AI NativeYour first scan is done and numbers are appearing. Here is how to read them without getting lost.
The three headline metrics
Every product page opens with four KPI tiles. Three of them tell you your current position.
Recommendation rate. The fraction of unbranded prompts where AI named your product as its top choice. This is the headline metric. A rate below 45% is flagged as needing attention. Above 62% is considered strong for most categories.
Mention rate. The fraction of unbranded prompts where AI named you at all, at any position. This is broader than recommendation. A mention is a necessary condition for a recommendation, so a high mention rate with a low recommendation rate means AI knows you exist but does not prefer you.
Sentiment. A score derived from the tone of answers that mention you on branded prompts. Positive is above 0.05, strong positive above 0.35. If the score is below zero, AI is regularly saying something unfavourable when your brand is asked about directly.
The outcome ladder
On the product overview you will see a ladder showing how unbranded answers are distributed across states. The states from weakest to strongest are: absent, mentioned, listed option, recommended, recommended first. Your goal is to move answers up the ladder over time.
A product that is absent in most answers has a discovery problem. A product that is mentioned but rarely recommended has a preference problem. These call for different actions.
The persona heatmap
Navigate to the matrix tab. This shows recommendation rate broken out by persona (rows) and funnel stage (columns). Cells in the heatmap are colour-coded from weak to strong.
A blank or red cell is a gap: AI does not recommend you to that persona at that stage. These are your opportunities. The heatmap is usually the most useful thing to read after your first scan because it shows where the gap is, not just how large the total gap is.
The "Do this next" prompt
The product overview page shows a single plain-English action at the top based on the highest-priority signal in your current scan. If there is a negative branded answer that also contains a factually wrong claim, that appears first. If a brand guardrail is failing, that appears. Otherwise the top opportunity from the matrix is shown.
This is a starting point, not a complete action list. The action links to the relevant view so you can dig in.
Checking accuracy
Go to the Accuracy tab. This shows branded answers that have claims the system could check against your ground-truth facts. Answers marked "inaccurate" mean AI stated something about you that contradicts a verified fact in your profile. A negative answer that is also inaccurate is flagged as the highest priority because wrong facts in a hostile answer do the most damage.
If you have no facts set up yet, the accuracy tab will be empty. Go to Brand Truth and fill in a few facts using the interview form to enable this check.
What a demo scan result means
If you ran a demo scan, the numbers are synthetic. They demonstrate the interface and the data structure but do not reflect how AI actually answers questions about your brand. The word "demo" appears on every scan card and in the trend chart labels. Demo results are held in a separate series and never shown alongside live scan results.
Run a live scan to see real numbers. This queries the actual AI assistants and uses credits. You can see the estimated cost before running from the product overview.
Related docs
Questions
What does a recommendation rate of 0% mean?
It means none of the unbranded prompts in the last scan resulted in AI naming your product as its top pick. You may still be mentioned, just not recommended. Check the mention rate and the outcome ladder breakdown to see if you are present at all or present but not preferred.
Why do I have mention rate but low recommendation rate?
AI knows your product but is choosing competitors over you when asked to recommend. The matrix heatmap will show which persona and funnel-stage combinations are weakest. The provenance tab shows which domains AI is citing, which points to where competitor authority is coming from.
How many answers make up each metric?
The product page shows the answer count (n) next to the headline KPIs. For a standard scan with three runs across two engines, you might see 30 to 60 answers. Each prompt is asked once per engine run, so the count scales with the number of prompts, engines, and runs you configured.
Are demo scan numbers comparable to live scan numbers?
No. Demo and live are separate series and are never mixed. Trend charts draw from one series at a time. Switch to live to get real numbers.
What should I fix first?
The "Do this next" block on the product overview page points to the highest-priority signal. If that is empty or you want more detail, check in this order: negative AND inaccurate branded answers, failing brand guardrails, the weakest cell in the matrix heatmap.
My first scan shows very low scores. Is something wrong?
Low first-scan numbers are normal for brands without an established AI presence. The platform measures where you stand today so you have a baseline to improve from. The opportunity score on each cell tells you which gaps are worth closing first based on business value, gap size, and how winnable the position is.
How often should I re-scan?
The product page has an auto-scan option that you can set to daily, weekly, or monthly. For most teams, weekly is a good default while running content and optimisation actions. More frequent scans build a denser trend line.