# VisiScan self-scan data card — September 9, 2026

## Purpose

This dataset records a VisiScan free scan of `https://www.visiscan.app`. It is published so readers can audit the observations behind the case study and measurement guide.

## Scope

- Scan ID: `cmtu0y8bg000004ibq7zdh632`
- Window: completed September 9, 2026 at 11:39:55 UTC
- Business profile produced by the scanner: `OTHER`, United States
- Questions: 5
- Engines with valid live answers: ChatGPT, Claude, Perplexity
- Samples per question and engine: 2
- Raw answer observations: 30
- Gemini: unavailable in this run and excluded from the denominator
- Source/provenance: all 30 rows are marked `LIVE`

## Fields

Each CSV row is one question–engine–sample observation. It includes the exact prompt and answer, engine, sample index, timestamp, mention classification and confidence, named competitors, cited domains, latency, and provenance.

## Summary

VisiScan was classified as named in 5 of 30 valid answer observations: ChatGPT 2/10, Claude 1/10, and Perplexity 2/10. Ten observations contained at least one cited domain. The product's published score was 8/100 because its versioned score uses question weights; the unweighted raw mention rate is 16.7%. These numbers answer different questions and should not be substituted for one another.

## Known limitations

- The profiler classified VisiScan as `OTHER`, creating generic prompts such as “What is the best business near US?” This is a real measurement-quality defect, not a representative SaaS prompt panel.
- The run covers one date, one locale, three engines, five questions, and two samples. It is a baseline, not a market estimate or trend.
- The export assigns the scan completion time where historical rows have no separate `observedAt` value.
- Mention classification is automated and should be checked against the preserved answer before publication or decision-making.
- No intervention or like-for-like rerun is included, so the dataset cannot establish causation or improvement.

## Reproduction

Run the same five prompts in fresh sessions, twice per engine, record unavailable providers explicitly, and compare only like-for-like engine and prompt cells. VisiScan maintainers can reproduce the export with:

`pnpm exec tsx scripts/export-public-ai-visibility-observations.ts cmtu0y8bg000004ibq7zdh632 public/downloads/visiscan-self-scan-2026-09-09.csv`

## License

The dataset structure and VisiScan annotations are licensed under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). AI-provider output remains subject to the relevant provider terms. Attribute the dataset as “VisiScan self-scan, September 9, 2026” and link to the canonical case study.