AI search guides
Long-form playbooks written from what we see scanning real sites — no recycled advice, no fluff. Start here if you want to be the source ChatGPT quotes.
Quick answer
Getting cited by AI engines comes down to four moves, in order: let the AI crawlers in, serve your content as real server-rendered HTML, answer each page's core question in a self-contained paragraph near the top, and describe your entities in valid structured data. These guides walk through each move with concrete examples, then show you how to measure whether citations actually increased.
Suggested reading order
The library is short on purpose. Read it in this sequence and you'll never be applying a fix whose prerequisite is still broken.
- 1Diagnose before you write
Run a scan of your domain. Note the crawler-access, rendering and schema failures — those are hard blockers and no amount of content work compensates for them.
- 2Fix the machine-readable layer
robots.txt, sitemap, canonical tags, JSON-LD, llms.txt. It's a day of work and it's what makes every later change visible to engines.
- 3Rewrite for extraction
Question-shaped headings, a 40–60 word answer under each, tables and lists instead of dense prose, sources for every factual claim.
- 4Track citations, then iterate
Measure share of voice per engine on your target prompts. Improve the pages that are retrieved but not quoted — they're the cheapest wins on the board.
The library — 3 guides
Every guide leads with a one-sentence answer, so you can decide in five seconds whether it's the one you need.
How to rank in ChatGPT
You rank in ChatGPT by making your pages retrievable (allow GPTBot, ChatGPT-User and OAI-SearchBot), quotable (short, self-contained answer paragraphs with clear entities), and verifiable (Organization and Article schema plus corroborating mentions on sites ChatGPT already trusts).
- ChatGPT Search retrieves live pages — crawler access is a hard gate, not a ranking factor.
- Answer-shaped paragraphs of 40–70 words are what actually get lifted into responses.
- Entity clarity beats keyword density: name your brand, product and category explicitly.
Brand mentions in ChatGPT: how to find, track and grow them
To find your brand mentions in ChatGPT, run a fixed panel of buyer-intent prompts on a schedule and log whether you were named, cited with a link, or absent — then grow mentions by fixing crawler access, publishing corroborated facts about your brand, and getting into the third-party roundups ChatGPT retrieves.
- A one-off prompt proves nothing — mentions vary run to run, so measure share of a fixed panel.
- ChatGPT describes your brand using whatever third-party pages it can retrieve, not your homepage copy.
- Wrong descriptions are usually stale corroboration, not model error — fix the sources.
The AI visibility checklist
AI visibility depends on four things you can audit in an afternoon: whether AI crawlers can fetch your pages, whether the answer is extractable from the HTML, whether schema and dates let engines attribute and trust the claim, and whether third-party sources corroborate it.
- Crawler access is a gate — everything else is wasted if the fetch fails.
- Server-rendered answers beat client-rendered ones every time.
- Freshness signals decide who wins current-events and 'best of' prompts.
Guides pair with free tools
Each guide has a companion tool so you can apply the advice the same afternoon you read it.
| If you need to… | Read | Run |
|---|---|---|
| Get quoted by ChatGPT | The ChatGPT citation guide | AI answer readiness checker |
| Let AI crawlers in | The crawler-access guide | robots.txt for AI generator |
| Publish an llms.txt | The llms.txt guide | llms.txt generator, then the validator |
| Prove your schema is valid | The structured-data guide | Schema validator |
| Know if you're being mentioned | The measurement guide | AI brand mention tracker |
The principles behind every guide
If you remember nothing else from this library, remember these.
- Retrieval comes before persuasion — an unreachable page cannot be cited, however good it is.
- Write the answer before the context; models lift the first self-contained paragraph.
- Structure beats length: a table an engine can parse outperforms three paragraphs of prose.
- Entities must be named, not implied — models can't resolve 'we' or 'the platform'.
- Cite your own sources; source-dense pages are disproportionately quoted by Perplexity.
- Measure citations per engine, not aggregate traffic, or you'll optimize blind.
Frequently asked questions
- Where should a beginner start?
- Start with the guide on getting cited by ChatGPT, then run a scan of your own domain so you're reading the rest of the library against real findings rather than in the abstract. Most teams discover two or three blocking issues — a disallowed crawler, JavaScript-only content, or invalid schema — that make everything else moot until they're fixed.
- How are these guides different from the blog?
- The blog is a broad, multilingual stream published daily. These guides are hand-written, deliberately opinionated, and maintained: each one is revised when engine behaviour changes and carries a visible update date. If the two ever disagree, trust the guide.
- Do I need a paid plan to apply any of this?
- No. Every technique in the library can be executed by hand with the free tools on this site — the llms.txt generator, the robots.txt builder, the schema validator and the answer-readiness checker. The paid product exists to do it continuously across a large site and to track citations over time.
- How often are the guides updated?
- We review the full library each quarter and revise individual guides immediately when a crawler, schema type or citation behaviour changes. The updated date on each card is the real last-edited date, not a rolling timestamp.
Explore further
Connected guides to keep going — short reads, all internally linked.