The assistants we measure, described accurately
6 surfaces across 5 vendors. Below is exactly what each one queries — including the two whose honest name is not the name you would expect.
TrackGeo measures answers from provider APIs — the models behind these assistants — not scrapes of their consumer apps.
We ask the models behind these assistants the same buyer questions your customers ask. It is a real, repeatable measurement of what the model says. It is not a recording of someone's app session, and the difference changes how you should read every number we give you.
Every engine, and the surface it actually queries
Surface is the fine print that usually gets dropped. It is the first thing we show you, because a measurement is only as meaningful as the thing it measured.
ChatGPT
OpenAI
- What we query
- OpenAI API
- What you should know
- We query OpenAI's API directly. Results reflect the model that powers ChatGPT, not a capture of the ChatGPT consumer app.
Perplexity
Perplexity
- What we query
- Perplexity API (sonar)
- What you should know
- Perplexity's sonar models return answers with citations, which is what makes source tracking possible here.
Gemini
- What we query
- Gemini API
- What you should know
- Ungrounded Gemini responses — the model's own answer, without a live web search step.
Azure OpenAI
Microsoft
- What we query
- Azure OpenAI Service
- What you should know
- This is Microsoft's Azure-hosted OpenAI deployment. We do not measure the Microsoft Copilot consumer product, and we don't claim to.
Claude
Anthropic
- What we query
- Anthropic Messages API
- What you should know
- We query Anthropic's API directly.
Gemini with Search grounding
- What we query
- Gemini API + google_search tool
- What you should know
- Gemini answering with live Google Search grounding. It is the closest available proxy for AI-Overview-style answers, but it is not Google AI Overviews and results will differ from what you see on a search page.
Supported
A production adapter exists and real prompt runs exercise it. The surface line tells you precisely what was asked.
Limited
Real, but the surface is an approximation of the thing you have in mind. Read the note before you trust the number — and know that we never quietly promote one of these to supported.
How many engines each plan measures
Running the same prompt on more engines is not only about coverage. Engine count feeds the sample behind a finding, so it raises how much confidence a result can earn.
| Plan | Engines | Included |
|---|---|---|
| Free | 2 of 6 |
|
| Essentials | 4 of 6 |
|
| Professional | 5 of 6 |
|
| Growth | 6 of 6 |
|
Enterprise includes every engine as well. See the full plan comparison.
A logo on a pricing page is not an integration
The easiest way to win a feature comparison is to list every assistant anyone has heard of and route them all somewhere convenient. This page exists so we cannot do that.
Listed means built
An engine appears here only when a production adapter exists and real prompt runs exercise it. If an assistant is not on this page, we do not measure it — there is no third state where we quietly send your prompt somewhere else.
Named for what it is
Two of the engines above carry a name that differs from how they are labelled inside the product, because the internal label would be a false claim in public. The honest name is the one you see.
Caveats stay attached
The note on a limited engine travels with it everywhere it is rendered. Nothing here gets promoted because a number would look better without the asterisk.
Because answers vary run to run, a single response from any engine on this page tells you very little. What makes engine coverage useful is repetition across engines over time — which is also how confidence is earned.
Engine coverage, answered plainly
Which AI engines does TrackGeo measure?
TrackGeo measures 6 surfaces across 5 vendors: ChatGPT (OpenAI API), Perplexity (Perplexity API (sonar)), Gemini (Gemini API), Azure OpenAI (Azure OpenAI Service), Claude (Anthropic Messages API), Gemini with Search grounding (Gemini API + google_search tool). Each one is listed with the exact surface it queries, because the surface is the part that determines what the number means.
Are these numbers taken from the consumer assistant apps?
No. TrackGeo measures answers from provider APIs — the models behind these assistants — not scrapes of their consumer apps. We ask the same underlying models the same buyer questions your customers ask. That is a defensible and useful measurement, but it is not a capture of anyone's app session, and we will not imply otherwise to make the number sound better than it is.
Why does an engine get marked limited rather than supported?
Limited means the measurement is real, but the surface approximates the thing you probably have in mind — so the caveat matters more than the checkmark. Today that applies to one engine: Gemini with Search grounding. Gemini answering with live Google Search grounding. It is the closest available proxy for AI-Overview-style answers, but it is not Google AI Overviews and results will differ from what you see on a search page. A limited engine is never silently upgraded to supported.
What happens if an assistant I care about is not listed?
Then we do not measure it. An engine appears on this page only when a production adapter exists and real prompt runs exercise it. We would rather lose the comparison checkbox than route a well-known brand name to some other model and let you believe you are seeing that assistant.
How many engines does my plan include?
Engine access is a plan lever: Free includes 2, Essentials 4, Professional 5, and Growth 6 — every engine we support. Running the same prompt on more engines is also what raises confidence in a finding, so the ladder is about evidence quality as well as coverage.
See what these engines say about you
Start with a free audit of your site, or talk to us about monitoring your brand across every engine on this page.