Opinion. Thesis: Crypto scam warnings should be judged by whether they change a decision, not whether people like, share or click them. New Canadian evidence suggests the best warning may be the one that adds enough friction to interrupt a bad choice—even when audiences say they prefer something smoother.

On November 17, the Canadian Securities Administrators published the results of a randomized online trial involving 2,500 Canadian adults. Participants saw a simulated social-media feed containing legitimate and fraudulent crypto investment opportunities, plus one of four anti-scam messages or no message. They then allocated a hypothetical $10,000 across the opportunities.

The result is uncomfortable for anyone optimizing fraud prevention like a marketing campaign. An inoculation video that demonstrated scam tactics reduced allocations to fraudulent opportunities by 11% relative to the control. An interactive “spot the scam” quiz reduced them by 8%. A reflection prompt and a message encouraging people to apply their usual double-checking habits did not produce statistically significant reductions.

Yet the effective interventions were the least liked and least likely to be shared. Only 29% of participants rated the video positively, while 77% rated the ineffective double-checking message positively. The study’s lesson is not that every disliked warning works. It is that sentiment, reach and engagement are weak substitutes for the behaviour a safety system is supposed to change.

Positive friction is a product feature

The CSA-commissioned research report describes the likely mechanism as “positive friction.” The video and quiz disrupted passive scrolling and required closer attention. They did not materially improve general knowledge of warning signs across the groups. Instead, they changed how people engaged with the next decision.

That distinction matters for exchanges, wallets, banks and social platforms. Most warning systems still assume the user’s problem is missing information. They place a paragraph beside a button, require a checkbox, or publish a checklist somewhere else. A scammer is designing the opposite experience: urgency, emotional momentum, apparent social proof and a straight path to payment.

My view is that fraud controls should deliberately interrupt that path at the moment authority or money is about to move. The warning should require a small act of verification: compare two claims, identify the source of an endorsement, confirm the recipient through an independent channel, or check whether a firm appears on the CSA’s list of authorized crypto platforms. A generic “be careful” banner can be acknowledged without being processed.

This is related to the principle in Wallets Should Sign Outcomes, Not Just Instructions. A signature proves authorization; it does not prove that a user understood the likely result. Warnings cannot guarantee understanding either, but a product can require evidence that the user engaged with the risk before a first transfer, a new recipient, an unusually large amount or an irreversible approval.

Measure the protected outcome

Safety teams need metrics that match their purpose. Impressions reveal distribution. Click-through rates reveal attention. Positive ratings reveal comfort. None establishes that fewer people sent funds to a fraudster.

A better evaluation stack would track progressively stronger evidence:

  1. Comprehension: can the user identify the specific risk in the offer?
  2. Verification: did the user consult an independent registry, contact or source?
  3. Decision quality: did the intervention reduce transfers to confirmed fraudulent destinations without blocking legitimate activity?
  4. Durability: does the effect persist days or weeks later, and after repeated exposure?
  5. Recovery: when prevention fails, can the provider preserve evidence, freeze what remains and route the victim to qualified help quickly?

The fourth and fifth measures matter because the Canadian trial stopped at an immediate simulated allocation. It was not conducted on a real social platform, used hypothetical rather than personal funds, required participants to allocate the full amount, and tested a single exposure. The researchers explicitly say they cannot determine whether the effects persist or grow over time. The November 17 publication date is also not a fieldwork date; the report does not identify when participants completed the trial.

Do not turn every warning into a roadblock

Friction has costs. Repeated prompts can become ritual clicks. Poorly targeted delays can deny timely access to legitimate users. A platform that interrupts everything teaches people that nothing is important. Controls should therefore be risk-based, measurable and reversible.

Start with moments that combine novelty, urgency and irreversibility: a first withdrawal to an external wallet, a newly added payee, a request initiated from a social-media message, a remote-access session, or a transfer that departs sharply from the account’s history. Use a brief interactive check, explain why it appeared, provide an exit, and record whether the intervention changed the decision. Publish both positive and null results so every institution does not repeat the same failed experiment.

This is an inference from the CSA study, not a result the study itself proved. The trial tested communications in a simulated feed, not exchange withdrawal flows or wallet interfaces. Applying the finding to transaction products requires separate real-world testing with privacy, accessibility, false-positive and consumer-protection review.

What would change my view

I would change this assessment if large, independent field trials using real decisions showed that interactive friction merely shifted losses to another channel, caused substantial harm to legitimate users, or lost its effect after repeated exposure. Evidence that a low-friction message produced equal or better durable outcomes would also favour the simpler design.

Until then, providers should stop treating the warning people like most as the warning that protects them best. In crypto fraud prevention, the useful question is not “Did this message perform?” It is “Did this person make a better decision?”

Disclosure: Jason Ansell is a co-founder of Vector Smart Chain and the author of The Dark Side of Web3, a book about recognizing crypto scams. This article does not assess VSC, any token or any investment opportunity, and it is not investment advice.

Sources