What AI Said.

The machines are answering questions about your business. Every day. Without you.

Buyers now ask ChatGPT, Google’s AI Mode, Claude and Perplexity — and trust the name the machine says. What AI Said captures what AI says about you, traces every answer back to the source that caused it, and hands you (or your engineer) the exact fixes. Then it re-tests until the thread runs clean.

What it does

Measure. Asks real buyer questions to ChatGPT, Google AI, Claude and Perplexity, stores every answer verbatim, and counts mentions of you and your competitors at exact character offsets in the transcript.
Diagnose. Every wrong answer has a cause of death: the stale page, the missing schema, the conflicting policy. Each finding is stamped with its evidence.
Fix. Diagnosis in, work orders out — one layer speaks merchant, the other speaks engineer.
Publish. Owner-approved facts become a hosted public profile with llms.txt, facts.json and JSON-LD — the formats AI systems actually read.
Re-test. Scheduled measurements show whether the fixes moved the needle, with the before/after on one page.

The one rule

Every number, measured — never estimated. Every claim, traceable to evidence you can read yourself. Every recommendation, born from something we actually observed. And where we cannot know — we say so, plainly.

Where it runs

In the browser at whataisaid.io (installable as an app), plus a Windows and Linux desktop app and a direct Android APK — all listed with real versions and SHA-256 checksums in the Field Kit section of this page.

How do I make my own website visible to AI?

The same way this page is: serve your content as real HTML, not an empty JavaScript shell; publish owner-stated facts in llms.txt and structured data (JSON-LD); keep them current. What AI Said measures whether that is working, finds what is not, and automates the publishing — this very page has its own llms.txt.

This plain-HTML version is intentional so any crawler or AI can read it. The interactive report loads with JavaScript.