Method
How RankAI Links Its howItWorks Page to AI Visibility
RankAI's howItWorks page explains a two-step path: first it reads your public pages, then it observes how AI engines answer. Here is how the two steps connect, and what the method can and cannot show.
If you run a site and want to know how AI visibility gets measured, RankAI's howItWorks page lays out a two-step path: first the public pages of the site are read, then the answers from AI engines are observed. The public pages name ChatGPT, Perplexity, Gemini, Grok and Google AI. What makes the measurement checkable is the link between the two steps: without knowing which pages can be read, we don't know what we are observing in the answers.
From reading pages to observing answers
The howItWorks page does not promise rankings. It describes a flow: the public pages of the site are read, and on that basis RankAI observes the answers from the AI engines it names. That is a context step, not a guarantee of results. The distinction matters because AI visibility is not a single number: it depends on which engine we observe and on how we phrase the question.
A working method proposed by RankAI
What follows is a working proposal from RankAI, not a universal standard. Its purpose is to make the path checkable step by step.
- Check that the pages can be read. Checkable criterion: the public page is reachable and its content shows up in the reading, without depending on restricted content or anything behind a login.
- Connect the engines that need a connection. Checkable criterion: Claude and Grok are measured only after you connect them; until the connection is active, there is no observation to record for these engines.
- Keep observations separate. Checkable criterion: Gemini and Google AI Overviews are not the same observation, so they should be kept apart when you read the result.
- Record the question alongside the answer. Checkable criterion: every observation is tied to the question that was asked and to the engine that answered, so any comparison stays between like elements.
What it means to observe an AI answer
An API response is not a person's screen. When RankAI observes answers from AI engines, it does so through a technical channel: what we see is the returned answer, not the interface a user would see in their own browser. That limit should be kept in mind every time you interpret an AI visibility figure, because the observation is confined to what the channel returns.
In the same way, Gemini and Google AI Overviews are not the same observation: they are two distinct viewpoints and should not be added together or confused. The howItWorks page exists precisely to make this chain clear, from the page that was read to the answer that was observed.
The contribution of this piece is a practical way to read the howItWorks page without turning it into a promise. The useful point is to separate two questions that often get merged: "did the engine read my page?" and "what did the engine answer?" The first is about whether public pages can be read; the second is about observing the answers. RankAI holds the two steps together but does not collapse them into one. Anyone looking at the results should do the same: note which pages could be read at the moment of observation, which engine was queried, which question was asked, and whether that engine needed a connection. Only then does a comparison between two different observations stay honest. RankAI's working proposal, then, is to treat every answer as a snapshot with a stated context, not as an absolute visibility score.
The limit to keep in mind
The checkable limit is this: RankAI observes answers from the named AI engines through an API channel, and Claude and Grok only after you connect them. It is not a person's screen, and Gemini and Google AI Overviews remain distinct observations. For the full picture of the path, see the in-depth page.