What the Free Growth Audit Actually Checks

The growth audit takes a domain and runs seven checks against it. This post is the list, in plain terms, so you know what it reads before you give it your URL. It is the same engine that produces the Brand Index, pointed at a domain you type instead of a checked-in registry.

The seven checks

The two most people have not checked

Agent access. Six stores in our public registry return 403 to every request, from an honest crawler user agent and a desktop Chrome user agent alike. None of that appears in their own analytics, because the request never reaches their origin. That is its own post, and it is the first thing worth knowing because every other check downstream depends on the store being readable. The audit reports what it found; the visibility baseline guide is where you go to fix it, since robots.txt and CDN bot categories are the two layers this collector reads.

AI visibility. The questions are unbranded on purpose, derived from your own catalogue rather than from a template, because a question naming your brand guarantees a mention and measures nothing. If you come up, you came up because the assistant volunteered you.

What the audit refuses to say

This is the part worth reading before the feature list, because it is what separates a measurement from a lead magnet.

• Not-assessed is a real state. When a collector fails or never ran, the report says so and locates the failure with us. It does not score you zero and let you assume the worst. We shipped the opposite behaviour once: a store was told it refuses automated browsers when the truth was our screenshot vendor being rate limited.
• A check returns ok only when it observed the thing it claims. Not when a tag exists that points at the thing. We used to score `og:image` on the tag being present, so a page whose card renders blank scored the same as one that renders.
• A zero is not reported until the matcher has been shown capable of a hit. A brand can score zero because assistants did not name it, or because our matching was broken. Those look identical on a dashboard and they are not the same fact.
• Counts refer to distinct products. We once published '104 of 102 product titles' had an issue, which is arithmetically impossible and was on the page for anyone to see.

All four are drawn from the five documented failure classes, every one of them a defect this engine actually shipped. They are published with the instance that produced each, which is the only reason you should give any weight to the claim that they are fixed.

An audit that cannot fail honestly is a sales document. The useful question is not what it checks, it is what it says when a check does not work.

What it does not do

• It does not need access to your ad accounts. Everything above is read from public surfaces: your storefront, your catalogue, the Meta Ad Library, and the assistants themselves.
• It is one observation, from one vantage point, on one day. Edge configurations change, assistants change, and your catalogue changes. A single run is a reading, not a verdict.
• It does not tell you what your revenue would be if you fixed things. We do not have that number and neither does anybody selling you one.

The reason this exists in the shape it does: most of what determines whether an assistant can recommend you is invisible from inside your own business. A blocked request logs nothing. An unnamed brand produces no session, no bounce, no line on any chart. The audit is a way of asking the questions your own instrumentation structurally cannot answer.