Get it in, track it down, follow up

Mike Caulfield, creator of the SIFT fact-checking methodology and co-author of Verified, developed this three-step framework as a practical approach to using AI for information verification. It's designed to be a habit rather than expertise: something anyone can apply without becoming a prompt engineer.

"Get it in" means submitting the claim to an AI with web search enabled. Don't agonise over phrasing, just get something in front of the model. The biggest barrier to checking things is the inertia of not doing so.

"Track it down" means following the AI's response back to actual sources. Don't accept the summary. Pick one claim from the output and verify it has a real, traceable origin. This is where you catch hallucinations and assess whether the sources are credible.

"Follow up" means treating the first response as a starting point, not a verdict. Ask for another round. Request a sources table. Ask what experts disagree about. The value comes from iteration, not a single query.

The framework sits alongside AI tools like Caulfield's own Deep Background prompt, but works with any model that has web search. His goal is to shift people from being passive consumers of AI summaries to active investigators who use AI to map the information landscape and then navigate it themselves.