AI search covers Google's AI Overviews and AI Mode, ChatGPT search, Perplexity, Microsoft Copilot and others. Each fetches pages, builds an answer and may link its sources. How it chooses is mostly unpublished.
So this guide separates what the engines document from what people guess. The documented parts are short, and they point to ordinary good practice. Read the sibling guides for each engine, linked below.
What is AI search?
AI search is any search experience in which an AI model writes an answer from pages it finds, and usually links to some of them. It differs from a classic results page, which lists links and leaves the reader to open them. Our glossary defines AI Overviews, answer engine optimization, generative engine optimization and LLM SEO.
The change in clicking is measurable. Pew Research tracked 68,879 Google searches by 900 US adults in March 2025 and found that users clicked a result link in 8% of visits with an AI summary, against 15% without. So some value now reaches a business without a visit.
What do the engines say about how they pick sources?
Less than most guides imply. The table below sets out what each engine's own documentation says and what it does not. We read each page on 12 November 2026.
Google: AI Overviews and AI Mode
Google's documentation says AI Overviews and AI Mode may use a technique it calls query fan-out, which issues several related searches across subtopics and data sources, so they can show a wider and more varied set of links than classic search. A page can appear as a supporting link if it is indexed and eligible to be shown in Search with a snippet. Google says there are no additional technical requirements, no special schema.org markup and no AI text files needed.
Google's May 2025 post adds practical advice: make unique, helpful content, give a good page experience, make sure Googlebot is not blocked and the page returns a success status, keep structured data matching the visible text, and look at the overall value of visits, not clicks alone. It also warns that more restrictive snippet controls limit how your content can appear in AI experiences. What Google does not publish is how it ranks or selects the links in a given answer.
ChatGPT, Perplexity and Microsoft
OpenAI's crawler documentation says OAI-SearchBot is used to surface websites in ChatGPT's search features, and that a site that opts it out will not be shown in ChatGPT search answers, though it may still appear as a navigational link. GPTBot is a separate crawler for content that may be used to train its models, and the two can be set independently in robots.txt. ChatGPT-User handles actions started by a user, and robots.txt rules may not apply to it. The page does not say how pages are ranked in answers.
Perplexity's bots guide says PerplexityBot is designed to surface and link websites in Perplexity search results and is not used to crawl content for AI foundation models. Perplexity-User supports user actions and generally ignores robots.txt, because a user triggered the visit. Its help page says answers gather information from authoritative sources and carry numbered citations, without explaining how sources are ranked.
Microsoft gives site owners something concrete: Bing Webmaster Tools has an AI Performance report. Microsoft's 16 June 2026 post says it shows where content is cited in AI-generated answers across Copilot, Bing and select partner experiences, and, in preview, groups the grounding queries by intent and topic. It does not show traffic.
Gemini and other assistants
We found no separate Google page explaining how the Gemini app chooses sources, only the Search documentation above, so we say nothing more about it. The same goes for assistants whose makers publish no source-selection guidance. Where an engine says nothing, any advice about it is a guess, and a careful guide labels it as one.
What works across all of them?
Four things hold in every engine's documentation or follow from it.
- Access. Let each engine's search crawler reach your pages: Googlebot, OAI-SearchBot and PerplexityBot are the documented search-side crawlers. Decide about training crawlers such as GPTBot separately.
- Clear answers. Put a direct answer to the page's main question in the first two sentences, then the detail. A reader or a model that stops there still has the point.
- Trust signals. Name the author and the reason they know the subject, cite sources with dates, and keep company details the same everywhere.
- Page quality. Fast, readable pages with the content in text, and structured data that matches what is visible.
What should you do first?
- Check access. Read your robots.txt and any CDN or firewall rules to see which crawlers are blocked, on purpose or by accident.
- Pick your ten most important pages. Make sure each is indexed and answers its question in the opening.
- Fix the facts. Prices, hours, service areas and names should match across the site, your Business Profile and directories.
- Add the author and the date. Show who reviewed each page and when, and change the date only when the facts change.
- Start measuring. Set up the reports in the tracking guide before you change anything, so you can compare.
What is not worth your time?
Formulas that claim to reveal how a model ranks sources, pages built only to fill a topic map, and paid promises of citations. Google's spam policies treat many pages made mainly to manipulate rankings as scaled content abuse, whether or not AI wrote them. A file such as llms.txt is optional and has no documented effect on Google's features; our llms.txt guide explains where it fits.
Also avoid treating AI search as a separate channel with a separate budget before the basics are sound. The same fixes help in classic search too, so the work is not lost if an engine changes its behavior.
How do you know if it is working?
Count outcomes first: calls, form sign-ups, bookings and sales, and ask new customers how they found you. Then add the engine reports that exist: Search Console, which counts AI feature traffic inside the Web totals, and Bing's AI Performance report. Our guide to tracking AI search visibility gives a repeatable method and its limits.
Where do the sibling guides fit?
Use these for the engine you care about most: Google AI Overviews, Google AI Mode, ChatGPT search, Perplexity, local businesses and being cited in AI search.
For the vocabulary, see GEO vs SEO vs AEO, and for markup see schema for AI search.
Conclusion
AI search rewards the same things good search always did, and the engines say little more. Keep your pages open to the right crawlers, clear, accurate and attributed, and measure real outcomes.
PromotEdge International is an international strategy, technology, marketing and design company, and our delivery team is in India. See our AI search service and our technical audits.
Facts we rely on: see our company facts for what we state about ourselves and where each fact comes from.