I asked ChatGPT "What's the best accessibility testing tool?" last week. It cited three products. Mine wasn't one of them. My site ranks on page 1 of Google for that exact query.
That's the problem. Google rankings don't guarantee AI citations. ChatGPT, Perplexity, and Google AI Overviews decide what to cite using different signals than traditional search. This shift has created a new discipline: Answer Engine Optimisation (AEO).
AEO is the practice of structuring web content so that AI-powered search engines can extract, understand, and cite it.
How answer engines differ from search engines
Search engines reward relevance. Answer engines reward extractability.
An answer engine ingests your content, extracts factual claims, evaluates authority via structured metadata and external corroboration, synthesises an answer from multiple sources, and cites only the pages that directly contributed. If your content isn't structured for extraction, it gets skipped — regardless of your Google ranking.
What AI systems actually look for
Here's what matters for AEO, broken into three categories:
| Category | Key Signals |
|---|---|
| Structural | JSON-LD schema (Article, Product, FAQPage), meta descriptions under 160 chars, clean heading hierarchy, llms.txt file, AI crawler permissions in robots.txt, datePublished/dateModified freshness signals |
| Content | Q&A headings with concise answers, definition sentences ("X is..."), lists and tables, specific numbers over vague marketing copy, outbound links to authoritative sources (W3C, OpenAI GPTBot docs), author attribution |
| Answerability | Self-contained answers extractable without context, one clear topic per page, claims backed by visible evidence |
Before and after
Low AEO score:
We offer flexible pricing plans designed to scale with your business. Contact us to learn more.
High AEO score:
Our Professional plan costs $99/month and includes 100 pages per scan, 100 AI tokens, and API access. Annual billing saves 20%. A free Starter plan supports 10 pages per scan with no time limit.
The second version has specific numbers, answers "how much does it cost?" directly, and stands alone — an AI can extract and cite it without surrounding context.
Five mistakes developers make
1. Client-side only rendering. AI crawlers fetch HTML and don't always execute JavaScript. Use SSR or SSG for content pages. In Next.js, keep content in server components.
2. Content hidden in tabs/accordions. Collapsed UI content may not get indexed. Render critical content (FAQs, pricing) as visible HTML.
3. Missing JSON-LD. No structured data means AI systems have to guess what your page is about. Add Article, Product, or FAQPage schema and validate with Google's Rich Results Test.
4. No llms.txt. This newer standard gives AI crawlers a structured summary of your site. Most sites don't have one — add it in 10 minutes:
# Your Product
> One-line description.
## Key Pages
- [Home](https://example.com): Overview
- [Pricing](https://example.com/pricing): Plans and pricing
- [Docs](https://example.com/docs): Documentation
5. AI crawlers blocked in robots.txt. Explicitly allow them:
User-agent: GPTBot
Allow: /
User-agent: Claude-Web
Allow: /
User-agent: PerplexityBot
Allow: /
How to test your AEO right now
- Pick a question your page should answer
- Ask ChatGPT or Perplexity that exact question
- Check if your site is cited
- If not, compare your page to the cited sources — they'll almost always have structured data, specific numbers, and self-contained answers that yours is missing
Tools like Blyxo can automate this across your whole site, but the manual test above takes 30 seconds and is usually eye-opening.
TL;DR
SEO optimises for ranking. AEO optimises for citation. You need both. Start with these five actions:
- Add JSON-LD to every page
- Create
/llms.txt - Allow AI crawlers in robots.txt
- Add FAQ sections with self-contained answers
- Replace vague copy with specific facts
Have you checked whether ChatGPT cites your site? Try it and let me know what you find in the comments.
Top comments (1)
Great observation Francis — and your own experience (ranking on Google page 1 but not getting AI citations) is exactly the pattern I see running an audit tool. We scan for both traditional SEO signals and AI-readiness, and the overlap is surprisingly small.
The technical root: Google evaluates pages in isolation via its index. AI answer engines evaluate your page in context of everything else it has ingested. If your content is well-structured but a competitor has a Wikipedia entry, a Crunchbase profile, and 50 backlinks from .edu domains, the AI will cite them over you even if your actual article is better.
The fix that tends to work: treat structured data as a first-class distribution channel, not an SEO checkbox. Schema.org with Speakable, About, and especially WebSite with potentialAction — these are the signals AI crawlers look for that Google largely ignores for rankings.
I wrote more about this in a crawl topology deep-dive, but the short takeaway: if you are not getting AI citations despite good Google rankings, the gap is almost always entity authority (who trusts you) vs page relevance (how good is this one post).