Direct answer
A dark pattern is a UI or UX design choice built to manipulate or deceive users into doing something they would not otherwise choose. My default recommendation on a genuine dark pattern is no, with a specific alternative that captures some of the lift honestly, because the trust cost tends to outlast the conversion win.
Who I consult
Design or UX, for a read on whether this matches known deceptive patterns rather than aggressive-but-honest persuasion. Legal, since dark patterns increasingly carry direct regulatory exposure in several jurisdictions. Customer support or research, for existing complaint signal on adjacent flows. My own leadership, since a staff-level call here shapes the org's risk posture, not just this one feature.
Short- and long-term signals
Short-term: the conversion lift itself, and how confident and reproducible it actually is.
Long-term: support-ticket and complaint volume, chargeback rate (customers disputing the charge directly with their card issuer, which is more damaging than a normal refund) if it applies, churn in the exposed cohort over the following months rather than just the immediate conversion event, and regulatory or press exposure risk.
What I would recommend
Reject the deceptive version, and counter-propose the honest version of whatever real need it is serving. Growth teams usually have a legitimate insight, friction, unclear value, hesitation, even when the proposed solution is manipulative; find the honest lever that addresses the same insight.
Worked example
Growth proposes making the decline option on a subscription upsell visually tiny and gray, styled to look disabled, next to a bright accept button, a visual-interference pattern, also called false hierarchy, in which the styling rather than the wording does the deceiving, so users misread the tiny option as unavailable rather than simply less prominent. That is a clear dark pattern, deceptive by design rather than merely persuasive, so I would recommend against it. Name the pattern precisely when you argue the case, because confirm-shaming is a different one: there the decline option is instead worded to make the user feel guilty, for example "No thanks, I do not care about saving money," so the fix is copy, whereas the fix here is contrast and hit area. Both are deceptive, and calling one by the other's name is the fastest way for design or legal to conclude you have not actually looked at the screen. Counter-proposal: keep both options equally legible, but improve the accept option's actual value communication, clearer benefit copy, a genuinely true limited-time framing if one exists. Expect a smaller lift than the manipulative version would produce, but without the complaint and chargeback costs similar patterns generate elsewhere, and without the longer-term regulatory exposure.
Trade-offs and pitfalls
Banning anything that increases conversion through a default or a nudge is a pitfall, since a well-designed, honest default is legitimate and different from a dark pattern; the test is deception, not persuasion. Accepting a "small test, we will pull it if it backfires" framing for something genuinely deceptive is the other pitfall, since running it at all is the actual ethical call, not just the decision to keep it afterward.
