Field note
AI got a fact about you wrong. The order of operations to fix it.
You can't edit the model. You can change what it reads. Here's the sequence: calm, in order.
An AI assistant says something false about your company: wrong founding year, a product you don't sell, a price that's years out of date. The instinct is to argue with the chat. Don't. Here's the order that actually works.
1. Confirm it's real, not a fluke. Ask the same question a few times, across more than one assistant, in fresh sessions. Answers drift between sessions, so one bad reply isn't a pattern. Fix patterns, not noise.
2. Find what it's reading. Ask the assistant for its sources, then actually open them. The error usually traces to a specific page (an outdated profile, a wrong directory entry, a third-party article), and sometimes the citation goes nowhere at all. The wrong source is your target.
3. Fix it at the source, not in the chat. Correct the authoritative pages: your own site (in plain text, with valid structured data that states the fact), and the third-party sources you can influence. You're not editing the model. You're changing the evidence it draws from.
4. Re-measure, and wait. Assistants update only after they re-read the web. Re-run your check over the following weeks and watch the answer move. Expect weeks, not minutes.
One honest limit: some sources you don't control, and no fix is instant or guaranteed. But the order matters: most people skip step 1, panic at step 2, and never reach step 3, where the actual fix lives.
This is the "Correct" in the method. See the overview.
New here? Start with the overview → · By Denes Csaszar · written in the open