AI quotes your market's numbers. Who checks the source?

WHAT HAPPENED

Trakkr traced 1,285 stats from five AI assistants

Trakkr asked ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews for numbers, then followed each citation back, up to five pages deep. One stat, 44% of salespeople give up after one follow-up, had no study behind it. All five assistants said it.

THE NUMBER

40%

of traced stats reached a study that says the same thing

Another 8% reached the study, but the number or context had changed.

THE REAL PROBLEM IN COMPANIES

What I see in companies: nobody owns the number AI quotes

1. 01Old figures about your market sit on blogs and roundups, and AI repeats them.
2. 02The real data is buried in a PDF or behind a form that crawlers can't read.
3. 03Nobody checks what AI says about the company, so wrong numbers go unseen.

I'VE ALREADY SOLVED THIS

One of Brazil's largest cosmetics companies: a report AI can cite

1. 01Annual report rebuilt: server-rendered HTML in 3 languages, Schema.org/JSON-LD, llms.txt
2. 0217 AI crawlers allowed by name, with an in-house log collector measuring everything
3. 0315 days: 1,427 AI reads from 8 engines, 553 real questions, 60 citation clicks

MY TAKE

Measure instead of assuming.

If AI repeats a stale number about your market, you only find out by checking. Publish clear, dated, machine-readable primary data. Then measure how often AI reads it, answers with it and sends people to it. Without that, you are guessing.