and verify: clear structure, headings that match the question, tables, specific figures, structured data, and named authors with verifiable profiles. Corroboration across independent sources also helps.
Does llms.txt do anything?
There is no public evidence that any major assistant uses it. It is a proposed convention that costs almost nothing to add, and it should not be the basis of a strategy. Structure, specificity and third-party presence have demonstrated effects; llms.txt does not.
How do I measure AI search visibility?
Imperfectly. Check analytics for referrals from assistant domains, periodically test questions in your category and note who is cited, and watch branded search trends. Be sceptical of tools claiming precise AI visibility figures.