AI fact checker that shows its sources for every claim
A viral post, a statistic in a client deck or a quote in your draft article can be wrong, out of context or years old. This prompt has your todo.is agent break the text into checkable claims, look for the original sources, rate each claim and explain what is actually true, with links you can open yourself.
The prompt
- Fact-check this: [TEXT, LINK OR FILE]. Break it into separate factual claims (numbers, dates, quotes, "studies show" statements). For each claim, search for the original source: official statistics, the actual study, the full speech or document, or reliable news reports. Rate each claim as true, mostly true, misleading, false, or can’t verify, and explain why in 1 or 2 sentences with the source links. Point out numbers that are outdated, quotes taken out of context, and claims where the source was misread. Put the most serious problems first. I need this for [PURPOSE], so [EXTRA INSTRUCTIONS].
What to change
- [TEXT, LINK OR FILE]: Paste the text, a link to the post or article, or attach a document, slide deck or screenshot.
- [PURPOSE]: e.g. "a blog post I am about to publish", "a family group chat" or "a client presentation".
- [EXTRA INSTRUCTIONS]: e.g. "suggest corrected wording for each problem" or "give me a short, friendly reply I can post".
Example result
- Fact check · viral post about coffee and memory
- Fictional example. Text checked: "A Harvard study proved that 3 cups of coffee a day cut dementia risk by 65%. Coffee is now officially recommended by doctors for brain health. It works within a week."
- Claim 1: "A Harvard study proved..."
- Misleading. The study behind the post was an observational study from a European university, not Harvard. It found an association, not proof of cause and effect. (link to the study)
- Claim 2: "...cut dementia risk by 65%"
- False as stated. The 65% figure comes from an older study of one small group, and it applied to people drinking 3 to 5 cups in midlife. Larger later studies found smaller or mixed effects. (links to both studies)
- Claim 3: "3 cups a day"
- Mostly true in that several studies looked at moderate intake around 3 to 4 cups. It is not a proven dose.
- Claim 4: "Officially recommended by doctors for brain health"
- False. No major health body recommends coffee to prevent dementia. Guidance generally says moderate coffee is fine for most healthy adults. (link to a health authority page)
- Claim 5: "It works within a week"
- Can’t verify. No source found. The studies followed people for years.
- Corrected version
- "Some long-term observational studies link moderate coffee drinking with a lower risk of dementia, but they don’t prove coffee causes the effect, and no health body recommends it for brain health."
- Friendly reply you could post
- "Interesting topic! The studies are real but they show a link, not proof, and the 65% number is from one old small study. Here’s the original: (link)."
How to do it with todo.is
- Copy the prompt and paste the text, link or file you want checked.
- Say what it is for, so your agent knows how strict and detailed to be.
- Send it to your agent on WhatsApp or Telegram, or paste it into todo.is. Screenshots work too.
- Open the source links for anything important, and ask follow-ups like "find the exact page in the report".
Tips for a better result
- Ask for the original source, not an article about it. Numbers often change as they get passed along.
- Check the date. A true statistic from 2012 can be misleading today.
- Look for what the source does not say. Many claims stretch an "association" into "proof".
- For a whole article, attach the file and ask for a table of claims with ratings and links.
fact checker: FAQ
- How do I fact-check something online? Find the original source of the claim, check who published it and when, read what it actually says, and look for other reliable sources that agree or disagree.
- Can AI fact checkers be wrong? Yes. AI can misread a source or miss a newer one. That is why this prompt asks for links on every claim, so you can check the important ones yourself.
- Can it check an image or screenshot? It can read the text in a screenshot and check those claims. Telling whether a photo itself is edited or AI-made is much harder, and it will say when it cannot be sure.
- Can it check claims in other languages? Yes. Paste the text in any language and ask for the fact check in your language.
JavaScript is required to use the todo.is app.