How AI Visibility Tracking Differs by Engine
How ChatGPT, Gemini, Google AI Overviews, AI Mode, Perplexity and Claude use the web and show sources, and what that means for B2B SaaS tracking.
Short answer: "Visibility in AI" is not one place. ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity and Claude look at the web in different ways and show sources differently. That is why tracking is set up per engine. For each engine, record four things separately: did the engine look at the web for that answer, did it show sources, which of its crawlers can reach your site, and when can a change show up? Botfusions measures ChatGPT, Gemini, Google AI Overviews, Google AI Mode and Perplexity today. Claude is not in our measured set.
B2B SaaS marketing teams often ask one question: "Does AI recommend us?" That question has a separate answer for each engine. We covered why the same question gets different answers across engines in our post on why AI answers show your brand only sometimes. This post looks at something else: what you can record on each engine, and how that changes your tracking plan.
Every engine behaviour below comes from the vendor's own documentation. If we could not find a behaviour in the documentation, we left it out.
1. Write down the engine set first
The first rule of tracking is to state, in every report, which engines were measured. Our measurement package covers five surfaces today: ChatGPT, Gemini, Google AI Overviews, Google AI Mode and Perplexity. Claude is not in that standard package. We do not write a rate for an engine that was not measured, because "zero" and "not measured" are not the same thing. Details are on our methodology page.
One note: our 476-answer agency measurement (29 October 2026) was run before Perplexity joined our set. That measurement ran on four engines. Do not put two measurements taken with different engine sets side by side.
2. ChatGPT: the search crawler opens the door
OpenAI's crawler documentation defines three separate agents. OAI-SearchBot is used to surface websites in ChatGPT's search features. Sites that opt out of OAI-SearchBot are not shown in ChatGPT search answers, though they can still appear as navigational links. ChatGPT-User is the agent that may visit a page when a user asks a question. OpenAI says this agent does not crawl the web automatically and that, because a user starts these actions, robots.txt rules may not apply. GPTBot is the crawler that collects content for training.
OpenAI also says that after a robots.txt update it can take about 24 hours for search results to adjust.
OpenAI's API web search documentation explains that the model can choose whether to search the web based on the prompt. According to the same document, the number of sources the model consulted is often greater than the number of citations it shows. This document describes the developer API, not the ChatGPT interface. It still gives a useful distinction for tracking: the pages read and the pages cited are not the same list.
What changes in tracking: check that OAI-SearchBot is allowed in your robots.txt. For every ChatGPT answer, record separately whether it contains links. In an answer without links, the citation signal cannot occur in the first place.
3. Google AI Overviews and AI Mode: the search index is the base
In its AI features documentation, Google says that to appear as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet. It also says there are no additional technical requirements and no special optimizations needed. Both surfaces may split a question into subtopics and run several related searches ("query fan-out").
Three details in Google's documentation affect tracking directly:
- AI Overviews are shown only when Google's systems determine they add something to classic Search. A search with no overview does not mean your brand dropped. That search should come out of the measurement's denominator.
- AI Mode and AI Overviews may use different models and techniques, so the answers and links they show can differ. Do not merge them into one "Google" row.
- Traffic from these features is reported in the Search Console Performance report, within the "Web" search type.
Google's AI Mode help page adds two more details. AI Mode may show a set of web links when there is not high enough confidence in an AI response. For the best experience, it recommends turning on search history and personalised recommendations. So a check done in a signed-in account may not match a check done in a clean session.
What changes in tracking: first confirm that your pages are indexed and do not carry a snippet block. Report AI Overviews and AI Mode as separate engines. Count searches with no overview separately.
4. Gemini: not every answer has sources
Google's Gemini help page says Gemini Apps sometimes show sources and related content within and below a response. The Sources button can sit at the bottom of the response or inline in the text. The same page says that not all responses include related links or sources.
What changes in tracking: on Gemini, the citation signal may not occur in every answer. Read mention and position as the main signals on this engine. Count the answers that do show sources separately, but do not treat a missing source as a bad sign on its own.
