AI search engines are not one thing, and this hub explains how 15 of them differ before you pick which ones to optimise for. Each engine below has its own guide covering where its web results come from, which crawlers it documents, how it shows its sources, and what is still unknown.
Why the differences matter
When people say "optimise for AI", they usually picture a single system that reads your page and decides whether to quote it. In practice there are at least four different designs behind the products on this list.
- Assistants that call a search tool. The model decides when to search, sends a query to a search provider, reads the results and writes an answer. ChatGPT, Claude, Grok, Mistral Le Chat and several others work roughly like this, though the provider behind the tool differs and is not always disclosed.
- Answers built into a search engine. Google AI Overviews and AI Mode are produced inside Google Search, on top of Google's own index and ranking systems. Brave and Yandex follow a similar pattern with their own indexes.
- Answer engines that retrieve first. Perplexity is the best-known example: the product is built around fetching sources and citing them in numbered form.
- Regional assistants. DeepSeek, Qwen and Kimi serve a web and a user base that Western SEO guides rarely cover, with different crawlers, different language expectations and different publishing norms.
This matters for a plain reason. If an engine uses Bing's index, a Bing problem is your problem. If it uses Google's index, a Google problem is your problem. If it has its own crawler, you can allow or block it separately. A single checklist cannot capture that, which is why each engine gets its own page.
How to use this hub
Start with the table at the bottom of this page. It lists each engine, who makes it, and what kind of evidence stands behind our guide. Engines with video test material are marked, and the others are built from documentation and published analysis. The distinction is deliberate. We want you to know when a guide rests on observed behaviour and when it rests on what vendors say.
Then read the guide for the engine your audience actually uses. If you do not know which that is, look at your analytics for referrals from AI products, or ask a handful of customers which assistants they use. For many sites the answer is two or three engines, not fifteen.
Four families worth knowing
Google's AI features. Google AI Overviews, Google AI Mode and Gemini are related but not identical. Overviews appear on the results page, AI Mode is a separate conversational view, and Gemini is the standalone assistant. Google's own guidance for the first two says there is no special markup you need beyond being indexable and useful. We cover what that does and does not mean in practice.
Microsoft and OpenAI. Microsoft Copilot is tied to Bing. ChatGPT documents its own crawlers and says it sometimes partners with other search providers. The relationship between the two is a recurring question in guides, and we flag which parts are confirmed and which are only reported.
Independent retrieval. Perplexity, Brave Search AI, Claude and Mistral Le Chat are grouped here because each has either its own index or its own documented fetchers. Their guides spend more time on crawler access than the others.
Chinese and Russian ecosystems. DeepSeek, Qwen, Kimi and YandexGPT operate under different rules and often different languages. These guides are the most cautious, because primary documentation in English is thin and agency claims are common.
Social platforms add two more: Grok leans on live posts from X, and Meta AI draws on Meta's own surfaces and partners.
What each overview contains
Every "how it works" guide follows the same promises, even though the outlines differ by engine:
- A plain definition of the product and the part of it that touches the open web.
- The crawlers and fetchers the engine documents, and what controls them.
- How it presents sources to a user, because that decides what a citation is worth.
- An "Analysis of the reviewed walkthroughs" section, when video test material exists.
- A frequently asked questions block, a limits section and a sources list.
When you finish an overview, the natural next step is its companion playbook on the how to rank hub.