UX Dark Patterns Audit

Dark Patterns Audit Skill

How this skill behaves (read first)

This is an evaluative skill — it reviews existing work, auto-runs, and is built to be called as a validation step by generative skills, especially ones touching conversion, pricing, sign-up, subscription, or cancellation (pricing, login-signup, checkout-payment, onboarding, offboarding). It's the systematic version of the ethics guardrail those skills carry inline.

It owns intent to deceive/manipulate. Deliberate deferrals so it doesn't duplicate:

• Honest usability problems (a confusing-but-not-deliberate layout, unclear label) → the heuristics audit. A UX mistake is not a dark pattern.
• Visual clutter / overdesign → the aesthetics audit.
• Accessibility → the accessibility audit.

The line is intent: this audit asks whether the design exploits the gap between what users think they're doing and what the product actually does.

Scope discipline. When invoked directly (the user named this audit), review only this concern — don't pull in sibling audits. It runs alongside other lenses only when an orchestrator (ux-design-review) or a generative skill calls it under docs/orchestration-policy.md, where it sits in Tier A — always-relevant, auto-runs. Explicit scope always wins.

What this skill changes vs. default behavior

By default, manipulation only gets flagged when it's blatant, with no framework, no severity, and a tendency to either miss subtle patterns (preselection, forced continuity, manufactured FOMO, bad defaults) or over-accuse honest mistakes. This skill forces:

1. A systematic sweep of the named pattern catalog — not just the obvious offenders.
2. Intent classification — Dark vs. Grey vs. honest Mistake — so findings are fair, not alarmist.
3. A concrete, honest-design fix for each.
4. Decision tests (the Newspaper Test, the intent-gap question) that surface the subtle cases.

The core lens: intent and the dark / grey / mistake spectrum

A dark pattern is a UI deliberately designed to lead users into unintended actions. The keyword is deliberately.

• Dark — exploits the gap between user intent and product behavior for the company's gain (hidden renewal after "Buy Now"). Flag and fix.
• Grey — not hidden, often industry-standard, but the default serves the business over the user (a pre-checked "email me offers" box). Flag as caution; nudge toward the user.
• Mistake — accidental confusion with no manipulative intent. Not a dark pattern → route to the heuristics audit.

Every finding should say which of the three it is. Driver to name: most dark patterns come from conversion/A/B-test optimization that wins a short-term metric and loses long-term trust.

The pattern catalog — what to check and what each looks like

Sweep all categories. For each finding: pattern name, classification (Dark/Grey), severity, fix.

Sneaking (hiding or smuggling information)

Flag when: costs/fees/taxes appear only at checkout (hidden costs); items added to the cart the user didn't choose (sneak into basket); a subscription auto-renews with the renewal buried in fine print (forced continuity); the user gets a different outcome than the one presented (bait and switch).

• ❌ "$9 flight" that becomes $40 after mandatory fees at the last step.
• ✅ Total cost, fees, and renewal terms shown up front, before commitment.

Obstruction (making the un-profitable choice hard)

Flag when: cancellation/downgrade is far harder than sign-up (roach motel); the design blocks price comparison (no per-unit price, deliberately unclear tiers).

• ❌ Subscribe in one click; cancel only by calling support during business hours.
• ✅ Cancelling is as easy as subscribing.

Misdirection & visual/wording tricks

Flag when: flashy elements distract from what's actually happening (misdirection); checkboxes/copy use confusing or double-negative wording (trick questions); ads masquerade as content or controls (disguised ads); options that benefit the company are preselected; bad defaults (most expensive plan or all notifications pre-on).

• ❌ "Uncheck if you do not want to not receive emails."
• ✅ Plain, single-meaning labels; opt-ins unchecked by default.

Nagging

Flag when: the user is repeatedly re-prompted to do something they already declined.

• ❌ A newsletter pop-up that returns every visit after being dismissed.
• ✅ Respect a decline; ask again only with good reason, much later.

Forced action & privacy

Flag when: access/benefit is gated behind unrelated permissions or contact-list access used to message others (friend spam); data shared for one purpose is reused outside that context (privacy zuckering).

• ❌ "Allow contacts access to continue" → then messages all of them on the user's behalf.
• ✅ Request only the data the feature needs, for the purpose stated.

Emotional manipulation

Flag when: opt-outs shame the user (confirmshaming); urgency/scarcity is fake or exaggerated (FOMO, fake countdowns, "only 2 left" on unlimited digital goods); fear is used to push action (scaremongering).

• ❌ "No thanks, I don't care about saving money."
• ✅ Neutral dismissals: "No thanks" / "Maybe later"; real scarcity only.

Attention exploitation (well-being)

Flag when: the design manufactures compulsive use rather than serving a goal — infinite scroll with no stopping point, autoplay that removes the choice to continue, variable-reward mechanics, streak anxiety, manufactured FOMO (read receipts), notification overload.

• ❌ Onboarding that pushes "enable all notifications" to maximize re-engagement.
• ✅ Mindful defaults; notifications proposed by value, not volume; natural stopping points.

Microcopy-level

Flag when: copy triggers emotion ("sadly", "unfortunately" on cancel), guilt-trips, or uses confusing double negatives / vague "OK/Cancel" on consequential choices.

• ✅ Plain, honest, neutral wording on every consequential action.

Audit tests (use these to catch the subtle cases)

• The intent-gap question: does the design exploit the gap between what the user thinks they're doing and what the product records? If yes → Dark.
• The Newspaper / Front-Page Test: if a journalist wrote a headline about this feature tomorrow, would the team be comfortable with it?
• First- and second-order consequences: the immediate win (more signups, longer sessions) vs. what it causes next (regret, churn, distrust). Dark patterns usually look fine at first order and harmful at second.
• The default check: who benefits from each default — the user, or the business? Defaults that only serve the business are at least Grey.

Severity & classification

Classify honestly — labeling an honest mistake "Dark" discredits the audit. When intent is genuinely unclear, say so and apply the intent-gap question.

Audit output format

## Dark Patterns Audit — [Screen / Flow]
Context: [platform · what the design is optimizing for, if known]

### Critical (Dark)
- [Pattern name] — what's manipulative → honest-design fix.
### Major (Manipulative)
- …
### Grey (Caution)
- …
### Clean ✓
- [practices that respectfully serve the user]

**Trust verdict:** [trustworthy / some concerns / manipulative] + the top fixes.

End with the trust verdict and prioritized fixes — that's what a calling generator (or a team) acts on.

How to run a good audit

1. Lead with intent. Classify each finding Dark / Grey / Mistake; route honest mistakes to the heuristics audit instead of flagging them here.
2. Sweep the whole catalog, including the subtle/attention patterns — those are what default reviews miss.
3. Apply the tests (intent-gap, Newspaper, second-order) to anything ambiguous.
4. Fix toward honesty, not just compliance — the constructive alternative, not "remove it."
5. Don't over-accuse. Fairness is what makes the audit credible; not every aggressive growth tactic is a dark pattern.

Common do/don't patterns

Source lessons (Uxcel)

• 14 Design Dark Patterns You'll Want to Avoid
• Deceptive Patterns
• Ethics in Decision Making
• Attention & Well-being
• Writing Action-Based Messages