Findings · 2026-08-20 · By VSNARY | Emmanuel Orta · 0 views

AI visibility tools that grade you 100% and still sell you the fix

When the call to action fires regardless of the result, the result was never the point.

Somebody ran crawlcheck.io through Adobe's AI Content Visibility Checker. The result was a clean pass: the agent view and the human view matched on essentially every row. Our homepage is server-rendered, so there is nothing hiding behind JavaScript for an agent to miss.

The report said this:

Citation Readability Score: Only 100% of your content is visible to AI agents

Only 100%. Directly above it, a red button: “See how Adobe Brand Visibility fixes this.”

Fixes what? The page scored full marks on their own instrument. The word “only” is a template that assumed a deficit, and the call to action fires whether or not one exists.

This is not a cheap shot at a typo #

The underlying measurement is a good one. Comparing raw HTML against the rendered DOM is genuinely useful, and it catches a real failure: content that exists only after JavaScript runs is content most crawlers never see. It is a check we have scoped and not yet shipped, and they ship it as their headline metric. Credit where it is due.

The problem is what the tool does with a good result. A measurement instrument that cannot report “you are fine, do nothing” is not an instrument. It is a lead form with a chart attached.

What we send instead, and why the shape differs #

The nearest thing here is a different question asked with a different instrument. Rather than compare one fetch against a rendered version of itself, the page is fetched as each named crawler alongside an ordinary control, and the section reports whether they all received the same thing. Divergence between identities is the finding; sameness is a pass with nothing to sell.

The two approaches catch different failures and neither subsumes the other. A render gap is the site withholding content from every machine equally. A divergence is the site withholding content from particular machines — an edge rule keyed to a user-agent, a challenge page served only to a named crawler, a shorter body for one identity than another. A site can pass either check and fail the other, which is why calling either one a visibility score is the category error.

Passing one of these says nothing about the other. A single number that claims to cover both is a marketing artefact.

When a row cannot be measured #

A row with no evidence either way is reported as unmeasured, never counted as a zero. That is the same discipline as refusing to fire a call to action on a clean result: the instrument must be able to return “nothing to report here”, or every reading it gives is worth less. A site that refused us scored 15 until that rule was applied properly — the low number was measuring our access, not their site.

Why we care about this specific failure #

Our own pricing page carries this line, and it predates the Adobe report:

Nothing a licence unlocks changes a measurement. The moment a payment can move a result, every result here is worth less — including the ones you would be paying for.

And our reports carry this one:

A row that could not be measured says so and scores nothing — it is never counted as a zero.

Those are not marketing lines. They are the reason the numbers mean anything. A scanner that always finds a problem has told you nothing when it finds a problem, because you already knew what it was going to say.

What a clean result should look like #

Our own FAQ answers the question directly, and the answer is deliberately bad for business:

What if my site is already fine? Then you will know in about a minute and pay nothing. That is the likeliest outcome for a well-built site.

That is the whole discipline. If the machine layer is clean, the correct output is “clean” and no invoice. A tool that cannot say that is measuring your willingness to buy, not your site.

The report, row by row #

The word only is a template that assumed a deficit. The button fires whether or not one exists.

The honest close #

We came out of their report at 100%, and we are not going to pretend that proves much on its own — it is one page, on one instrument, measuring one thing. But it did tell us something useful about the category we are in, and it cost us nothing to learn. The same category error shows up in share-of-voice scores, where the number describes an API’s answers rather than the product people actually use.

If you run us and your site is fine, we will say so, and you will not hear from us again.

Questions this post answers

How can a tool give a perfect score and recommend a fix?

When the recommendation is attached to the page rather than to the result. If the call to action fires regardless of what was measured, the measurement was never the point.

How should I read any visibility score?

Ask what was fetched, from where, and as whom; whether the tool can grade the seller's own site the same way; and whether the score would move if you fixed what it recommends. A grade that cannot go down is not a grade.

Related findings

Comments

Comments are read before they appear. Nothing is published automatically, and no account is needed.

Writing about this? Facts, live figures and marks — every number on that page is dated and traceable to a scan.

All findings · The dataset · How the dataset works