Free checklist

The AI fact-checking checklist

How to catch AI mistakes before they catch you. A practical guide to verifying AI outputs.

When AI is most likely to be wrong

Specific dates, numbers, and statistics
Niche or highly technical facts
Recent events (after the model's training cutoff)
URLs, web links, and email addresses
Quotes attributed to real people
Legal, medical, or financial specifics
Academic citations and references

The 5-step fact-check

Run through these before you use any AI output for anything important.

1

Does it sound too perfect?

If the AI gives you a very specific statistic, a perfectly fitting quote, or a suspiciously convenient fact, treat it with extra scepticism. The more specific and convenient the claim, the more likely it was fabricated. AI is confident even when it is wrong.

2

Cross-reference the key facts

Pick the 3-5 most important claims and verify them with a quick search. You do not need to check every sentence. Focus on the facts that matter most: statistics, dates, names, and any claim you plan to share publicly or use in a professional context.

3

Ask the AI for sources

Tell the AI: "Provide sources for the claims you just made." If it cannot provide real, verifiable sources, or if the links do not work, treat those claims as unverified. Real sources have real URLs that load real pages.

4

Use Perplexity for research tasks

Perplexity.ai provides sourced answers by default. Every claim is linked to a real source you can check. When accuracy matters more than creativity, use Perplexity instead of ChatGPT or Claude for research.

5

Trust your own expertise

If something does not sound right based on your own knowledge and experience, it probably is not right. AI can be confidently, convincingly wrong. Your expertise is the final filter.

The trust traffic light

Use this to decide how much checking an AI output needs.

Green: trust more, check less

Brainstorming and idea generation. First drafts and outlines. Reformatting or restructuring your own content. Summarising documents you provided. Code structure and boilerplate. Grammar and spelling corrections. Explaining concepts in simple language.

Amber: verify the key facts

Blog posts and articles. Presentations and slide content. Reports and proposals. Competitor research. Product descriptions. Social media content with factual claims. Email campaigns with specific details. Job descriptions and company profiles.

Red: always verify everything

Statistics and data points. Legal information or contracts. Medical or health advice. Financial guidance or projections. Academic citations and references. Anything published as fact under your name. Content about real people. Regulatory or compliance information.

Before you publish or send: the final check

Have I verified all statistics and specific numbers?

Do all dates and names check out?

Are any quotes attributed correctly and accurate?

Do any URLs or links actually work?

Does the content match my own knowledge and experience?

Would I be comfortable if someone asked me to prove this?

Have I removed or rewritten anything I could not verify?

Want to get more from AI?

We help businesses and individuals use AI with confidence. Here are a few ways we can help you go further.

Visit us at smashyourai.com

Follow us for daily AI tips and practical guides