AI engines
How AI search engines pick the brands they name
Each AI engine builds its answers differently: ChatGPT, Perplexity and Claude search the web when it helps and cite their sources, Gemini grounds answers in Google's Search index, and Google's AI Overviews and AI Mode are built from Google Search itself. That is why a brand can be first in one engine and missing in another, and why each one needs its own crawler rule and its own measurement.
| Engine | Where answers come from | Crawler that matters | How sources are shown |
|---|---|---|---|
| ChatGPT | Live web search when it helps the question; otherwise model knowledge | OAI-SearchBot (search), GPTBot (training) | Inline citations and a sources panel |
| Google AI Overviews & AI Mode | Google Search index, several related searches per question | Googlebot (Google-Extended does not apply) | Supporting links shown with the answer |
| Perplexity | Real-time web search for every answer | PerplexityBot (search index) | Numbered citations in every answer |
| Gemini | Grounding with Google Search when it helps | Googlebot crawls; Google-Extended controls Gemini use | Sources button when sources are available |
| Claude | Web search when current information helps | Claude-SearchBot (search), Claude-User (user requests) | Citations in every answer that uses web search |
Search crawler OAI-SearchBot, live web search, sources shown with the answer.
How it worksBuilt from Google Search; Googlebot and snippet rules decide eligibility.
How it worksSearches for every answer, numbered citations, crawler PerplexityBot.
How it worksGrounds answers in Google's index; Google-Extended controls Gemini use.
How it worksSearches when it helps, cites every search answer, three separate bots.
How it worksWhat all engines have in common
- Access first: if the engine's search crawler cannot reach your pages, nothing else matters for live answers.
- Sources decide: with search on, the pages an engine reads for a question shape which brands it names. A large study found the AI assistants cite mostly third-party (earned) sources.
- Answers vary: the same question can name different brands from one run to the next, so measure rates over repeated answers, not screenshots.
- Language and country matter: a German question for Germany can name other providers than the same question in English.
Training crawlers are a separate decision
Every vendor separates the crawler for search from the one for training. GPTBot, ClaudeBot and the Google-Extended token are about training (and, for Google-Extended, Gemini app grounding). OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot are about being found. Blocking a training crawler is a legitimate choice; blocking a search crawler by accident is the most common reason a site never gets cited. The free AI crawler checker shows your current rules.
How Nodpick measures each engine
Nodpick asks every engine the same buyer questions in your buyers' country and language, and reports a score per engine, who is named instead of you and which domains each engine cites. The free check covers ChatGPT, Perplexity, Gemini and Claude; paid plans add Google AI Overviews and AI Mode and ask every week. The method explains the score and its range.
Common questions
Which AI search engine matters most for my brand?
The one your buyers use, and that differs by market. Measure all of them for the same questions before you decide where to spend effort; the gaps between engines are often larger than expected.
Do I need a separate strategy per engine?
Mostly not. Access, clear facts and presence on the cited sources help everywhere. What differs is the crawler rule per engine and which sources each one prefers, which you only see by measuring.
Why does one engine name me and another not?
They search different indexes and read different pages. Compare the domains each engine cites for the same question: the engine that ignores you usually cites pages that do not list you.