A gentle reminder as we all explore AI tools: use them, but use them thoughtfully. Even when an AI cites sources, it’s still up to us to confirm they’re real.

During a talk I gave last year on why AI is likely to create more jobs than it displaces, I asked ChatGPT for background material with references. Two citations turned out to be phantom papers. When I pointed this out, the model apologized and quickly offered five reliable sources—but only after I checked.

Lesson learned: always open the links and verify each reference yourself. If you’re using a reasoning model like ChatGPT o3, ask it for live links at the end of its answer. One click should take you straight to the actual paper or article, so you can judge the evidence directly. With that simple habit, AI becomes a powerful assistant rather than an unreliable narrator.