SEO for LLM: Get Cited by ChatGPT, Gemini and 13 More Engines

Why this site exists

SEO for LLM exists because the advice on AI search is a mess. Search for how to get cited by an AI engine and you will find confident lists of ranking factors with percentages attached, usually from companies that sell a tool or a service. Some of those numbers come from real studies. Many come from nowhere. A reader has no way to tell the difference, and neither does the engine that later summarises those pages.

So we do something simple. For every engine we separate three things: what the engine's own documentation says, what independent studies or hands-on tests show, and what vendors only claim. Each guide says which is which. When we do not know, we say that too. It is slower than copying a checklist, but it is the only version of this advice we would trust ourselves.

What you will find here

There are two guides for each of the 15 engines on this site, so 30 in-depth articles in total.

• How it works. What the engine is, where its web results come from, which crawlers it documents, how it shows sources, and what that means for a publisher. These pages answer the question "should I care about this engine at all?"
• How to rank. A step-by-step playbook, ordered by what blocks you first. Crawler access comes before wording, wording comes before schema, and measurement closes the loop. Each playbook includes at least one worked example you can copy.

The engines covered are ChatGPT, Gemini, Claude, Microsoft Copilot, Grok, Meta AI, DeepSeek, Mistral Le Chat, Qwen, Kimi, Perplexity, Google AI Overviews, Google AI Mode, Brave Search AI and YandexGPT. They behave differently. Google's AI features sit on top of Google's own index. Perplexity and Brave lean on independent retrieval. The Chinese assistants serve a different web, with different crawlers and different expectations about language. Treating them as one "AI search" blob is the first mistake most guides make.

What most engines have in common

Across the engines we studied, a few points came up again and again, and they are worth knowing before you open any single guide.

1. Access comes first. If a crawler or fetcher cannot read your page, nothing else matters. Check robots.txt, your CDN's bot rules and whether your key content is in the HTML the server returns.
2. Answers get lifted from passages, not pages. A paragraph that states a clear answer, with the number or name that supports it, is easier to quote than the same fact buried in a long story.
3. Specifics beat generalities. Prices, dates, measurements, named sources and first-hand detail give an engine something it can attribute. Generic copy gives it nothing to cite.
4. Freshness and honesty both count. Update pages when facts change, and show the real date. A fake "updated today" label helps nobody.
5. You can only manage what you measure. Most engines do not give you a ranking report. You will have to ask real questions, read the sources, and keep a log.

None of this is magic, and none of it is new. Good SEO has always rewarded pages that answer a question well and can be trusted. What changes from engine to engine is the route by which your page reaches the answer, and that route is what the guides map.

How we write the guides

Every guide starts from collected sources: official documentation where it exists, saved articles from agencies and publishers, and transcripts of video walkthroughs in which practitioners test engines on screen. We read them, check them against primary sources where we can, and record each fact with its origin. Claims we cannot confirm are marked as reported or unconfirmed, and the weakest ones are left out.

We also say plainly where our evidence is thin. For some engines, such as Mistral Le Chat, Kimi or YandexGPT, we had no hands-on video material and relied on documentation and published analysis. Those guides say so in a "Limits" section. You can read the full process on the methodology page.

Who this is for

The guides are written for people who publish on the web and want to be found in AI answers: SEO specialists, content marketers, founders with a documentation site, and editors at publishers who are wondering what to allow or block. You do not need a technical background to follow them. Where code helps, such as a robots.txt rule or a JSON-LD block, we show the snippet and explain what it does.

Where to start

If you have an hour, read the overview of how AI engines differ and then pick the two engines that matter most for your audience. If you already know your target, go straight to its playbook from the ranking hub. The most-read starting points are the guides for ChatGPT, Google AI Overviews and Perplexity.

The site is edited by Cisse Mohamed. If you spot an error, or you have evidence that contradicts a guide, please get in touch. Corrections are applied and dated.