Nothing announces itself when this breaks. A readiness regression produces no error, no alert and no drop you could attribute — which is why it is normally found long after it mattered.
One blanket Disallow added while chasing a scraper takes out GPTBot, ClaudeBot and PerplexityBot at the same time. The site still works perfectly. Nothing in your analytics changes shape for weeks.
A theme or framework update can shift copy from server-rendered HTML to client-side JavaScript. Humans see no difference. Models stop being able to parse the page at all.
A plugin deactivates, a template is edited, a migration drops a field. Schema types vanish from pages that had them, and nothing on the site looks broken.
A staging domain, a www variant, a CDN rewrite. Each is a routine change, and each can tell every AI system and search engine that the authoritative copy of your content lives on a host you do not control.
Every one of those is observable from outside, costs nothing to re-read, and is invisible unless somebody is looking on a schedule. That is the entire product.
01
Every signal we read last month, re-read, with changes marked. Added, removed, altered.
A month where nothing changed is a valid and useful result, and it is reported as plainly as a month where something broke. A report that has to find something wrong to justify itself will always find something.
02
Per-agent crawler directives, Content-Signal, llms.txt, the .well-known endpoints, schema types, canonical, server-rendered content, headings, Open Graph.
Stored verbatim with the date attached. An observation is not editable after the fact — an edited observation is an opinion, and the point of a dated record is that you can go back and ask why a decision was made.
03
Named explicitly, every month, with the reason. A blocked path, an unreachable endpoint, a JavaScript-rendered section.
Silence is not a result. A missing finding otherwise means either the signal is fine or we never got to it, and those are indistinguishable from the outside.
04
When a reading moves in a direction that matters, you get told what moved, what it was before, and what it now permits or prevents.
No severity score. The change itself, quoted, so your developer can act on it without translating anything.
There are a lot of AI visibility products now. Most of them are solving a different problem to this one, and a few are solving it in ways worth knowing about.
The cheapest honest thing in this category, because it only promises what can be read.
If three months pass with no meaningful change on your site, we will tell you the monitor is not earning its fee and you should cancel. A subscription that nobody will ever advise you to leave is not a service.
The monitor only makes sense once there is a baseline to diff against. The audit is that baseline.
You probably should buy one of those as well, and we will say so. They answer "am I appearing in AI answers", which this does not. This answers "did the things that govern whether I can appear change on my own site this month" — which none of them track, because it is unglamorous and cannot be turned into a score. The two are complementary, not competing.
Because we would be handing you a number we cannot explain. Our whole position is that a figure with no stated method is not evidence, and reselling somebody else's composite score would be exactly that with our name on it.
Largely yes, and we will show you how during the audit. What you are paying for is that somebody does it on a schedule, keeps the history, and notices the diff. Most teams genuinely intend to and do not.
You get a report saying nothing changed, which is the correct answer and takes us very little time to produce. We would rather send you that than invent a finding to look busy.
Monthly by default. We do not pretend to daily monitoring, because the signals here do not move daily and a daily cadence would mostly bill you for noise. If you are shipping platform changes weekly, tell us and we will discuss a tighter cadence and what it would actually cost.
Yes, at no additional charge. It runs on the same reading we need anyway to do the rest of the work.
Get the baseline first — the audit is what makes a diff meaningful. If you have already had it, the monitor starts from your existing reading.