Stop misinformation before it spreads. Verify claims directly on social media.
Verify is a browser extension that adds fact-checking to scoial media. With a single click, get instant verification results for any claim along with sources.
Hover over any post and hit Verify. Within seconds, you'll see:
- ✅ Claims entailed by credible sources
- ❌ Claims refuted with contradicting evidence
⚠️ Unclear claims flagged for further investigation
All results appear right on the post.
A user verifies a claim that a user posted about from a satirical article. Verify finds alternate, trustworthy sources that directly contradict it.
Verify checks if these quotes were actually said
Verify can find multiple documents with different results, allowing the user to check sources for themselves
Verify searches a wide range of reputable sources to ensure accurate verification. However, it excludes certain sites known for unreliable information.
- FactCheck.org
- Snopes
- PolitiFact
- Reuters Fact Check
- Wikipedia (with caution)
- Reputable news outlets (BBC, CNN, The New York Times)
- The Onion (satirical content)
- x.com (Formally Twitter)
- reddit.com (user-generated content)
- theguardian.com (due to frequent opinion pieces)
- foxnews.com (due to known biases)
-
Clone this repo
git clone https://github.com/EunoiaC/verify.git cd verify -
Set up your API keys — Create a
.envfile in the root directory:GEMINI_API=your_gemini_key_here GOOGLE_SEARCH_API=your_google_search_key_here -
Load into Chrome
- Open
chrome://extensions - Enable Developer mode (top right)
- Click Load unpacked
- Select the
chrome_extensionfolder
- Open
-
Navigate to Reddit and look for the Verify button
- A finetuned gemini-2.5-flash model processes the social media post to extract the main claims.
- Using the Google Custom Search API, Verify searches for relevant fact-checking articles from reputable sources.
- An embedding model compares the claims against the retrieved articles to determine which parts of the documents are relevant.
- Once we built the context, a deberta model finetuned on nli tasks classifies each claim as entails, contradicts, or neutral based on the evidence.
Found a bug? Have feedback? Open an issue
Verify: Because social media deserves better information.



