Free Chrome Extension

See through manipulative
websites in real time.

Dark Pattern Protector detects and highlights deceptive UI tricks as you browse — using blazing-fast DOM rules and Claude AI analysis.

🔍 11 pattern types detected ⚡ Simple & Expert mode 🤖 Powered by Claude AI ✓ Free & open source Your data stays in your browser
How it works

Built for
everyone.

Choose Simple mode for plain-English summaries, or Expert mode for technical detail. Either way, fast DOM rules catch the obvious tricks instantly and Claude AI catches the subtle ones.

1
Page loads
The extension activates automatically on every page you visit. No setup, no clicking — always on.
2
DOM rules scan
Eleven hand-crafted rules instantly scan for pre-ticked checkboxes, fake timers, cookie tricks, forced sign-ups, and more. Zero latency.
3
AI deep analysis
Claude AI reads page text to catch subtle manipulation — guilt-trip copy, vague privacy promises, suspicious reviews, and deceptive framing.
4
Your health score
Simple mode shows 🟢 Looks clean, 🟡 A few things to watch, or 🔴 Manipulative elements found. Dismiss anything you don't care about — per site or forever.
What we detect

11 tricks
in the playbook.

From sneaky sign-up checkboxes to fake countdown timers, we catch the patterns companies use to manipulate your choices — and tell you exactly what to do about each one.

☑️ Sneaky sign-up checkbox
Pre-ticked box silently signs you up for newsletters or data sharing
CRITICAL · Forced consent
🍪 Cookie banner trick
Large "Accept All", no easy reject option, or buried opt-out
CRITICAL · Forced consent
🔐 Forced sign-up
Makes you create an account just to browse or complete a task
CRITICAL · Access barrier
😬 Guilt-trip button
"No thanks, I prefer paying full price" — shames you for saying no
HIGH · Manipulation
🔀 Confusing opt-out wording
"Uncheck if you do not want to not receive…" — designed to confuse
HIGH · Confusion
👁️ Hidden opt-out link
Tiny text, low contrast, or near-zero opacity to hide the exit
HIGH · Obstruction
📉 Fake "only X left"
"Only 2 left!" — stock counts that are frequently fabricated
MEDIUM · False urgency
⏱️ Fake countdown timer
Resets on page reload — the deadline was never real
MEDIUM · False urgency
💳 Free trial with hidden catch
Requires a card "just in case" then charges you when the trial ends
MEDIUM · Hidden cost
📈 Price increase warning
"Price goes up tomorrow!" pressure used to rush your decision
MEDIUM · False urgency
⭐ Suspicious reviews
Review patterns or wording that suggest fake or incentivised feedback
AI · Claude analysis
Real world examples

Dark patterns
are everywhere.

E-commerce checkout
"☑ Yes! Sign me up for exclusive partner offers via email and SMS."
Pre-ticked marketing consent
Subscription cancellation
"No thanks, I don't want to save money."
Confirm-shaming
Hotel booking site
"🔥 Only 1 room left! 14 people are viewing this right now."
Fake scarcity + social pressure
Cookie consent banner
Large green "Accept All". "Manage preferences" buried in 8pt gray text.
Cookie banner obstruction
Flash sale page
"⏰ Offer ends in 00:47:12" — resets to 00:59:00 on refresh.
Fake countdown timer
Newsletter signup
"Uncheck this box if you do not wish to not receive our weekly digest."
Double negative consent
AI analysis

Powered by
Claude AI.

DOM rules catch the obvious patterns instantly. For subtle manipulation — roach motel cancellations, misleading pricing, guilt-trip copy — the extension sends page context to Claude for deeper analysis.

🔑
Your own API key
Your Anthropic key is stored locally in Chrome. No data ever touches our servers.
⚡
Works without AI too
All DOM detections run locally with zero latency. AI analysis is an optional enhancement layer.
🟢
Health indicator
Simple mode shows an instant verdict — clean, a few things to watch, or manipulative — so you know at a glance.
💬
Plain-English labels
"Guilt-trip button" not "confirm-shaming". Every finding uses plain language plus a "What to do" tip.
🙈
Dismiss what you don't care about
Ignore a finding on just this site, or never show it again — you stay in control of what gets flagged.
🎛️
Pattern preferences
Choose which of the 11 pattern types to detect. Toggle each one on or off in plain English from the options page.