Tracked platforms
The six AI assistants RankX AI tracks, what each is good for, how the per-Website and per-prompt platform choice works, and why Google AI Overviews sits apart.
RankX AI tracks six AI assistants: ChatGPT, Gemini, Claude, Perplexity, Grok and Microsoft Copilot. All six are on every plan, with no engine add-ons, by policy. What changes between plans is how many prompts you track, never which assistants you may ask.
Which of the six a Website runs is a per-Website setting, and each prompt carries its own list on top of that. A prompt follows the Website until you give it one, and a prompt's list can only narrow the Website's, never widen it. So a question that only matters on ChatGPT and Gemini need not be asked of the other four, and does not cost you for them. Check both lists before reading any figure on this page: an assistant that was not asked has no verdict, which is not the same as a miss.
Google AI Overviews is a seventh surface RankX AI watches, and it is deliberately not in this list. It is measured a different way, and folding it in would overstate your assistant coverage by a sixth.
The six
ChatGPT is the one every reader recognises, and it covers the broadest range of everyday consumer and business questions. If you track one assistant closely, this is usually the one your buyers actually used.
Gemini is integrated with Google Search and Workspace, which makes it the assistant that matters most to brands whose buyers already live in Google. It is also the one whose behaviour is most entangled with your ordinary search presence.
Claude skews to nuanced research and document-heavy work, and is strong in B2B contexts. A considered-purchase brand often finds its clearest signal here, because the questions asked of it are longer and more specific.
Perplexity is AI-native search with real-time citations, so it is the clearest view of which sources are being surfaced alongside your brand. If you are working the citation side of this problem, Perplexity's answers are the most legible evidence.
Grok is xAI's assistant and is built into X, so it reaches buyers who ask their questions where they already follow the conversation about your category. Tracking it shows how that audience hears your brand described.
Microsoft Copilot sits in Windows, Edge and Bing, which means it reaches people who never chose an assistant at all: they typed into the search box that was already there. It answers from a search-grounded index, so it behaves more like a search engine with an opinion than like a chat product, and it is the assistant most likely to name a brand your Bing coverage supports.
Microsoft Copilot, and what differs about it
Three practical differences, and none of them is a caveat about quality.
Whether it runs is a per-Website choice. Like every assistant, Copilot is switched on or off in the Website's platform list, and a prompt can narrow the list further. Every assistant on the list adds the same step to the cost of each run, so a check that costs money stays a decision the setting leaves with you.
An empty Copilot column is unknown, not zero. This follows from the setting. If the assistant was not asked, no verdict exists, and a blank is the honest rendering of that. It is the same rule the product applies to any missing verdict.
A Copilot check settles after the run rather than inside it. The answer is read from the interface a person uses, and that read is asynchronous: a full check settles in roughly two and a half minutes rather than immediately. So a run can report Copilot as still on its way while every other assistant has already answered. That is a result you have not received yet, never a failure.
A night when Copilot's source is unavailable leaves your rates clean. Copilot is read through one route, so on a night that route is unavailable, RankX AI removes Copilot from that night's check set rather than recording a failed check. Your rates keep a denominator you can name, and a gap in the Copilot series is never mistaken for a miss.
Why per-assistant figures are the real figures
The six disagree with each other by design. They retrieve differently, they were trained differently, and they answer the same question differently on consecutive runs.
That is not a defect in the measurement, it is the shape of the thing being measured. So:
Read per assistant. A 40% mention rate on one and 10% on another is two findings. Averaging them to 25% describes neither.
Do not average a missing assistant in. A blank rate is unknown, not zero, and folding it into an average silently invents a measurement. It matters whenever a Website runs fewer than all six: an assistant you have not switched on has no data at all, and its blank stays blank.
Expect movement between runs. Two runs of the same prompt frequently return different brand lists. The reliable unit is a rate across a panel of prompts over several runs, not a figure from one.
Google AI Overviews is separate, and it is measured differently
| The six assistants | Google AI Overviews | |
|---|---|---|
| What it is | An assistant answering a question | A summary above a search results page |
| Measured by | The AI-visibility run, on your tracked prompts | The rank tracker, on your tracked keywords |
| Where it lives | AI Visibility | Google Results |
| Widened by adding | Prompts | Keywords |
Gemini and Google AI Overviews are both Google's and are still two different things. One is an assistant answering a tracked prompt in the scheduled run; the other is a summary above a results page, captured by the rank tracker. A brand can do well on one and badly on the other, and that gap is a finding rather than an inconsistency.
What every check costs
An AI visibility check is charged per prompt, per enabled assistant, per run, so your prompt list is multiplied by the number of assistants that are actually switched on.
On all six, twenty-five prompts is 150 checks a run; on five it is 125. Each assistant you add or remove moves every run by the same step, for as long as the setting stands. That is the whole arithmetic, and it is set out with the rest in credits and metering.
Prices are on the credit cost reference, which is generated from the product.
Turning an assistant off
You can disable an assistant for a Website if it genuinely does not matter to your buyers, and doing so removes it from every future run and from that run's cost.
Two things to weigh. It stops the spend on that assistant immediately, which is a real saving on a large prompt list. And it ends the series: a rate you stop collecting is a comparison you cannot make later, and the history you already have does not extend itself.
Where to go next
- The tracked platform reference, which is the generated table.
- AI Prompts, for the panel the assistants are asked.
- AI Overviews, for the Google surface.
Last updated
Answer accuracy
Whether what AI assistants say about your business is true. Seven verdicts, judged in code against the facts you have published rather than by a second model.
AI Readiness
Whether AI assistants can reach, read and understand your site. Five scored categories, the checks RankX AI actually runs, and why a score can be withheld.