Pull Every Claim Out of a Video
A 40-minute interview can hold thirty factual claims. Revenue figures, launch dates, “a study of 500 people”, “twice as fast as the competition”. Finding them means rewatching the whole thing with a notepad.
Paste the video link or upload the file, and ScreenApp’s AI goes through it and lists every claim as a table: the timestamp, the exact words, what kind of claim it is, and how you could check it. Click a timestamp and you are at the moment it was said.
Every claim, timestamped
Each row links to the moment the claim is made, so you hear it in context.
The exact words
The quote as it appears in the transcript, with the speaker when there is more than one.
Sorted by type
Numbers, dates, studies, comparisons and quotes, so you can check one kind at a time.
A way to check it
For each claim, the kind of source that would confirm or contradict it.
A Claim Extractor, Not a Fact Checker
This distinction is the whole design, so I will be blunt about it.
Gives you a verdict
Reads the claim, searches the web and says true, false or misleading. Fast, but you are trusting its search and its judgement, and you often cannot see why.
Gives you the evidence
Finds every claim and the moment it was said, sorts it by type and suggests where to check it. Slower to finish, but you can defend every conclusion you reach.
An AI that says “true” or “false” about a claim is guessing unless it checked a real source, and even then it can pick the wrong one. So this tool does not give verdicts. It finds the claims, shows you exactly where each one is, and tells you what kind of evidence would settle it. The checking is yours, and every claim keeps a timestamp so anyone can hear it in context.
What the Table Looks Like
These rows are made up to show the format. They are not taken from a real recording.
interview-with-founder.mp4
| Time | Claim | Type | How to check it |
|---|---|---|---|
| 02:14 | "Our revenue doubled last year." | Numerical | Annual report or filing |
| 05:38 | "This product launched in 2022." | Historical | Press release or archived product page |
| 09:12 | "The study involved 500 participants." | Research | The study itself, methods section |
| 14:50 | "We are twice as fast as the market leader." | Comparative | The benchmark behind it, if one is named |
Ask for changes in the same chat: only the numerical claims, only what one speaker said, or the claims grouped by topic.
The Kinds of Claims It Finds
Numerical
Revenue, growth, percentages, user counts, prices.
Historical
Dates, launches, "we were the first to".
Research
"A study found", "scientists say", survey results.
Comparative
"Faster than", "cheaper than", "the biggest in".
Attributed
What someone else supposedly said or did.
Causal
"X caused Y", "because of this policy".
Opinions and predictions are not claims you can check, so the AI flags them separately instead of mixing them into the list. “This is the best phone ever made” is an opinion. “This phone sold 10 million units” is a claim.
How to Extract Claims From a Video
- Paste the link or upload the file. Links from YouTube, TikTok, Instagram, Facebook, Vimeo and Loom links, direct audio and video links, PDFs and web articles work. Sign in with a free account.
- Read the claims table. ScreenApp transcribes the video and opens a chat that asks for every claim with its timestamp.
- Check what matters. Click a timestamp to hear the claim in context, then follow the suggested check for the ones you need to verify.
The AI watches videos up to 50 minutes long, and the AI reads video frames, not just the audio, so a statistic on a slide or a chart on screen is caught as well as what is said out loud.
Check a claim in three steps
Hear it in context
Click the timestamp. Half the problems are a claim taken out of its sentence.
Find the primary source
The filing, the study, the original statement, not an article about it.
Write down what you found
Note the source next to the claim, so the next person does not start over.
Start with the numerical claims. They are the easiest to check and the most often wrong.
Check Claims Across Several Videos
One video is useful. Five videos from the same speaker is where it gets interesting.
Put every interview, keynote or earnings call in one folder. ScreenApp has one AI chat that searches and answers from every recording in your library, or a single folder, so you can ask questions that no single video answers:
What It Can and Cannot Do
Good at
- Finding claims in long videos you do not have time to rewatch
- Catching numbers on slides and charts, not only speech
- Separating claims from opinions and predictions
- Comparing what a speaker said across several videos
Not built for
- Verdicts: it does not say whether a claim is true
- Live checking: it works on a recording, not while you watch
- Whole channels: one link at a time; playlist URLs are not imported as a batch
- Streaming services: Spotify and other DRM-protected streaming services cannot be imported
Video Claim Extractor vs Other Tools
| Feature | ScreenApp | Facty | FactScope | TranscriptAPI workflow |
|---|---|---|---|---|
| Gives true or false verdicts | No | Uses five verdict levels instead of true or false | Labels videos Real, Fake, Mixed or Story | No |
| Where claims come from | The video’s audio and frames | YouTube captions, auto-generated ones included | Not stated | YouTube transcripts |
| Uploads and non-YouTube links | Yes | YouTube only | Not stated | YouTube only |
| Questions across many videos | Yes | Not stated | Not stated | Not stated |
| Free option | 10 AI chats a month | 30 minutes of video a month | 2 checks | 100 free credits |
Sources, checked 2026-10-12: Facty on Product Hunt, FactScope Firefox add-on, ScreenApp pricing, TranscriptAPI fact-check workflow
- vs Facty: a browser extension for Chrome, Edge, Brave, Arc and Opera that reads YouTube captions, auto-generated ones included, searches the web and shows a verdict over the player. Good while you watch YouTube. It needs captions, and you get its verdict rather than the evidence.
- vs FactScope: a Firefox add-on that labels a video Real, Fake, Mixed or Story, with the sources it used. Quick for a gut check on one viral video.
- vs TranscriptAPI’s workflow: the closest idea to this page. It gives one claim per row with a timestamp link and a suggested way to verify it, and no truth rating, but you need an API key and an AI agent or automation tool to run it. ScreenApp does the same job from a link or a file, with nothing to set up.
Who Extracts Claims From Videos
Journalists and fact-checkers pull every checkable statement out of a press conference, interview or viral clip, then spend their time on verification instead of transcription.
Researchers and students list the claims in a lecture, documentary or debate and trace each one back to its source for an essay or a literature review.
Investors and analysts extract every number from an earnings call or a founder interview and compare it with the filings.
Marketing and legal teams review their own ads, webinars and product videos for claims that need substantiation before they go out.
Debate and media literacy teachers use the table to show students the difference between a claim, an opinion and a prediction.
