LLM SEO: How to Optimise Your Site for ChatGPT, Claude & the Next Wave of AI Search
A guide to shaping your content for accurate AI citations
Why This Matters
Google still matters, but the first place many users ask a question is now a chatbot. If ChatGPT, Claude, or Perplexity can't find a clean, trustworthy answer on your site, it paraphrases someone else—or worse, hallucinates. LLM SEO (sometimes called Generative Engine Optimisation, GEO) is the practice of shaping your content so language models quote you accurately and often.
1 · Classic SEO vs. LLM SEO at a Glance
| Classic SEO | LLM SEO / GEO | |
|---|---|---|
| Primary gatekeeper | Google/Bing index | ChatGPT, Claude, Perplexity, Gemini |
| Ranking signal | Backlinks, on-page keywords, Core Web Vitals | High-quality content chunks, structured summaries, citation confidence |
| Crawl mechanism | Search-bot follows links & sitemaps | AI agent fetches context on-demand, often with limited token budget |
| Optimisation artefacts | robots.txt, sitemap.xml, Schema.org | llms.txt, Markdown mirrors, embedding-ready chunks |
| Winning metric | Click-through rate to site | Accuracy of answers & frequency of brand citation |
2 · Core Pillars of LLM SEO
- Surface clean text
- Provide Markdown versions of docs (
/page.mdor GitHub raw). - Strip navigation chrome, ads, and cookie banners—they waste tokens.
- Provide Markdown versions of docs (
- Publish
llms.txt- Curate 20–50 canonical URLs per language under clear H2 headings.
- Add an Optional section for deep dives an agent can skip.
- Chunk and label content
- Keep chunks under ~2 k tokens; add H1 / H2 hierarchy.
- Use descriptive file names (
pricing_fi.md,refund_policy_en.md).
- Machine-readable meta
- JSON-LD still helps—some AI agents ingest Schema.org.
- Include author, date, and canonical URL to boost citation trust.
- Proof & citation hooks
- State facts in short sentences; AI tends to quote sentence-level snippets.
- Provide numbers, tables, and bulleted lists—easy for models to lift verbatim.
3 · Content Formats That Play Well with ChatGPT & Claude
| Format | Why LLMs Like It | Pro Tip |
|---|---|---|
| Markdown docs | Minimal markup; headings become natural chunk boundaries. | Add a one-line summary at the top of each file. |
| FAQ blocks | Short Q-A pairs map directly to user queries. | Keep one question per <h3> so agents can jump-link. |
| Policy snippets | Users ask "What's the refund window?"—models prefer exact clauses. | Use numbered lists; start key sentences with the policy keyword. |
| Code samples | LLMs love copy-pasting runnable code. | Supply both request and response when documenting APIs. |
4 · The SurferSEO Angle: Keyword Research Still Matters
Surfer recently rolled out an AI SEO panel that tracks how often brands are cited in model answers. Their advice:
- Target intent, not just keywords. Optimise for questions like "How do I integrate ContextKit with Vercel?"
- Chunk by sub-topic. One long pillar page makes sense for Google, but a bot with a 16 k-token window may only load the first half.
- Monitor answer accuracy. Surfer's SERP-like "Answer Engine" report shows if a model's reply matches your current content—use it to spot outdated docs.
5 · Six-Step LLM SEO Checklist
- Audit – Ask ChatGPT & Claude five common questions about your product. Note wrong or missing answers.
- Create
llms.txt– Point to the exact pages that contain the correct info. - Convert to Markdown – Export key pages or use a converter (Pandoc, Firecrawl).
- Chunk & label – Split >2 k-token files; add language code to filenames.
- Embed facts – Add explicit numbers, pricing tables, version notes.
- Re-test monthly – Rerun the questions; track citation and accuracy.
(Customers on ContextKit's Continuous plan get Steps 2-6 automated.)
6 · FAQ
Will this hurt my Google rankings?
No. Nothing in llms.txt blocks search crawlers. You're simply providing an extra map for AI agents.
Do I need backlinks for LLM SEO?
Less so. LLMs prioritise freshness and clarity over link juice—but authoritative mentions still help the broader web graph.
What about opt-out?
If you want to block training but allow answer-time crawling, whitelist GPTBot in robots.txt and disallow Google-Extended. llms.txt complements, not replaces, those directives.
Takeaway
Classic SEO isn't dead, but Generative Engine Optimisation is the new front-line for brand trust. A tidy llms.txt, clean Markdown pages, and fact-rich chunks make chatter-bots your ally instead of a liability. Start with one high-intent question today—"What's our refund policy?"—and make sure ChatGPT answers exactly the way you wrote it.
Keywords touched in this post: LLM SEO, AI SEO, generative engine optimisation, llms.txt. Feel free to adapt and republish with attribution.
Optimise Your Site for AI Search Today
Ensure AI models like ChatGPT and Claude cite your content accurately with a tailored llms.txt and optimised content strategy.