Jiwon Jeong

@broadriverstudiosllc Thanks Richard, appreciate the good wishes. If you end up testing it, news and health videos surface the most checkable claims, so those give the clearest picture of whether it actually holds up.

@td2375671 Thank you Tommy. The timing is part of why we built it, there is simply more confident-sounding content now than anyone can check by hand. If you give it a run, we would rather hear where it gets things wrong than where it works.

@mersesarvari Thank you, that means a lot on launch day. If you do try it, the honest test is a video you already know something about, so you can tell when a verdict is wrong rather than just taking our word for it.

@thridy That's a fair summary, thank you. The "needs further context" part is the one we care most about getting right, so if you try it and a verdict feels off, send us the video and we will look at it.

@bhattbuilds Good question, and worth taking properly.

Gemini's YouTube integration and DeepVerify overlap, but they answer different questions.

Gemini answers the question you thought to ask. You have to already suspect something, pause, and go ask. DeepVerify surfaces the claims you didn't think to question, because it reads the captions and pulls out every checkable statement on its own.

Second difference is where the answer comes from. Ask a model about a video and the answer is largely produced from what the model already knows. DeepVerify retrieves sources at the moment of the check and writes the verdict only from what came back, then shows you those links so you can disagree with it.

Third, it is allowed to fail. When the evidence doesn't settle a claim it returns Uncertain and leaves it out of the score rather than guessing. A general assistant will usually give you an answer either way.

And each claim is pinned to a timestamp so you can jump to the moment it was said.

Honestly, if you just want a summary of a video, the built-in integration is fine and free. DeepVerify is for when you want to know which specific sentences hold up and what the evidence actually was.

Hi StartupBase 👋

We built DeepVerify because we kept doing the same thing: pausing a video, opening a new tab, typing in a claim that sounded off, and giving up halfway through the search results.

Checking something takes more effort than hearing it does. That gap is part of why bad information travels so easily.

So DeepVerify does the checking in place. It reads the captions or the page, pulls out the claims that can actually be verified, and puts a verdict next to each one with the sources attached.

One deliberate choice: it is allowed to say "Uncertain." Always returning true or false would look more impressive, but a fact-checker that is confidently wrong is worse than none at all. When the evidence isn't there, it says so.

Two things we'd genuinely like feedback on:

1. Where does it get things wrong? Send us the video or article and we'll dig in.
2. Is a sidebar the right place for this, or does it get in the way?

Free to install. Works on YouTube, articles, and live audio.