Free AI Visibility Audit — See What AI Agents Recover
Free AI Visibility Audit
Before you fix anything, see what an AI agent actually gets from your homepage today. This is the same instrument behind every number on this site, offered free at the single-model tier.
What you get
- One AI model fetches your homepage cold, with no framing and no hints about what to look for.
- A recovery score: how much of what a buyer would ask actually came back correct.
- The specific list of facts it could not find. Pricing is the single most common gap we measure. On eight real client sites across 48 recorded visits, not one produced a price, and that finding reproduced across three separate passes (Notes 12, 13).
What you don't get
The free tier shows you the problem; it does not fix it. No fix package, no generated surfaces, no monitoring. That is the paid tiers, starting at a one-time 149 USD Audit + Kit. See the full breakdown on our AI SEO services page.
The limits, stated plainly
- One audit per domain every 30 days.
- One model, one run. The paid audit uses six models and repeats the test, which is what makes a result stable enough to act on (Note 13 reproduced the same headline finding across three independent passes; single-model, single-run results carry more noise).
- Self-serve since 2026-08-11: the run starts when you submit and the result appears on this page in about five minutes, with a permanent link you can keep.
What you get back
This is the shape of it. The run below is a real one, on a live trade contractor's site, before and after a fix was installed; your free audit is the "before" column of the same instrument, on your own site, with your own questions.
| Buyer question | Before | After | |
|---|---|---|---|
| How long have you been in business? | 1/5 | 5/5 | was missing from the page |
| Do you do free estimates? | 0/5 | 4/5 | was missing from the page |
| Are you licensed? (license number) | 5/5 | 5/5 | already at ceiling |
| Where do you work? | 5/5 | 5/5 | already at ceiling |
| What does it cost? | — | — | deliberately not published; unchanged |
Measured 2026-07-29, five models per arm, one run (Note 30). Your report is that first column plus the list of everything the model could not determine, which is the part people actually act on.
How to run it
Two fields, and the run starts itself. Nothing is passed to a third party, no account is created, and your network address is never stored — the queue keeps a salted fingerprint of it so one connection cannot flood the list. The email is kept with the request so the result stays attributable to whoever asked; it is used for nothing else unless you also join the waitlist.
The form is the path. If your browser will not run it, write to
orders@aigentcandor.com with the site you want audited and a person runs
the same instrument by hand.
What happens after
If the score is what you expected, you now have a number instead of a guess, worth knowing on its own. If it isn't, the same instrument scales up: a six-model audit, a generated fix built from your own signed-off facts, and a before-and-after re-run against the frozen baseline, so you see the receipt, not just a promise. Details on the full package and the monitoring dashboard.