How AI Search Results Work: AI Overviews, AI Mode, ChatGPT and Perplexity

AI search products look different on the surface, but they share a pattern: find relevant pages, read them, write an answer, and link to some of the sources. The details matter, because they decide which controls work and what a site owner can actually influence.

This guide explains each step using the platforms’ own documentation, and flags where they behave differently.

The basic pipeline

1. Crawling and indexing: the platform’s crawler fetches pages it’s allowed to access and stores them.
2. Understanding the question: the system interprets the query, sometimes breaking it into subquestions.
3. Retrieval: it searches its index for relevant pages and passages.
4. Generation: a language model writes an answer grounded in what was retrieved.
5. Attribution: some sources are shown as links or citations.

Google AI Overviews and AI Mode

Google’s AI features use its normal search infrastructure. Google says AI Overviews appear only when its systems decide they add to classic results; AI Mode is designed for deeper exploration and comparisons. Both use query fan-out: multiple related searches across subtopics and data sources, which Google says lets it show a wider and more diverse set of links.

• Eligibility: the page must be indexed and eligible to show with a snippet.
• No special markup, files or optimizations are required, per Google.
• Controls: nosnippet, data-nosnippet, max-snippet and noindex limit what Google shows.
• Reporting: included in Search Console’s Performance report under Web.

ChatGPT search

OpenAI uses OAI-SearchBot to surface websites in ChatGPT’s search features. Sites that disallow it won’t appear in ChatGPT search answers, though they may still appear as navigational links. GPTBot is a separate crawler for model training; blocking it doesn’t affect ChatGPT search. ChatGPT-User visits pages when a user asks, and OpenAI notes robots.txt rules may not apply to those user-initiated visits. OpenAI says robots.txt changes for search take about 24 hours.

Perplexity

PerplexityBot crawls to surface and link websites in Perplexity’s results and, according to Perplexity, isn’t used for training foundation models. Perplexity-User fetches pages when a user’s question needs them and generally ignores robots.txt, because the fetch was user-requested.

Why some pages get cited

No platform publishes a full selection formula, so be skeptical of anyone claiming one. What the documentation and the GEO research suggest is consistent with general SEO:

• The page is accessible to that platform’s crawler and indexed.
• It answers the specific question or subquestion clearly.
• The relevant passage makes sense on its own.
• The information is specific, accurate and supported.
• The source looks trustworthy for the topic.

See how to make content easier to cite for applying this.

What you can and can’t control

You can control	You can’t control
Which crawlers can access your site	Which sources a system picks for a given query
Whether pages are indexable and snippet-eligible	How an answer is worded
How clearly and accurately you state information	Whether a link is shown for every use
Consistency of your brand facts	When AI features appear for a query

For crawler rules, see robots.txt mistakes; for the bigger picture, AI SEO vs traditional SEO.

Frequently asked questions

Sources

• Google Search Central: AI features and your website
• Google Search Central: Control your snippets in search results
• Google Search Central: Google's common crawlers (Google-Extended)
• OpenAI: Overview of OpenAI crawlers
• Perplexity: Perplexity crawlers
• Aggarwal et al.: GEO: Generative Engine Optimization (arXiv, 2023/2024)

Last reviewed October 12, 2026.