See through phishing
before it sees you.
PhishGuard AI inspects every site, link, QR code and download the moment you encounter it — and tells you, in plain language, exactly why something looks dangerous.
One extension, every entry point.
Phishing doesn't only arrive as a suspicious link — it hides in look-alike domains, urgent login forms, QR codes on posters, and files disguised as invoices. PhishGuard AI watches all of it.
Real-time site analysis
URL structure, length, and obfuscation patterns are checked the instant a URL is scanned.
Brand impersonation detection
Catches look-alike domains and typosquats of banks, payment providers and government portals, and names the brand it's mimicking.
Page & form behavior (Coming soon)
[Planned] Will read the DOM for password fields, OTP prompts and urgent language.
QR-code inspection (Coming soon)
[Planned] Will decode any QR code rendered on a page and check its destination.
Download protection (Coming soon)
[Planned] Will flag double extensions, mismatched file types and risky archive contents.
Webmail link scanning (Coming soon)
[Planned] Will highlight dangerous links directly inside Gmail and Outlook web.
From page load to protection, in milliseconds.
Every URL passes through the same pipeline — a Random Forest ML model trained on 60,000 real URLs, scored 0–100 and explained before you see a warning.
Capture
The extension reads the URL, DOM structure and any links, QR codes or downloads on the page.
Analyze
A RandomForest ML model scores dozens of lexical features extracted directly from the URL.
Explain
Feature importances turn the model's output into the plain-language reasons you actually see.
Act
Depending on the score, PhishGuard AI allows, warns, or blocks the page — you can always continue anyway.
Never just "trust us."
Every verdict comes with the specific signals that drove it — ranked by how much each one contributed to the score.
Feature attribution
Model feature importances show exactly which signal moved the score.
Plain-language reasons
No jargon — "this domain was registered 3 days ago," not a raw feature name.
You decide the last step
High-risk pages are blocked by default, with an explicit "continue anyway" if you disagree.
Install PhishGuard AI in 10 seconds.
Free, lightweight, and runs quietly in the background. You'll only hear from it when something's actually wrong.