All AI use cases Better decisions

Get 3 new AI use cases every weekday

One full use case with its copy-paste prompt, plus 2 quick picks. Free, and one click to unsubscribe.

You're in. Watch your inbox on weekday mornings.

Make smarter decisions

Catch the claims your AI research stopped backing with sources

It marks each claim Supported, Partly supported or Not in the sources, then searches fresh for the weak ones.

About 3 minutes Just copy and paste

Published . Steps as of October 2026.

How do I know if my AI research answers are actually sourced?

Ask the AI to audit its own thread: list each claim, the source behind it, and a strict verdict, then search again for the weak ones. Perplexity fits best when you're already researching there, since every answer shows its sources. ChatGPT, Gemini and Claude work the same way when web search is on. If lots of claims fail, start a fresh thread and ask your key question on its own.

A picture made with AI to show the idea.

Your prompt

Fill in your details. Skip any you like.

ChatGPT opens with your prompt already typed in. Perplexity and Gemini can't be filled in from a link, so those buttons copy your prompt: paste it into the new chat.

The prompt that gets copied

Before I rely on this conversation, audit it. Go back through every answer you've given me in this thread about [YOUR TOPIC, e.g. which heat pump to buy]. Make a table with three columns: the claim, the exact source that supports it, and a verdict (Supported, Partly supported, or Not in the sources). Be strict: if a claim goes beyond what the linked page actually says, mark it Not in the sources, even if it sounds right. Then search the web again, fresh, for every claim marked Partly supported or Not in the sources, and tell me what you find, with new links. Finish with a short summary: what I can trust, and what I should still check myself before I [MY NEXT STEP, e.g. book an installer].

After a few follow-up questions, AI answers can drift past what their sources say. Paste this at the end of a research chat to get every claim checked against its source and searched again.

Step by step

  1. Finish your research chat as usual in Perplexity or any AI with web search
  2. Paste the prompt below as your last message
  3. Open the new links it finds for anything marked Not in the sources
  4. Start a fresh thread with the key question if many claims fail

What you'll get

A claim-by-claim table with a verdict for each, fresh sources for the weak claims, and a short list of what you can trust.

Made for
Everyone, Consultants, Small business owners
Works in
Perplexity, ChatGPT, Gemini, Claude

Frequently Asked Questions.

How do I know if my AI research answers are actually sourced?

Ask the AI to audit its own thread: list each claim, the source behind it, and a strict verdict, then search again for the weak ones. Perplexity fits best when you're already researching there, since every answer shows its sources. ChatGPT, Gemini and Claude work the same way when web search is on. If lots of claims fail, start a fresh thread and ask your key question on its own.

Which AI tools does this work in?

It works in Perplexity, ChatGPT, Gemini and Claude. On this page, ChatGPT and Claude open with your prompt already typed in. No paid plan is noted for this one.

How long does it take?

About 3 minutes. It's rated "Just copy and paste". There are 4 steps, listed above.

What do I need before I start?

Just your details for 2 blanks: your topic and my next step. You can skip any, and your AI will ask for them.

What will I get back?

A claim-by-claim table with a verdict for each, fresh sources for the weak claims, and a short list of what you can trust.

Find more like this in Better decisions, For small business, For consultants and Claude prompts.

More ways to make smarter decisions.

See all 56 ways to make smarter decisions