--- title: Why AI Disclosure Makes Your Work Seem Less Authentic slug: technostress-ai-disclosure-authenticity-builder date: 2026-06-11 excerpt: A large 2026 study found people rate the same writing lower once they learn AI helped make it, and the reason is a drop in perceived authenticity, not quality. Here is what that says about the part of a finished piece a working person most wants to call their own. featured_image: "https://bbtxujdxvidaghmhxkqs.supabase.co/storage/v1/object/public/generated-images/blog-1781168767368-technostress-ai-disclosure-authenticity-builder.webp" featured_image_alt: A single printed page resting on a wooden desk beside a closed laptop in soft morning light, with one sentence on the page lightly highlighted by hand. canonical_url: https://cerevisor.com/blog/technostress-ai-disclosure-authenticity-builder updated_at: 2026-06-11T09:06:08.386159+00:00 --- # Why AI Disclosure Makes Your Work Seem Less Authentic TLDR A large 2026 study found people rate the exact same writing lower the moment they learn AI helped make it. The drop is about perceived authenticity, not quality, and saying the tool was only an assistant does not undo it. The honest catch: in other settings, disclosure does not dent trust at all, so this is a tendency worth noticing, not a rule about anyone's worth. A developer I know shipped a clean piece of internal documentation last week, the kind that takes an afternoon and saves everyone a month. Then she paused over one line in the message: whether to mention that the tool had helped write it. She added the line, deleted it, added it again. That small hesitation has a surprising amount of research behind it now. --- ## What 27,000 readers did with the same writing This spring, three researchers, Manav Raj, Justin Berg, and Rob Seamans, published the largest test of this I have seen, in a long-running psychology journal. Across sixteen experiments, they showed people short pieces of [writing](/blog/technostress-ai-rival-or-partner-writers-builder-research) and asked them to rate the work. The texts were identical. The only thing that changed was a label saying a person wrote it, an AI wrote it, or a person used AI to help. The work with an AI label got marked down, again and again. The part that stays with me is what did not rescue it. When people were told someone used the tool only as an assistant, they still judged the piece about as harshly as if the machine had written the whole thing alone. The quality never moved, because the words never moved. What moved was perceived [authenticity](/blog/ai-authenticity-labor-sound-like-you), the sense that the writing came from a real person. That is the same instinct sitting underneath the hidden work of making AI output sound like you, and underneath the quiet question of whether a finished piece still carries a sense of authorship. > "To explore these questions, the scientists carried out sixteen separate experiments involving a total of 27,491 participants." PsyPost, reporting on a study in the Journal of Experimental Psychology: General, April 2026 Key Insight The work did not get worse. The label changed, and with it the sense that a real person stood behind the words. --- ## Where the AI disclosure penalty stops Two honest limits. First, this measured how other people judge writing, not how the writers felt on the inside, so reading it as a verdict on a person’s worth is my extension, not the study’s finding. Second, the penalty is not a law. A separate set of experiments on persuasive messages found that disclosing AI help did not reliably reduce [trust](/blog/ai-adoption-trust-not-training) at all, and people who used AI built [trust](/blog/ai-resistance-trust-not-training) just as well, only faster. Even there, the AI-assisted messages scored a little lower on authenticity. So the felt-authenticity dip is real and repeatable. What it does to the outcome depends on the room. > The quality never changed. What dropped was the sense that a person was in there. --- ## What to notice about your own voice today The reaction those thousands of readers had to a label is a quieter version of the one a working person feels from the inside. When a tool returns something clean and fine, there is often a small, almost unnoticed pull, a wish that some of the work be unmistakably ours. That pull is not a problem to fix. It is information. The same way leaning hard on a tool can let reliance erode self-belief, the authenticity question is worth sitting with rather than arguing away. Notice the half-second of hesitation before you mention the tool. Notice which parts of a piece you most want to be able to call your own. That noticing is most of the practice, and it is figured out one piece of work at a time. #### Sources - [The Artificial Intelligence Disclosure Penalty: Humans Persistently Devalue AI-Generated Creative Writing](https://psycnet.apa.org/fulltext/2027-12675-001.html) - Journal of Experimental Psychology: General, 2026-04-05 - [People consistently devalue creative writing generated by artificial intelligence](https://www.psypost.org/people-consistently-devalue-creative-writing-generated-by-artificial-intelligence/) - PsyPost, 2026-04-05 - [Writing with AI boosts trust-building efficiency](https://pmc.ncbi.nlm.nih.gov/articles/PMC12765382/) - iScience, 2025-11-20