AI visibility

When someone asks AI for a recommendation, does your product enter the conversation?

Unbranded Portuguese buying prompts, repeated answers and per-product mention vs recommendation counts with the exact answer behind each count.

Explore sample data Open the labeled sample

Sample available; live runs need provider setup

Gengine · sample

Demo report (sample fixtures)

Illustrative example — not a client result

For the brand asking whether its exact product enters an AI recommendation conversation

One screenshot cannot reveal repetition, failure rate or whether the model merely mentioned the brand instead of recommending the exact variant.

Read the prompt and recorded answer behind every count, with product-level attribution and a visible denominator.

See the workflow

Make the evidence tangible

Use the controls to move through a deterministic sample. No customer result or live provider data is implied.

Recorded answer · prompt 2 · repetition 3SAMPLE

Prompt

“Which laptop is suitable for video editing and mobility?”

Among the options, the Aurora Pro 14 is recommended for buyers prioritizing mobility and 32 GB of memory. Compare battery life and local support too.

Illustrative example — not a client result · API records may differ from consumer chat interfaces.

A reproducible question set

Five buying prompts, repeated three times

Portuguese buying prompts are reviewed before the run. Branded and unbranded intent remain distinguishable in the report.

Illustrative example — not a client result

1Choose brand, subcategory and up to 3 products + 3 competitors

2Review 5 prompts × 3 repetitions before running

3Failures excluded from the denominator, coverage shown

Exact attribution

Mentioned is not recommended

Mentions, comparisons, recommendations and citations have separate labels and traceable excerpts. Publishers and generic phrases are excluded from product rankings.

Illustrative example — not a client result

1Choose brand, subcategory and up to 3 products + 3 competitors

2Review 5 prompts × 3 repetitions before running

3Failures excluded from the denominator, coverage shown

Visible denominator

Failures never disappear

Only valid completed observations enter the recommendation rate; failed or unknown rows stay visible in coverage.

Illustrative example — not a client result

1Choose brand, subcategory and up to 3 products + 3 competitors

2Review 5 prompts × 3 repetitions before running

3Failures excluded from the denominator, coverage shown

What you take away

Useful output for the next decision

Recorded answers

Prompt, repetition, original answer, provider, model and timestamp.

Product metrics

Recommendation numerator/denominator, mentions, comparisons and citations.

Review queue

Ambiguous family-to-variant matches remain unresolved until reviewed.

Before you start

Straight answers to practical questions

Does this reproduce the consumer ChatGPT interface?

No. Live diagnostics use a configured provider API and are labeled with that provider and model. Consumer interfaces can differ.

Can I explore it without a provider?

Yes. The sample is clearly labeled and derived from fixed fixtures. A live run stays blocked until setup is verified.

Start with evidence you can review

Sample available; live runs need provider setup

Explore sample data