How AI engines actually pick sources
ChatGPT (search mode), Perplexity and Google AI Overviews all do roughly the same thing: run a real-time search, fetch the top 5–20 results, extract quotable sentences, and assemble an answer with citations. To be quoted you need to be (a) in the top results, and (b) easy to quote.
1. Win the underlying search first
If you don't rank in the top 10 organically for a query, you almost never get cited for it. The boring news is that AEO starts with traditional SEO: a fast site, clear topical pages, and a few credible backlinks.
2. Make your pages extractable
- Open every key page with a 40–60 word direct answer to its title question.
- Use clear H2/H3 structure — one idea per section.
- Add bullet lists and short comparison tables. LLMs love them.
- Mark up FAQs with FAQPage schema and HowTo content with HowTo schema.
3. Build a brand footprint LLMs can see
Large language models weight third-party mentions heavily. Get your brand mentioned in:
- Reddit threads in your niche (genuinely — not spam).
- Industry roundups and "best X in the UK" lists.
- Review sites: Trustpilot, Google reviews, niche review hubs.
- Podcasts and YouTube interviews — transcripts get indexed.
4. Publish an llms.txt and keep your HTML semantic
Add an /llms.txt file describing your site, key URLs, and policies. Avoid wrapping your main content in giant JavaScript apps that hide the text from crawlers — server-render or pre-render the core copy.
5. Measure it
Track a list of 20–30 core questions monthly. For each, run them in ChatGPT, Perplexity, and Google AI Mode. Note: are you cited, mentioned, or absent? Improve the worst performers first — usually by sharpening the intro answer and adding FAQ schema.