Most people do not realize when a personal message they receive was written by AI, study finds

2026-04-20 · Source: Artificial intelligence (AI) – The Conversation · Field: Science & Research — Social Sciences & Behavioral Studies, Artificial Intelligence & Machine Learning · Depth: Fundamental Awareness, quick

Summary

Two new experiments involving over 1,300 U.S. participants, aged 18 to 84, reveal that most people do not consider a personal message to be AI-generated, even if they use AI themselves. Researchers showed participants AI-generated messages, such as an apology email, under various disclosure conditions. A significant "AI disclosure penalty" was observed: when participants knew a message was AI-generated, they rated the sender negatively ("lazy," "insincere"). However, when authorship was undisclosed, impressions were as positive as when messages were believed to be human-written. Surprisingly, participants' own frequent use of generative AI had little impact on their default skepticism, with heavy and light AI users alike assuming human authorship when not informed otherwise. This lack of skepticism creates a moral dilemma where undisclosed AI users face no detection risk, while those who disclose suffer reputational damage.

Key takeaway

For AI Product Managers developing communication tools, recognize that users face a dilemma: undisclosed AI use carries no social penalty, but disclosure leads to negative perceptions. Your product design should consider how to mitigate the "AI disclosure penalty" or facilitate communication methods that are less susceptible to AI detection, such as voice or in-person interactions, to preserve authenticity and trust in personal exchanges.

Key insights

People generally lack skepticism about AI-generated personal messages, leading to an "AI disclosure penalty."

Principles

• Undisclosed AI use avoids negative social judgment.
• Disclosure of AI authorship incurs reputational costs.

Method

The study recruited 1,300+ U.S. participants, showing them AI-generated messages under four disclosure conditions to assess social judgments.

In practice

• Consider in-person communication for heartfelt messages.
• Be aware of the "AI disclosure penalty" in professional contexts.

Topics

• AI-Generated Text
• Social Judgment
• AI Disclosure Penalty
• Generative AI Use
• Skepticism

Best for: AI Product Manager, Product Manager, AI Scientist, AI Ethicist, Research Scientist

Related on AIssential

• If using ChatGPT is cheating, what about ghostwriting? The old debate behind a new panic — Societal discomfort with ghostwriting and AI-generated text stems from a desire for authentic authorship.
• Is your AI chatbot manipulating you? Subtly reshaping your opinions? — Generative AI's hyper-personalized persuasion poses significant, unmonitored risks to individual agency and societal discourse.
• Things You Should Never Tell ChatGPT (Privacy & Safety Guide for Smart AI Use) — Never share sensitive personal, financial, or security information with AI chat tools.
• A uni professor admitted using AI to write an opinion piece. Here’s what it revealed about trust in the technology — Undisclosed AI use erodes public trust and undermines industry credibility.
• Correct Yourself, Keep My Trust: How Self-Correction and Social Connection Shape Credibility in Social Chatbots — Social chatbots maintain credibility and amplify correction effectiveness by self-correcting and fostering social connection.
• "That's AI Slop, You Bot!" Studying Accusations, Evidence, and Credibility in Online Discourse Towards LLM-Generated Comments — Online "AI slop" accusations function as social gatekeeping for perceived authenticity, not accurate AI detection.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial intelligence (AI) – The Conversation.