Not measured is not failed: how to read an AI visibility report

A signal marked not measured in an AI visibility report means the file could not be read at the moment of the scan, not that your site failed it. The distinction looks small and it changes how the whole report reads: a measurement failure asks nothing of you, a confirmed absence does. Mixing the two means paying to fix a problem that is not there, or ignoring the one that is.

Why should a serious report separate an absence from a measurement failure?

Because an unreachable file and a missing file produce the same blank, for opposite reasons. A timeout, a passing server error, a firewall that refuses the request: the scanner read nothing, but it also observed nothing. Scoring a zero in that case invents a defect. A Biencité scan marks the signal not measured and drops it out of the calculation instead of counting it as a failure.

The consequence shows up in every report: it states how many signals were actually verified out of the 19 in the reference list published on /en/methode. A score built on 15 verified signals and a score built on 19 do not describe the same thing, and you need to see which one you are holding before you decide anything. The rule has a cost, and it is deliberate: it makes it impossible to always hand over a round, reassuring, complete-looking number. A stated partial coverage beats a tidy wrong score.

What do the four verdicts mean, and why is the score alone not enough?

The four verdicts of a Biencité scan are invisible to AI, visible but not citable, readable and citable, ready for AI agents. They do not follow from the score: they follow a cascade. A site whose content AI assistants cannot read is invisible whatever the rest of the score says, even if it ticks the most advanced signals in the reference list. Order matters more than the total.

The cascade exists because the three axes measure three different problems that call for three different answers. A site can be perfectly visible and still not be citable: AI assistants read it, but nothing on its pages gives them anything to recommend you with. A comfortable overall score can therefore sit on top of an entire axis at zero. Read the weakest axis before you read the average, because the weakest axis describes your real ceiling.

Why can your platform make a signal impossible to earn?

Shopify, Wix, Webflow and Framer lock certain files at the root of the domain, and no setting on the merchant side unlocks them. Those signals are marked impossible on your platform rather than counted against your site, and they never appear in your action plan. You will not be handed advice your site cannot act on, and you will not pay for work that runs into a wall. The article Shopify, Wix, Webflow, Framer: what your platform actually lets you do for AI walks through that trade-off platform by platform.

A report that holds the impossible against you is not stricter, it is less useful. It lengthens the fix list, it inflates the quote, and it leaves you wondering for months why that one line never turns green.

Does a high score guarantee a citation from ChatGPT or Google?

No, and no tool can claim otherwise. Google says so in its own documentation for site owners, updated on December 10, 2025: the fact that a page meets every requirement, every best practice and every policy does not mean Google will crawl it, index it or serve it. An AI visibility scan measures eligibility, meaning the conditions without which a citation is very unlikely. It does not measure an outcome.

The stake is real all the same. Ahrefs compared 300,000 keywords between December 2023 and December 2025 and found that when an AI Overview is shown, the click-through rate of the top organic position is 58% lower than it is without one. The traffic that no longer arrives through the blue link is now decided inside the answer itself, and technical eligibility is the entry condition, not the winning ticket.

Should you trust a report that demands a file Google says you do not need?

In that same documentation, Google states that there are no additional requirements and no special optimizations needed to appear in AI Overviews or AI Mode, and that you do not need to create new machine readable files or AI text files. That is accurate, and it is true of Google. An honest report tells you so rather than leaving it out.

Answer engines are not all Google. ChatGPT, Claude, Perplexity, Gemini and Le Chat each read the web in their own way, and the Biencité weekly tracking queries those five surfaces precisely because they do not answer the same thing. A dedicated reading guide for AI will never make up for a home page that crawlers cannot read, however. That is exactly what the verdict cascade says: readability first, the agent layer last. A report selling you the agent layer while your first axis is failing is selling you the roof before the walls.

When is an AI visibility report of no use to you?

There are situations where the right answer is to wait, and an honest vendor names them before taking the money.

• Your site has nothing to cite. Three sentences and a contact form: no technical file will create the substance an AI could quote. The budget goes into writing first.
• Nobody is looking for you yet. If no outside source talks about your company, technical work will not stand in for a reputation that does not exist.
• Your platform is locked and you do not intend to move. Better to know the ceiling before paying to reach it.
• Your site is already ready for AI agents. Rescanning it every week produces nothing new, and the question becomes what AI assistants say about you, which is a different measurement.

Why is the weight of each signal not published?

The 19 signals are published, their weighting is not, and that is a deliberate choice. The weight of a signal shifts as generative engines change their practices, and publishing it would invite optimising the score rather than the actual visibility. The practical consequence fits in one sentence: an AI visibility score compares to itself over time, not to another site and not to another tool. Two vendors that do not check the same signals do not produce comparable numbers, and nobody gains from pretending otherwise.

What should you look at first when you open your report?

The verdict, then the coverage, then the weakest axis. The verdict tells you which step your site is stuck on. The coverage tells you how many signals the score was calculated from, and therefore how much it can be trusted. The weakest axis tells you where the first pound goes. The overall score comes last: it is the easiest of the four numbers to remember and the least actionable.

The scan runs read only, on the home page and a handful of public files, with no account and nothing installed. That deliberately narrow scope has limits of its own, set out in the article Our scan reads one page of your site. Here is why, and what that costs. A report that states its scope, its coverage and its blind spots is worth re-reading in six months. A report that only states a score is worth reading once.