AI NativeEach product in AI Native has its own overview page. This is the first stop after a scan and the screen most teams check regularly.
What a product is
In AI Native, a product is one distinct offering inside a brand. A brand might have many products: a loan product, a savings product, an insurance product. Each product has its own competitors, personas, prompt tree, and scan history. Measurements do not bleed across products.
You reach a product's overview from the sidebar product list or from the Products index, which shows a grid card for each product with headline recommendation rate and competitor count.
The KPI tiles
The top of the overview page shows four tiles.
Recommendation. The fraction of unbranded prompt answers where your product was the top pick. Trend line shows movement across your scan history. Benchmarks at 45% (caution) and 62% (strong).
Mention. The fraction of unbranded answers where your product was named at any position. A product can be mentioned without being recommended.
Sentiment. Average tone score across branded answers. Ranges from negative (below zero) through neutral to positive. Benchmarks at 0.05 and 0.35.
Open gaps. The number of persona-stage cells where your product is not yet the recommended choice. This is how many gaps are in the opportunity queue, not a rate.
Each tile shows the current value and a small sparkline. If the tile has a recommended action attached, it appears in a small note beneath the number.
The auto-insight banner
If the scan found a high-priority signal, a banner appears above the tiles. Signals are ranked: a negative branded answer that also contains a factually wrong claim comes first, then failing brand guardrails, then the top ranked opportunity. The banner links directly to the relevant sub-page.
The two surfaces
Below the tiles, the overview splits results into two surfaces.
Visibility. Questions the AI answers entirely in-chat, with no need to visit a site. The product's success on these prompts is a measure of how well AI represents you as a category answer.
Execution. Questions where the user is moving toward an action (booking, purchase, comparison). These prompts require AI to recommend a path and often lead somewhere. Performance here reflects whether AI is pointing buyers toward your product at the point of decision.
The two surfaces have separate scores and separate action suggestions.
The outcome ladder
A small section shows the distribution of unbranded answers across the outcome states: absent, mentioned, listed option, recommended, recommended first. The percentage in each state tells you immediately if you have a discovery problem (high absent), a presence problem (mentioned but low recommended), or a positioning problem (recommended but rarely first).
Per-engine comparison
A small table compares your unbranded recommendation rate across the AI engines included in the scan. If performance varies strongly between engines, that points to where to focus improvement work.
Sub-page navigation
The product overview links to six sub-pages: matrix (persona-stage heatmap), answers (full answer list with filters), provenance (citation source analysis), accuracy (fact-check of branded answers), sentiment breakdown, guardrails compliance, and the why ledger (why AI rates you as it does). There are also pages for AEO audit, execution readiness, and A/B experiments.
Navigation tabs appear in the sidebar under the product name when you are inside a product section.
Running a new scan
Two scan buttons are visible on the overview page for strategists, admins, and owners. Demo scan runs a synthetic pass with no AI calls and no credits. Live scan queries the real AI engines and deducts credits. The estimated cost is shown next to the button and calculated from your current prompt count, scan surfaces, and the platform's per-answer rate.
If your credit balance is too low for a live scan, the button shows a warning and links to the billing page.
Auto-scan schedule
A dropdown on the product overview lets you set an automatic scan cadence: off, monthly, weekly, or daily. When enabled, the next scheduled scan date is shown next to the dropdown. Scheduled scans always run live. If credits are insufficient when the scheduled time arrives, the scan is skipped.
Related docs
Questions
What is the difference between mention rate and recommendation rate?
Mention rate counts every answer where your product appears by name at any position. Recommendation rate counts only answers where your product is the top choice returned. A product can have high mention but low recommendation if AI knows it but prefers others.
Can I see the individual AI answers behind a tile?
Yes. Click the Answers tab on the product sub-navigation. You can filter by stage, persona, engine, class (branded/unbranded), outcome state, accuracy, and sentiment to drill into specific subsets.
What are the benchmark thresholds?
The recommendation tile benchmarks are 45% (caution) and 62% (strong). The sentiment benchmarks are 0.05 (neutral boundary) and 0.35 (strong positive). These are platform defaults reflecting median performance across the product's category.
Why does my product show "No scan yet"?
No completed scan exists for this product. Either a scan has not been run, or an in-progress scan has not finished. From the overview page, click Live scan or Demo scan to start one. The page refreshes when results are ready.
What does "engine comparison" show?
It shows your unbranded recommendation rate broken out per AI engine for the latest scan, alongside the rate from the previous scan for comparison. Large differences between engines point to where a targeted content or presence fix would have the most impact.
How is the auto-insight banner ranked?
The priority order is: (1) a negative branded answer that is also factually inaccurate, (2) a failing high-severity brand guardrail, (3) the highest-scored open opportunity from the matrix. Only the top-priority signal is shown at once.
Do the two surfaces use different prompts?
Yes. The platform classifies each prompt at onboarding time as either contained (visibility surface) or action (execution surface) based on the question's intent. The surfaces are not separate prompt sets you manage, they are a view over the existing prompt tree.