A to-do list for AI visibility, graded by evidence
Every action names the answers behind it and how strong the evidence is that the change helps. Work the list, mark it done, and see what moved.
- All eight AI engines on every paid prompt run
- A 95% interval on visibility and citation rates
- Every run priced before it starts
Opening soon: sign-up opens with live measurement, after our provider pilot.
- Prompt gapApple is named more for earbuds to use with an iPhone24 of 29 answers against 2Experiment
- Source gapTravel publication cites competitors, not you6 answersCorrelation
- Fan-out gapNo JBL page shown for "best earbuds for iphone"Searched in 12 answersCorrelation
- Wrong claimAnswers overstate the Tour One M3's battery with noise cancelling on3 of 17 checked answersWeak
RookAtlas turns measurements into actions: prompt gaps where a competitor is clearly named more, source gaps on sites that cite competitors, searches engines run that no prompt covers, cited pages that broke or went stale, claims that contradict your facts, and crawler access problems. Each carries an evidence grade (official, experiment, correlation, vendor or weak), and weak ones never count toward a score.
Five grades, stated plainly
A grade says how strong the evidence is that this kind of change moves AI visibility: from the engine maker's own documentation down to anecdote, which is kept out of every score.
- Official, experiment, correlation, vendor, weak
- The study or document behind each
- Weak items never count toward a score
- OfficialThe engine's maker documents it.
- ExperimentShown in controlled tests, though results vary by engine.
- CorrelationSeen alongside visibility in large studies; not proven to cause it.
- Vendor dataReported by a tool vendor from its own logs or tests.
- WeakAnecdotal; kept out of every score. Never counted.
Only gaps the numbers support
A prompt gap needs at least six answers and a 95% interval for you wholly below the leading competitor's. A few unlucky answers don't raise one.
- Non-overlapping intervals, not a hunch
- Counts in words, never inflated percentages
- Priority by how many answers you're missing
- Apple
- JBL
- Bose
- JBL
Open, done, resolved
Actions live as a list per brand. Mark one done or dismiss it; when the numbers no longer show the gap, it resolves itself, and it keeps its history across recomputes.
- Kept by kind and key across runs
- Done, dismissed or resolved by the data
- Linked to drafts and outreach
How it works
- Step 1
Measure
Regular runs on your prompts and competitors.
- Step 2
Get the list
Gaps, page health, wrong claims and access findings, ranked, each graded.
- Step 3
Act and check
Draft, reach out or fix, then watch the gap close in the numbers.
What's inside
Prompt gaps
Prompts where a competitor's interval sits clearly above yours.
Source gaps
Sites citing competitors but not you.
Fan-out gaps
Searches engines run that no prompt asks.
Broken and stale pages
Your cited pages that fail or are over a year old.
Wrong claims
Answers contradicting your fact sheet.
Access findings
Crawler, directive and rendering problems from the audit.
What it can't do
No grade promises a result. A grade says how strong the evidence is that this kind of change moves AI visibility in general, not that it will for you. After you act, RookAtlas compares answers before and after as an observation.
Works with
All featuresQuestions
Prices, credits and plans are on the pricing page .
What do the grades mean?
Official: the engine's maker documents it. Experiment: shown in controlled tests, with results varying by engine. Correlation: seen alongside visibility in large studies. Vendor: a tool vendor's own logs or tests. Weak: anecdote, kept out of every score.
Is there an AI visibility score?
Not an opaque one. Actions are ranked by their measured gap and graded by their evidence; any composite score would publish its formula with its parts and interval beside it.
How fresh are the actions?
They're recomputed from the last 30 days of answers after runs, and an action the data no longer supports resolves on its own.