What We Test
For each audit, we define buyer-intent prompts around one product category. The prompt set is small by design, but it covers different buyer behaviors.
| Prompt type | Question it answers |
|---|---|
| Open discovery | Does the product appear when the buyer asks for options without naming it? |
| Feature-specific | Does the product appear when the buyer names requirements such as free tier, setup time, compliance, or integrations? |
| Named comparison | How does the AI compare the product against direct competitors? |
| Alternative search | Does the product appear as an alternative to a better-known competitor? |
| Final choice | Which product does the AI choose, and why? |
AI Surfaces
We use available consumer AI surfaces and disclose access status. If a surface is blocked, login-gated, or cannot produce a usable answer, it is marked as unavailable. We do not invent missing outputs.
What We Measure
- Whether the product appears.
- Which competitors appear.
- Where the product appears in ranked or comparative answers.
- Whether the description is accurate.
- Which claims are stale, too specific, or unsupported.
- Which source types appear to shape the answer.
- What public evidence may be missing.
What Findings Mean
A finding is directional evidence, not a guarantee. If a product is absent from open discovery, it may have weaker default category visibility. If it appears only when named, the model may understand it but not retrieve it unaided. If a model misdescribes a feature, the public product facts may be ambiguous, stale, or over-inferred from third-party pages.
What We Do Not Do
- We do not guarantee AI ranking improvement.
- We do not fake citations, reviews, or third-party mentions.
- We do not bypass login or human-verification gates.
- We do not modify the customer website in the first audit.
- We do not claim that one test represents every model or geography.
Deliverable
The early-access audit includes 5 buyer-intent prompts, up to 4 available AI surfaces, target product vs. 3-5 competitors, inclusion and ranking observations, description accuracy review, source and evidence observations, a prioritized evidence-gap roadmap, and one follow-up question by email.