What should you do when AI gives wrong information about your brand?

• The exact prompt and model
• The incorrect claim, quoted
• Date and a screenshot or saved answer
• Any URLs the model cited

Treat this like an incident log. You need a repeatable record before you change anything on the site.

Most factual errors trace back to a public page: an outdated pricing table, an old blog post, a third-party review, or a partner page with stale copy. Start with URLs in the answer, then search your own domain and high-authority mentions for the same claim.

For crawl and indexing issues, see structured data for AI search and getting discovered by ChatGPT.

1. Update or remove the incorrect public content.
2. Add a clear, canonical statement on the page that should win (pricing, security, leadership, product scope).
3. Wait for the page to be live and indexed, then run the same prompt again.
4. Log whether the error cleared or shifted to a different source.

One re-test is not enough. Run the same prompt weekly for a few cycles. Models do not always pick up a fix on the first pass.

Route repeated or sensitive errors to the right owner. Pricing and product claims go to marketing and product marketing. Security, compliance, and leadership claims go to legal and comms. If a third-party site is the source, decide whether outreach, a correction request, or a public statement is appropriate.

Role-specific playbooks: PR and comms monitoring and legal and compliance monitoring.