Measured. Never guessed.
How a verdict is reached, what each number divides by, and what each one leaves out.
How does Inciteful measure AI visibility? In plain words.
Inciteful measures AI visibility by putting a client’s tracked questions to each AI engine, storing every answer with its sources, and giving each answer one of three states: named, read but not named, or not found. Every number on the dashboard and in a report is worked out from those stored answers, and anything that could not be measured is shown as unmeasured.
Measured, never guessed. The rest of this page is the detail.
- ChatGPT
- Gemini
- Google AI Overviews
- Google AI Mode
- Perplexity
- ClaudeCopilotAdd-ons
A check, a run, and a stored answer.
- A check
- One tracked question put to one engine. The full answer and its sources are stored.
- A run
- One scheduled pass over a client’s questions. It records its plan, so it can be called complete or incomplete.
- A change
- A change shows against the run before it, with the answer behind it.
Where buyers ask, and what each one shows.
- ChatGPT
- The answer ChatGPT gives a buyer, with the sources it cites.
- Gemini
- Google’s assistant. A separate answer from Google Search.
- Google AI Overviews
- The summary above Google’s results. A search with no Overview counts as not named.
- Google AI Mode
- Google’s conversational search, answered in full.
- Perplexity
- The answer and every source it cites.
- Microsoft Copilot
- An add-on. Copilot’s answer in Bing; a search it does not answer counts as not named.
- Claude
- An add-on. Anthropic’s assistant, with the sources it cites.
- Asked for the client’s market
- Setup records where the business sells, local, national or international, and the questions are written for that market.
Any other engine that returns no answer is recorded as not measured and left out of every rate. It is never counted as the client being absent.
Named is not the same as cited. So we keep them apart.
Named
The answer recommends your client by name.
For same-day appointments in Bristol, Harbour Dental is the most consistently recommended.
Counts toward their visibility
Read, not named
The engine used their page as a source, then recommended a rival.
Clifton Smiles offers Sunday appointments and a weekend emergency line.
The near miss, and the fastest win
Not found
The answer was built without them.
Redland Dental and Whiteladies are regularly named for whitening.
Needs a page worth citing
What each number divides by.
- Visibility
- Named answers over all answers in the latest completed run, branded questions excluded. A trend point is one run.
- Average position
- The mean place of the client’s own site in the answers’ source lists, where it is cited. Source order, not a search ranking.
- Sentiment
- The tone of each mention, from 0 to 100, only where there is tone evidence. No evidence shows no score.
- Buyer-journey stage
- Named answers over answers at each stage: discovering, comparing, choosing. Branded questions excluded.
- Change
- Measured against the previous completed run in the window. Every change names that baseline.
What the numbers leave out. By design.
- Measured after is not caused by
- When a page ships and visibility moves, we say what was measured after it. Never that the page caused it.
- Branded questions stay out
- A question that contains the client’s name is measured and shown, never counted in a headline.
- Unmeasured is not zero
- A dimension that could not be measured is shown as unmeasured. Never as 0, 50 or 100.
- Four kinds of data stay apart
- Observed runs, historical provider data, modelled estimates and generated text are labelled as what they are.
A change to the method is dated and shown on any chart that spans it. Since 12 October 2026, a search for which Google showed no AI Overview counts as not named, and since 14 October so does a search Copilot gave no answer for.
