Fake stats target AI citations

A post about publishing fabricated statistics to see whether large language models would cite them drew a 267-point discussion on Hacker News. The point for a business owner isn’t that every AI answer is false. It’s that a number can look credible in an AI-generated response without being true. One commenter said the made-up numbers had also appeared in an academic paper and a bank’s investment research, showing how a false claim can travel beyond the original post. The discussion is a useful reminder that an AI citation isn’t the same as verification.

If your team uses AI to prepare reports, compare vendors, or brief managers, check important claims against the underlying evidence before acting on them. Ask where a statistic came from, open the cited material, and confirm that it supports the claim being made. A source-checking process can help teams keep track of what AI says and what still needs checking.

The risk isn’t limited to dramatic errors. A fabricated number could be repeated after appearing in research or being cited by an AI system. Treat AI output as a starting point, not proof. When a claim could affect a decision, verify it with a source your business can trust.