DeepVerify Reviews and Details

Questions & Answers

1. What makes DeepVerify unique?

   Two things.

   First, DeepVerify checks claims while you are still watching or reading, not after. On YouTube it reads the captions as the video plays and marks each checked claim on the timeline, so you can jump straight back to the moment a statement was made.

   Second, it has a third verdict. Most tools give you true or false. DeepVerify labels a claim Uncertain when the evidence genuinely is not there, instead of forcing a verdict. In practice that label comes up often, and it is the honest answer.

   It also works on live audio, transcribing what is playing in a tab or coming from your microphone and checking it as it is spoken.

2. How would you describe the primary audience of DeepVerify?

   People who already stop mid-video to check something, and want that to be less work.

   In practice that has meant three groups:

   • Heavy YouTube and podcast viewers who watch a lot of explainer, health, and commentary content and have noticed how much of it is asserted without a source.
   • Students and researchers who need the sources, not just a verdict, and use the check history as a record of what they looked into.
   • Journalists and analysts who need to check a live stream or an interview as it happens rather than afterwards.

   It is built for the person who wants to see the evidence and decide for themselves, not for someone looking to be told what to think.

3. Which are the primary technologies used for building DeepVerify?

   Extension — Chrome Manifest V3, React, TypeScript, Tailwind CSS, Zustand for state, Vite for the build. The UI lives in a side panel rather than a popup so it can stay open while you keep browsing.

   Verification backend — a claim detection and fact-checking service that pulls sources and returns a verdict per claim, with a knowledge graph behind it for entities and relationships.

   Live audio — audio captured from the tab or microphone is downsampled and streamed over an encrypted WebSocket, then transcribed with Google Cloud Speech-to-Text. The audio is relayed in memory and is not stored.

   Visualisation — d3-force and Pixi.js for the claim graph view.

4. Why should a person choose DeepVerify over its competitors?

   Most fact-checking tools rate the source rather than the claim. They tell you a publication leans one way or has a poor track record, which is useful context but does not tell you whether the specific sentence you just heard is true.

   DeepVerify works at the claim level. It pulls out individual statements, checks each one, and shows you the sources it used, so you can disagree with a verdict and still see the evidence behind it.

   It also runs where the claims actually reach you. Video and live audio are where most people encounter unchecked statements now, and those are the hardest formats to verify by hand because you have to stop, remember the wording, and go search. DeepVerify does that part while the video keeps playing.