We Pointed Our Own Tool At Ourselves. We Scored 42/100.

Most vendors publish case studies about their best customer. We do not have one yet, so we did the next most useful thing: we pointed the product at ourselves and published the result without editing it.

The score was 42 out of 100.

What we actually ran

We recently shipped a feature called the AI Focus Group. It reads a website the way a crawler would, infers who that business sells to, generates a panel of synthetic customer personas grounded in evidence from the site itself, then has each persona walk the pages and report where they stop believing you.

These are simulations, not interviews. No real person was surveyed. They are directional, and we say so on every screen that shows them. But they are grounded in specific evidence from a specific site, and that turns out to be enough to be uncomfortable.

We ran an eight-person panel against visibilityradar.ai: a VP of Marketing, an SEO manager, a growth engineer, two founders, a growth lead, a solo indie founder, and a CMO. Then we read what they said about us.

Every single one stopped at the same place

Eight participants. Eight drop-offs. All at the stage labelled Trusts brand.

Not pricing. Not confusion about the product. The hook landed with all eight of them — several described the anxiety it names as "exact" and understood what the product does within seconds. Then every one of them stalled before paying.

Here is what they agreed on, independently:

The moderator's summary was blunter than anything we would have written about ourselves: the hook converts curiosity into free demos and nothing into revenue, because a faceless vendor asking buyers to trust a scoring methodology has given them no human and no track record to trust.

The competitive result was worse

We asked each persona who they would actually choose today. Our win rate was 25%.

Two would trial us. The rest went elsewhere, and the reasons were specific:

That last one is the fair hit. We sell a number. We had never shown what the number is worth.

Then we found the thing that actually stung

Our product tells customers to make their content readable and citable by AI models. One of its core recommendations is structured, crawlable FAQ content.

While preparing the panel's input, we fetched our own FAQ page the way a crawler sees it. The served HTML contained every question and not one answer. The answers were rendered only after a click, in the browser, in JavaScript.

So the AI models we exist to help brands win over could read our questions and never our answers. We were failing the exact test we sell.

That one is fixed. Answers now ship in the markup, with FAQPage structured data — the same markup we tell customers to add.

What we changed before publishing this

Publishing a bad score is not brave if nothing follows it. So:

What we have not fixed

We still have no customer case study, because we still have no customer result we are allowed to publish. Writing one we could not stand behind would fail the exact test this whole exercise was about. That gap stays open until it can be closed honestly.

We also have not restructured pricing or navigation, even though both could be better. The panel was clear that neither is what stopped anyone.

Why we are telling you this

Two reasons, and only one of them is noble.

The honest one: a tool that scores websites should be willing to be scored. If we had run this and quietly filed the result, the product would be a rhetorical device rather than a measurement.

The useful one: everything above is reproducible. Run the free scan on your own site and you will see the same machinery applied to you, including the parts you would rather not read. The number will not flatter you, which is the only reason it is worth anything.

Our score was 42. We would rather tell you than have you find out.