Home / How it works
How checkthis checks a Reel
checkthis is a free, automated fact-checker for Instagram Reels. You send a Reel to @checkthis.chat in a DM; it replies in a minute or two with what is true, what isn't, the sources, and the trick the Reel uses.
What happens to a Reel
- Listen and read. The speech is transcribed in its own language (English, Hindi, Tamil and other languages), and the caption and on-screen text are read.
- Find the claims. Every checkable statement is pulled out, along with what the Reel implies without saying it.
- Find the original. If the Reel is a clip of someone speaking, checkthis looks for the full video and compares what was said right before and after the cut.
- Check the sources. Each claim is searched in published fact-checks, established news outlets and official records. A quote only counts if it is found word for word in the source.
- Name the trick. The persuasion technique the Reel relies on is named in plain words, with the Reel's own words, and why the logic doesn't hold. See all 36 tricks.
- Flag what's concerning. When the sources show a lapse or questionable action by anyone, an agency, a party or a company, it is listed with that side's own response.
The verdicts
What it claims isn't true, according to reliable sources.
Real footage or facts, twisted into a story they don't support.
It leaves out something that changes what it means.
A real fact, made bigger than the sources show.
Some of it is right, some isn't.
The main points check out; some details are inflated.
The facts are real, but they don't lead to its conclusion.
It accuses someone, and nothing backs the accusation: no case, investigation, witness or report.
Nobody reliable has confirmed it.
An acquittal or a closed case: not proven, which is not a finding that it didn't happen.
Nothing wrong or misleading was found.
Satire, a skit or commentary, with its factual premise checked.
Neutral by design
The same rules apply whichever side a Reel favours. A party's own statement is shown as its position, never as proof, and that includes a government, a police force or a company explaining its own conduct. Motives and secret deals are never asserted without evidence, and neither are their absence.
Red flags
When the record shows someone in public office ducking the accountability a democracy expects (no press conferences, refusing questions, skipping Parliament, withholding records), or an institution's statement contradicted by the evidence, the check flags it, whoever it is.
Editorial standards
- Automated, with sources. Checks are made by AI models from public sources: published fact-checks, established news outlets and official records. Every fact links to its source, and quotes are checked word for word.
- Fake media needs grounds. A video is called AI-made or edited only when the creator admits it, an independent check shows it, the original footage is found, an AI tool's mark is visible, or the scene is impossible. Someone in a video calling it fake is not proof.
- Opinion is labelled as opinion. The reasons a creator gives for someone's conduct are their reading; the facts they cite are what get checked.
- An accusation needs something behind it. An accusation or label pinned on someone (foreign-funded, paid, staged, anti-national) is "No proof yet" only when something backs it: a complaint, a named witness, a document, an open investigation. With nothing behind it but the say-so of those who make it, it is "Baseless", whoever it targets.
When it gets something wrong
Every check has a "This looks wrong" button in the DM and a form on its page. Reports are reviewed: when a verdict was wrong, the method is fixed and every affected check is re-run, and checks are re-run as new reporting appears. Every verdict change is listed with its date in our corrections log. Results link to every source so you can verify them yourself.
Who runs checkthis, and how to reach us: About.