Effective Follow-up Prompts

By Mike Caulfield

How to Use

Step 1: Navigate to Google's AI Mode in your search.

Step 2: Input any claim or statement you want to verify. No editing required – just paste it as is.

Example claim:

"When archeologists first encountered the Dogon people, they described the White Dwarf Sirius B, despite the star not being visible to the naked eye – and claimed they had learned about it through an ancient visitor."

Step 3: While the initial AI response is often sufficient, use these follow-up prompts when you need accuracy, depth, or nuance

Simple & Battle-Tested

I've tested these methodically across a wide variety of situations and they work reliably. Many are designed to be memorable even without a copy/paste file. I'll describe the thinking behind these (vs. more ornate prompts) and how they work (mostly because when you ask LLMs to sort evidence they find more relevant facts and come to better conclusions).
  • Facts and misconceptions about what I posted
  • Facts and misconceptions and hype about what I posted
  • What is the evidence for and against the claim I posted
  • Look at the most recent information on this issue, summarize how newer evidence, reporting, or other information shifts the analysis (if at all), and provide link to the latest info
  • Find me a link to the original source. If not available, find a link to the closest thing to the original source.
  • Give me the background to this claim and the discourse on it that I need to understand its significance (and veracity).

Good when it works

When this works it's amazing, but for reasons I don't understand about 1 out of 7 times it results in a denial to return a response.
  • Where did this claim come from? Use the I in Caulfield's SIFT Method (I for Investigate the Source) to do a lateral reading analysis of what the various people involved with this tell us about the claim. Find Better Coverage (F) to see what those most "in the know" say about the claim, subclaim, and assumptions of the original post, highlighting and commenting on any disagreement in the sources.

Under development

Experimenting with these but haven't tested near enough (I like to test a lot and methodically)
  • What's the overarching claim of what I posted? What is the evidence for and against?
  • Categorize the elements of this event as unprecedented, unusual, noteworthy, or normal. It's OK to have empty categories.
  • Are there any definitional or metrics/measurement issues I should be aware of to interpret this event, study, or evidence? Any terminological confusions to flag for the public?
  • Identify presuppositions and attempt to source evidence for (or against) each of them, presented in a table
  • List all compelling evidence and expert/witness testimony and categorize it as being either for or against the claim
  • What did the study or article actually say, what evidence did it present, and what can the evidence show or not show?
  • What would be a more reasonable claim given the evidence? For example, if a claim is overstated what is a more reasonable estimate? Give specific numbers, quotes, and sources where possible.
  • Read the room: what do a variety of experts think about the claim? How does scientific, professional, popular, and media coverage break down and what does it tell us?

Output Formatting

Some simple output formatting follow-ups, edit as needed
  • Five bullet points, short sentences, summarizing the above, with sources
  • Summarize the above
  • The above, but easier reading level, with more popular (less scientific) sources
  • Translate for non expert
  • Boil the above down to 7 short bullet points with sources, but provide expert links

Part of a method under development by Mike Caulfield