1Evidence for every finding
Every result in a report links to what we actually saw: the page and the element, the AI answer, or the step in an agent test. If we cannot show the evidence, we do not report the finding.
We keep different kinds of evidence apart. Technical checks, AI answers and agent tests answer different questions, and one is never presented as another.
2Human-reviewed
Engineers review findings and every proposed fix before they reach you. Fixes are delivered as clear changes your team approves. Nothing is published to your site without your approval.
3Re-checked over time
After changes, we check again the same way, so a difference reflects your site and not a change in how we measured. Each report is dated, so you can compare like with like.
4Trends, not guarantees
AI answers vary from person to person and from day to day. We report how AI assistants describe your brand across many answers over time, not a single screenshot.
No one can guarantee a mention, a ranking, traffic or revenue, and we never do. We fix what is verifiably broken and show what changed afterwards. A change seen after a fix is reported as an observation, not as proof that the fix caused it.
5How results are labelled
We use plain words instead of a 0–100 score, because one number hides which problem you have.
| Label | What it means |
|---|---|
| Accessible · Partly blocked · Blocked | Find: whether AI can reach your pages. |
| Extracted · Partly extracted · Unavailable | Understand: whether AI can read your key facts, such as prices, services, hours and contact details. |
| No blockers found · Issues found | Use: whether the page has obstacles for agents, such as buttons without names. Only an agent journey test shows whether a task can be completed. |
| Observed | Your brand was named in the AI answers we looked at. |
| Not observed in this sample | Your brand was not named in the answers we looked at. It does not mean you are never mentioned, so we never write “not mentioned”. |
| Inconclusive | The evidence is not strong enough to say either way. |
| Not tested | That part was not included in this report. |
6Safe agent testing
- Agent tests never submit a form, book, pay, log in or enter personal data.
- They run on your website only after you confirm you own it, or on our own demo site.
- They stop at logins, CAPTCHAs and bot challenges, and report it.
- The free preview only reads public pages, the same way any visitor can.
7Sample data
Examples on this website use a fictional business, Harborline Dental Studio, and fictional competitors. Every screen that shows sample data is labelled as a fictional sample. We never show a real company as a demo subject without permission.
8Limits
- What one person sees in an AI app can differ from our results, because answers depend on the person, the time and the settings.
- The free preview is a quick check of public pages, not a full audit.
- Pages behind a login are not checked unless agreed in writing.
Questions about how we work? Contact us. How we handle data is in our privacy notice.