How Dropstore measures AI visibility

The basic unit is one saved prompt run against one configured AI provider at a recorded time. Results from different providers or different runs are not assumed to be identical.

• Mention: the response names a tracked brand or product.
• Recommendation: the response presents that brand or product as an option for the stated need.
• Citation: the response links to or identifies an external source.
• Position: the order in which a tracked entity appears in the analyzed answer.
• Share of voice: the tracked entity's presence relative to named competitors in the selected prompt set.

AI output can vary because models, retrieval indexes, prompt interpretation, source availability, and provider behavior change. A trend across a stable prompt set is more useful than one screenshot, but it still describes only the measured sample.

Observed AI answers are combined with Shopify catalog data and live storefront evidence. Technical and content audits identify discrepancies or missing context. Suggested changes must rely on supplied store facts and remain reviewable before application.

• Dropstore cannot observe every response shown to every user.
• A correlation between a store change and a later answer does not by itself prove causation.
• Crawler access and valid schema do not guarantee indexing, citations, or recommendations.
• Provider outages, rate limits, and model changes can affect scheduled observations.
• Sentiment, mention, and competitor extraction can require human review when an answer is ambiguous.

Define prompts before evaluating changes, keep the prompt set stable long enough to compare results, review the original answers, and separate technical eligibility from external authority. Decisions should be based on several observations and business relevance rather than a single composite score.