AI search is a query experience where a language model reads your question, gathers information from web pages (and sometimes its own training data), and returns a written answer with cited sources. You get a paragraph, not ten links. The industry also calls this an "answer engine," and the practice of getting cited in one is called AEO (answer engine optimization) or GEO (generative engine optimization).
This page is part of our AI search visibility pillar. Here we cover what AI search is and how the major engines differ. For tactics, see how to appear in AI search.
How AI search differs from classic search
Classic search matches your keywords to indexed pages and ranks them. You do the reading and synthesis. AI search does the synthesis for you.
| Classic search | AI search | |
|---|---|---|
| Output | Ranked list of links | One written answer |
| Your job | Click, read, compare | Read the answer |
| Sources shown | Titles and snippets | Inline citations |
| Follow-ups | New search | Conversational |
| Winner | Top-ranked pages | Cited pages |
The practical shift: in classic search you compete for the top of a list. In AI search you compete to be one of the handful of sources the model quotes. A page can be cited in an AI answer without ranking in the top 10 organic results. As of early 2026, only 38% of AI-cited pages rank in the top 10 organic, down from 76% in mid-2025.
The major AI search surfaces
Google AI Overviews
An AI-written summary that appears at the top of a normal Google results page. You do not opt in; Google triggers it based on query type. AI Overviews now appear on roughly 18-20% of searches and over 50% of long-tail queries. When one appears, click-through to links below drops sharply.
Google AI Mode
A separate conversational surface inside Google, closer to chatting with Gemini. Google reported AI Mode passing 1 billion monthly users at I/O 2026. It fans a query out into sub-queries, retrieves results for each, and merges them. When AI Mode is fully active, up to 93% of those searches end without a click to an external site.
ChatGPT Search
Web search built into ChatGPT, available on the free plan. It leans on the model's training data for settled facts and runs a live web search when freshness matters, attaching citations after the answer is generated. ChatGPT Search handles an estimated 250-500 million queries per week.
Perplexity
An answer engine built around sourcing from day one. It runs a real-time web search on every query, reads candidate pages, and writes an answer with numbered inline citations. Perplexity crossed 1 billion monthly queries in Q1 2026.
Microsoft Copilot
Microsoft's assistant, grounded in Bing. It queries Bing in sequential steps, each informing the next, and attributes sources from that pool. It is embedded across Windows, Edge, and Microsoft 365.
Why the engines disagree
These are not interchangeable. They pull from different source pools and rank with different mechanisms. In cross-platform tests, only about 12% of cited sources match across engines for the same query. Being cited by Perplexity does not mean you are cited by ChatGPT or AI Overviews. Visibility has to be earned surface by surface.
What powers an AI answer
Most AI search runs on retrieval-augmented generation (RAG): the engine retrieves candidate documents, then the model writes an answer grounded in them. Two design choices explain most of the behavior differences:
When it retrieves. Perplexity retrieves on every query. ChatGPT retrieves only when it decides freshness is needed, otherwise answering from training data.
When it cites. Perplexity embeds citations before writing. ChatGPT attaches them after. This affects how tightly the answer tracks its sources.
See where you stand
SEO AEO Specialist runs a free AI-visibility audit and hands you the exact fixes. One-off report €9, unlimited €19/mo.
Run your free audit →