How does the assistant recognize and position the named brand?
How doping.ai measures AI recommendations: from Probe to Classified answer
Every buyer Probe becomes a structured record of the assistant answer, recommendation outcome, competitors, and Citations. This guide shows how those parts move through doping.ai.
Each answer becomes a structured observation.
Each observation starts with a typed buyer Probe sent to a configured assistant. doping.ai stores each usable response with its Profile and Tracker context, then turns it into a Classified answer.
That record gives teams a precise unit to compare across Probe types, assistants, Trackers, and Runs.
One answer, complete context
Open the Probe, response, recommendation outcomes, competitors, and Citations together in Evidence.
The Profile and Tracker define the monitored context.
A Profile represents one monitored business context: its domain, brand, markets, competitors, and Trackers. A Tracker is a public homepage, product, service, collection, or solution page that can be checked.
Profile and Tracker context shapes the buyer questions and connects each answer to the public page a team can review and improve.
Six Probe types cover different parts of a buyer decision.
Which brands surface when the category is asked without naming the brand?
Does the assistant endorse the named brand for a buyer need?
Are material public claims about the named brand correct?
Does the brand or offering surface for a specific need without being named?
How does the named brand compare with relevant alternatives?
Monitoring keeps a Tracker's Probes and runs them again over time. A question's wording never changes: the Probe review can suggest same-type options instead, and each question is compared only with its own previous Runs.
Start with a baseline, then monitor a Tracker.
| Dimension | Free sample | Monitored Tracker |
|---|---|---|
| Scope | Homepage baseline | One selected Tracker |
| Probes | One question per type | The Tracker's Active Probes, as the set grows or shrinks |
| Assistant coverage | One assistant per Probe in rotation | All six assistants per Probe |
| Answers | One per Probe | Up to six per Probe Run |
| Cadence | One-time baseline | Up to weekly on Starter and Scale |
Enable monitoring for a Tracker to run its Probes repeatedly. Each Run sends one Probe to all six configured assistants, at most once per week. Results update as each Run completes, and a weekly report reviews what changed.
Each usable answer receives a recommendation classification.
Recommendation
Recommended, Mentioned, or Absent describes the audited brand or Tracker in the answer.
Recommendation describes whether the answer presents the business as a suitable choice for the requested buyer need. It does not verify the answer’s claims. Accuracy Probes ask whether material public claims about the brand are correct; their answers remain available for review and are left out of the Recommendation score, like Recall answers.
Competition outcomes add the alternative in the same answer.
Classification also records the exact competitor names observed in the answer and whether an alternative was recommended. The combination creates five Competition outcomes.
- 01Owned recommendation
- The Tracker is Recommended and no competitor receives a recommendation.
- 02Shared recommendation
- The Tracker and at least one observed competitor are Recommended.
- 03Recommendation loss
- The Tracker is Mentioned while an observed competitor is Recommended.
- 04Displacement
- The Tracker is Absent while an observed competitor is Recommended.
- 05Recommendation gap
- The Tracker is Mentioned or Absent and no observed competitor receives a recommendation.
Open Competition to rank observed alternatives and find the Trackers under pressure. Then return to Evidence for the exact answer behind the outcome.
Evidence shows how each result was formed.
A Classified answer stays connected to its Probe, assistant, recommendation state, reasoning, competitor mentions, and available Citations. Teams can move from a summary outcome to the exact answer and context behind it.
Citations show which cited or search-result URLs accompanied the answer. Together with stored reasoning and Profile context, they provide a direct path to the material behind each classification.
The measurement moves directly into product decisions.
- Overview summarizes the strongest patterns for the current Tracker.
- Competition shows which competitor brands appear and where they create pressure.
- Evidence opens the exact Probe, answer, recommendation outcomes, reasoning, competitors, and Citations.
- Website connects answer gaps to source-backed improvements for relevant Tracker pages.
- Trackers run the same Probes again as Probe units allow and send a weekly report.
- Radar adds recent, source-backed developments about observed Competitors.
Together, these views turn assistant answers into a repeatable workflow for review, prioritization, and monitoring.
For the Agentic readiness audit on 100 real sites, read What online shops tell AI agents.
Explore the measurement in the product.
Follow an answer from its outcome into Competition, Evidence, and Website.