5. Perplexity: numbered sources in every answer
Perplexity's help center says the tool searches the internet in real time and that each answer includes numbered citations linking to the original sources. Perplexity's crawler documentation defines two agents. PerplexityBot is designed to surface and link websites in Perplexity search results, and Perplexity recommends allowing it in robots.txt. Perplexity-User fetches pages to answer a user's question and, because a user requested the fetch, generally ignores robots.txt rules.
What changes in tracking: Perplexity is the engine where the citation signal is richest. Record the numbered source list of every answer. Look at which of your pages appears, on which question, and as which numbered source. Check that PerplexityBot is allowed in your robots.txt.
6. Claude: did the answer use web search?
Claude is not in our measured set. This section covers what to watch if you want to track Claude yourself.
Anthropic's web search help page says Claude uses a search tool to ground its responses in content from the live web. Responses that use web search include citations to sources. The page describes two cases. In the classic experience, web search is turned on from within a chat; on Team and Enterprise plans, an owner must first enable it for the workspace. In the new Claude experience there is no web search toggle, and Claude searches the web when it helps. Anthropic's crawler page defines three agents: ClaudeBot for training, Claude-User that fetches pages at a user's request, and Claude-SearchBot that crawls the web to improve search result quality. Anthropic says blocking Claude-SearchBot may reduce a site's visibility in user search results.
What changes in tracking: if you measure Claude, record for every answer whether it used web search. The easiest sign is whether the answer carries source citations. An answer built on web search is not the same kind of answer as one without it. Also note which experience and plan you used.
7. Per-engine summary
| Engine | How it looks at the web (per vendor docs) | How sources are shown | Crawler that reaches your site | Measured by Botfusions |
|---|---|---|---|---|
| ChatGPT | Through OAI-SearchBot in search features | Inline citations when it searches (API docs) | OAI-SearchBot, ChatGPT-User | Yes |
| Google AI Overviews | Google index, query fan-out | Supporting links | Google Search index | Yes |
| Google AI Mode | Google index, query fan-out | Links, a link list at low confidence | Google Search index | Yes |
| Gemini | Sources in some answers | Sources button, not in every answer | Not stated in the doc | Yes |
| Perplexity | Real-time search | Numbered sources in every answer | PerplexityBot, Perplexity-User | Yes |
| Claude | When turned on (classic), when it helps (new experience) | Citations in search answers | Claude-SearchBot, Claude-User | No |
The source note in the ChatGPT row is based on OpenAI's API documentation, not on the ChatGPT interface. For the Google rows, the indexing and snippet requirement in Google's documentation applies.
What this means for your tracking plan
- Fix the engine set and write it down. Every report should show which engines were measured.
- Check crawler access once. Confirm that OAI-SearchBot and PerplexityBot are allowed in your robots.txt and that your pages are in Google's index. If you also target Claude, add Claude-SearchBot.
- Interpret the citation signal per engine. On Perplexity, expect sources in every answer. On Gemini, not every answer has sources. On ChatGPT, if the model decides not to search, the answer has no web sources.
- Separate Google's two surfaces. Keep AI Overviews and AI Mode in separate rows. Count searches with no overview separately.
- Record the session conditions. Signed in or not, search history on or off, did Claude answer with web search?
For the full monitoring workflow, see our AI visibility monitoring guide. If you want us to run this measurement per engine for you, see our GEO services for B2B SaaS page.
Limits of this post
We read the engine behaviours from the vendors' own documentation on 13 November 2026. These products change often and the documentation can change too. We could not verify from OpenAI's own help pages how the ChatGPT interface displays sources, so we wrote nothing about it. What vendors say about freshness is limited: Perplexity says it searches in real time, OpenAI says a robots.txt change takes about 24 hours to show in search results, and Google states the indexing requirement. How long it takes for your page to enter an answer is something only your own measurement can show.
Sources
- [1] OpenAI docs: crawlers (bots)
- [2] OpenAI API docs, "Web search"
- [3] Google Search Central, "AI features and your website"
- [4] Google Search Help, "Get AI-powered responses with AI Mode in Google Search"
- [5] Gemini Apps Help, "View related sources from Gemini Apps"
- [6] Perplexity Help Center, "How does Perplexity work?"
- [7] Perplexity docs: crawlers (bots)
- [8] Claude Help Center, "Enable and use web search"
- [9] Claude Help Center, "Does Anthropic crawl data from the web, and how can site owners block the crawler?"
