Say It's AI: What the Research Actually Says About Disclosing AI-Written Messages
Thirteen experiments say telling people a message was written with AI makes them trust you less. The same research says being found out costs far more, that suspicion now falls on everyone, and that the EU, Utah and California no longer leave you the choice. Here is how to disclose without losing the room.
The disclosure penalty is real
The largest body of evidence is a 2025 paper in Organizational Behavior and Human Decision Processes by Oliver Schilke and Martin Reimann: thirteen pre-registered experiments and a meta-analysis across 4,093 observations, with students, managers and investors as the judges. The finding is blunt: "actors who disclose t…
Being found out is worse
If the research stopped there, the lesson would be to keep quiet. It does not stop there.
Where it matters least, and most
Context changes the size of the penalty, and the pattern is useful.
What the law now says
Preference has become obligation in several places this year.
How to disclose without losing the room
The evidence does not say "disclose and suffer". It says the penalty depends on how, where and when. What it supports:
Why the message should carry it
That last point is why Kanvo treats disclosure as a property of the message rather than a line in a policy. When an agent sends on someone's behalf, Kanvo records at that moment who authorised it, what it was allowed to say, and whether it told the recipient it was AI. The notice travels with the message, the record i…
Read next
Workslop and the one-ask messageWhat to hand to the agent, and what to keepEvery AI-sent message into the EU now needs a disclosureAgent mandates and the EU AI Act
Kanvo Team
Writes here about what Kanvo is building and why — the send path, the evidence it leaves, and the decisions behind both.
Related Articles
AI made polished text free, so the cost of a message moved to the person reading it. 40% of desk workers received "workslop" last month and spent nearly two hours on each piece. The message that still works in 2026 carries one ask, in the first line, with the context a human or their assistant needs.
Workslop and the One-Ask Message: How to Write for a Reader Who Has an AI Too
AI made polished text free, so the cost of a message moved to the person reading it. 40% of desk workers received "workslop" last month and spent nearly two hours on each piece. The message that still works in 2026 carries one ask, in the first line, with the context a human or their assistant needs.
What to Hand to the Agent, and What to Keep: The Messages an AI Should Never Write for You
We build an agent that reads your inbox and drafts your replies. Here is the research on the messages it should not touch: apologies, bad news, first contact, anything where the effort is the point. And the caveat on summaries that even careful people miss.
Enjoyed this article?
Subscribe to our newsletter for more insights and updates.