The three terms side by side
| SEO | AEO | GEO | |
|---|---|---|---|
| Goal | Rank in search results | Be the extracted answer | Be cited in AI answers |
| Where | Google, Bing | Featured snippets, voice, AI answers | ChatGPT, Perplexity, AI Overviews, Claude, Copilot |
| Unit | The page | The answer passage | The passage and the source’s trust |
| Success metric | Rankings, clicks | Snippet ownership | Citations, AI referrals |
What is actually different for AI search
1. Different crawlers, different rules
ChatGPT Search relies on OAI-SearchBot, Perplexity on PerplexityBot, Claude on Claude-SearchBot. Each reads its own robots.txt group, and CDN bot protection often blocks them while letting Googlebot through. A site can rank well on Google and still be missing from ChatGPT. See whether to block GPTBot, how ChatGPT Search finds a page, and the AI crawler list.
2. Little or no JavaScript rendering
Googlebot renders JavaScript; OpenAI’s and Anthropic’s crawlers have been observed fetching but not executing it. Text that only appears after scripts run is invisible to them.
3. Passages, not pages
AI answers quote or paraphrase short passages. A clear definition in one sentence, a numbered list of steps, or a comparison table is easier to use than the same facts spread across paragraphs of marketing copy.
4. Trust is judged across the web
Generative answers tend to cite sources that are consistent with other sources and clearly identified — named authors, an organization with a real address and profiles, dates, and references. On-page work can’t substitute for being mentioned elsewhere.
A practical checklist
- Access: allow Googlebot, Bingbot, OAI-SearchBot, PerplexityBot and Claude-SearchBot in robots.txt and in your CDN.
- Rendering: make sure the main content is in the HTML the server returns.
- Indexability: 200 status, self-referencing canonical, no stray
noindex, an XML sitemap. - Answer-first writing: question-shaped headings and a direct answer in the first sentence below each.
- Structure: lists for steps, tables for comparisons, short paragraphs.
- Entity clarity: Organization or Person schema with
sameAslinks, author bylines, visible dates. - Evidence: specific numbers, sources and first-hand detail that other pages don’t have.
- Measurement: track AI referrals in analytics and AI Overviews traffic in Search Console.
The free audit checks items 1–6 automatically and shows the evidence for each.
What doesn’t work
- Keyword stuffing for prompts. AI answers are generated from meaning, not exact-match phrases.
- Hidden text or schema that doesn’t match the page. It violates search guidelines and erodes trust.
- Mass-produced pages. Near-duplicate content adds nothing a model would cite over the original.
- Treating llms.txt as a ranking factor. It isn’t documented as one by any major engine.