AI NativeRunning a scan produces a result you read on the run page. This guide covers what that page shows, how to tell a mention from a recommendation, and how to trace any number back to the sources behind it.
The run page
Every scan has its own run page. While the scan works, the page updates through its states, from queued to running to done, so you are never staring at a frozen screen wondering whether anything is happening. When the scan finishes, the page fills in with the result.
At the top are the headline numbers for this scan: mention rate, recommendation rate, sentiment, and the count of gaps it found. From your second scan onward, each number is shown against the previous scan of the same mode, so you see the change, not just the level. That before-and-after is the point of scanning on a cadence: it is how you prove a fix worked.
Mention versus recommendation
The two numbers people most often confuse are mention and recommendation, and the run page keeps them apart on purpose.
- A mention means the answer named your brand. You showed up.
- A recommendation means the answer actually put you forward as a choice, the assistant suggested you, not merely listed you.
Being mentioned is not being chosen. An answer can name you in passing while recommending someone else, and that gap, mentioned but not recommended, is usually where the work is. The run page shows both so you can see the gap, not hide it inside a single visibility score. There is a separate guide that goes deeper on the difference.
The attack queue
A scan does not just report; it ranks the work. The gaps it finds are turned into an attack queue, an ordered list of the specific cells where AI does not yet recommend you, sorted so the highest-value, most winnable problem is at the top. Each item can be marked open, in progress, or done as your team works it, so the queue doubles as a record of what was acted on. AI Native does not silently fire a new scan when you mark something done; the impact only shows when you explicitly re-measure, so the change you see is real and not fabricated.
Sources behind the answer
A number you cannot trust is a number you will not act on, so every result traces back to evidence. You can open the underlying answers a scan read and the sources that shaped them, the pages and domains an AI engine drew on when it formed its response. This matters because the sources are diagnostic: if a competitor keeps appearing because one authoritative page recommends them and nothing equivalent recommends you, that page is the lead. Reading the sources turns a low recommendation rate from a complaint into a to-do list.
Accuracy on branded scans
When a scan covers branded prompts, the ones where a buyer names you directly, it also fact-checks those answers against the facts you verified in the Brand Truth Studio. The run page shows how many answers were checked and how many stated something inaccurate, so a confident but wrong answer about your brand surfaces as a fixable problem rather than slipping past unnoticed.