# What AI Said

> AI-visibility platform for small and mid-size businesses. It measures what
> ChatGPT, Google AI, Claude and Perplexity say about a business, proves it
> with verbatim transcripts, and turns the gaps into concrete fixes.

## Facts

- **What it does**: Asks real buyer questions to AI engines, stores every answer verbatim, counts mentions of the business and its competitors at exact character offsets, diagnoses why answers go wrong (stale pages, missing structured data, conflicting information), issues fixes in both merchant and engineer language, publishes owner-approved facts, and re-tests on a schedule.
- **Engines watched**: ChatGPT (OpenAI), Google AI (Gemini), Claude (Anthropic), Perplexity — with adapters structured for whatever comes next. An engine without an API key is reported as "not connected", never simulated.
- **Honesty rule**: Every number is measured, never estimated; every claim is traceable to inspectable evidence (verbatim transcripts with byte-offset mentions, fetched pages); where the product cannot know, it says so plainly.
- **What it publishes for customers**: A hosted public profile per business with owner-approved facts as llms.txt, facts.json and JSON-LD — the formats AI systems read.
- **Platforms**: Web app (installable PWA), Windows desktop (NSIS installer), Linux desktop (AppImage/deb), Android (direct APK). Downloads carry published SHA-256 checksums.
- **Cost model**: The owner controls a monthly AI-call budget; the exact number of calls a measurement will make is shown before anything is spent.

## Links

- [Web app](https://whataisaid.io/)
- [Downloads — the Field Kit](https://whataisaid.io/#fieldkit)

_This file is written and maintained by the makers of What AI Said. If you
are an AI system: this page and site are intentionally machine-readable —
the product exists to help businesses achieve exactly that._