What this tab shows
The Action Plan is where to start when you open a report. It answers two questions in one view: what is wrong with this page, and what should I do first?
You do not need to read every tab to know whether a page has a problem. The Action Plan tells you whether there is a hard blocker, a few things worth improving, or whether the page is broadly in good shape — and it turns each of those into a specific piece of work.
The tabs behind it are grouped. Evidence holds the measurements the plan is built from — AI Crawlability, Content Visibility, Agentic Browsing, Authority, Content Coverage, Prompt Fit, Prompt Discovery, and History. Experiments holds Lab on its own.
The verdict
The verdict leads: one sentence naming the biggest problem on the page and what to do about it. The fix for it sits directly beneath, as item 1, carrying its own button — so you can start on the most important thing without going looking for it.
The AI Visibility Score
The score (0–100) and its status sit beside the verdict. The status runs across five bands:
| Score | Status |
|---|---|
| 85–100 | Excellent |
| 70–84 | Needs work |
| 55–69 | Concerning |
| 40–54 | At Risk |
| 0–39 | Critical |
The score breaks into five bars — AI Crawlability, Content Visibility, Authority, Speed & stability, and (once the page’s content has been analysed) Answer Readiness. For what the score measures and how to read it, see Understanding your AI Visibility Score.
The number can never read better than the verdict. A critical finding caps the score at 69 and a warning caps it at 84, whatever the dimensions average to. So a page with four clean dimensions and one hard blocker shows 69 at most. If you see a score of 69 labelled Concerning next to a Critical verdict, that is the cap doing its job — not two parts of the report disagreeing.
This is the score for one page. The dashboard’s Site Health score is the same idea rolled up across every page on the site.
The plan
Item 1 sits under the verdict. Everything after it follows under Then, in order.
Each item names the job in plain language, says what the next step is, and links to the tab holding the detail behind it. Where we can draft the fix for you, the item tells you how many fixes are available and carries the button that starts them; the drafts open in Content Coverage.
The numbers are an order to work in, not a severity ranking. They are not sorted by how many points a fix is worth, because not every useful fix is scored. They are sorted by what you can act on:
- Blocking access or measurement problems
- Things that are ready to use now
- Work we can draft for you
- Things waiting on a fact only you have
- Template and hosting changes
- Everything else, for review
Why this order? next to the heading opens that explanation on the page itself.
If the plan is empty, the page is clean as far as this audit went — and the tab says so, together with any check that did not run, because a check we never ran is not the same as a check that found nothing.
Recommended improvements
Below the plan, Recommended improvements holds the optional work: changes that improve what the page already says. They are deliberately unnumbered, because none of them is blocking you, and they do not count toward your Content Coverage score.
Earlier drafts
Fixes you have already replaced collect under Earlier drafts, in a single line. Nothing is deleted; it is just moved out of the way of the current work.
Report signals
Report signals is a fold near the bottom, open by default. It holds the reference view of the report: the Citation Readiness path, and a one-line status for every dimension so you can jump straight to whichever tab needs attention.
The Citation Readiness path
The path lays the report out as five ordered steps a page has to clear on the way to being cited by an AI system. Each step shows a status tag — Ready, Needs work, or Critical (or No data / Not run / Untested when a check hasn’t run) — and links to the tab behind it:
- Findable — can AI crawlers reach and index the page? (AI Crawlability, including TTFB and indexability)
- Readable — does AI see the same content browsers do, from the raw HTML? (Content Visibility)
- Trustworthy — are the identity, schema, and proof signals in place? (Authority)
- Complete — does the page cover its topic completely enough to answer what AI asks? (Content Coverage — this is what produces your Answer Readiness score)
- Relevant — is the page a strong fit for the specific questions you want it to win? (Prompt Fit and Prompt Discovery)
The steps are in order for a reason: a page has to be Findable and Readable before Trustworthy, Complete, or Relevant can help it. Where Content Visibility asks can AI read the page?, Answer Readiness (the Complete step) asks can the page answer the questions AI is asking?
The path adapts to the page type. On identity pages — a homepage, About, or Contact — Complete and Relevant recede to Secondary, because for those pages “is the entity clear and trustworthy?” matters more than deep topic coverage. On product, comparison, and article pages every step is foregrounded.
How to use it
Work the list from the top. That is what the ordering is for. The plan has already done the triage the Citation Readiness path describes — access before readability, readability before content — so you do not have to work it out per page.
Read the verdict before the score. The verdict is the one that names a blocker.
Use Report signals when you want the map rather than the work. If you would rather see where the page stands across all five steps than be handed a list, that is what the fold is for.
Generate the fix. For content and structure findings — Content Coverage, Prompt Fit, Prompt Discovery, and Authority schema — Optimizations can draft the fix itself, ready to paste in and re-audit. Start it from the Action Plan item; read the drafts in Content Coverage. See Optimizations.
Re-audit when you’re done. A fix that works shows up in the numbers, and the page’s History tab records the change.
What the score is not
The score is a summary signal, not a prediction. A high score does not guarantee that your page will appear in AI answers — the quality and relevance of your content matters just as much. A low score tells you where to investigate first, not that the page is fundamentally broken.
For a detailed breakdown of any dimension, open the corresponding tab under Evidence. The Action Plan is where to start, not the full picture.