AI NativeEvery product page offers two scan buttons: Demo scan and Live scan. They run the same pipeline and produce the same layout, but they measure very different things.
What a demo scan does
A demo scan runs the full pipeline without contacting any AI engine. It injects pre-built fixture answers, scores them, builds the attack queue, and populates the charts. Every answer is clearly labeled with a "demo (synthetic)" badge on the run page so nothing is mistaken for real data.
Demo scans cost nothing. No credit hold, no settlement, no ledger entry.
The fixture answers show how the screen reads and what the gap-to-action flow looks like. They are not drawn from your product's real prompt set. Because demo results are kept out of the live trend, running them does not affect the recommendation-over-time chart.
Use a demo scan when evaluating the platform, showing a colleague around, or exploring before spending credits.
What a live scan does
A live scan fires your product's prompt set at the real AI engines and scores each answer. Everything on the run page reflects what those engines actually said.
Before the scan starts, an estimated cost appears on the product page as a credit chip. The tooltip shows the math: measured prompts multiplied by active surfaces equals answers, at a per-answer rate. That rate is the median cost of your own recent live scans of this product. If no history exists, it falls back to your workspace's recent scans, then the platform default. The basis is always disclosed.
When you click Live scan, the estimated amount is held from your balance. When the scan finishes, the hold is reconciled against the real measured cost. The final bill reflects the scan's real measured cost. If the real cost is lower than the hold, the difference is refunded. If the scan fails entirely, the full hold is returned.
The low-balance state
When the estimated cost exceeds your balance, the Live scan button switches to amber and shows the shortfall. Demo scan remains available. Top up at Workspace > Billing.
If auto-scan is scheduled and the balance is insufficient when it fires, that scan is skipped.
Errors excluded
Some engine calls fail or time out. Those are counted and shown on the run page as an "errors excluded" chip. Failed calls are excluded from every rate and score: the platform does not count a failed call as an absent answer.
The mode label stays with the run
Every run carries its mode permanently. The re-measure button shows the mode it will use. Trend charts and delta comparisons only pair runs of the same mode, so a demo baseline cannot be accidentally compared to a live result.
Questions
When should I use demo vs live?
Use demo to explore the platform or show a colleague without spending credits. Use live when you need real measurements to act on or to track progress over time.
Does a demo scan affect my live trend?
No. Demo results are excluded from trend charts and delta comparisons. They do not alter your recorded history.
What happens to my credits if a live scan fails?
The full hold is refunded. The ledger records this as "full refund: the scan failed, you were not charged."
How is the estimate calculated?
Prompt count multiplied by active surfaces multiplied by a per-answer rate derived from the median cost of your own recent live scans. The basis is shown in the estimate chip tooltip.
What does "errors excluded" mean?
Engine calls that failed or timed out are excluded from every rate and score rather than being counted as absent answers. The chip shows the count so you can decide whether to re-measure.