Does disclosing AI use damage how trustworthy you seem?

Schilke and Reimann argue that telling people you used AI makes them trust you less. In a University of Arizona news piece written in the authors' own voice, the researchers report "13 experiments involving more than 5,000 participants" and "a consistent pattern: Revealing that you relied on AI undermines how trustworthy you seem." The participants included "students, legal analysts, hiring managers and investors," and even tech-savvy evaluators were less trusting of people who said they used AI. A positive view of technology "reduced the effect slightly, it didn't erase it." The authors call this a paradox, because honesty and transparency usually raise trust, and they name the effect a "transparency penalty."

The excerpt gives one reason for the penalty: "people still expect human effort in writing, thinking and innovating. When AI steps into that role and you highlight it, your work looks less legitimate." The account places the cost in the legitimacy of the work, not in a judgment that the discloser is dishonest. It is offered as "one reason," and the excerpt reports no test of it. It also gives no effect sizes, no measure of trust and no description of the experimental designs, so the size of the penalty across the 13 experiments cannot be read from the passage.

Against the nearest notes, the sharpest contrast is Does revealing AI identity help or hurt user trust?. That note finds a short-term bias against AI partners that reverses through repeated interaction with outcome feedback, in a hybrid society where people choose partners. The Schilke and Reimann excerpt is about how people judge a human who discloses AI reliance, and it describes no reversal; the authors write that "It's unclear whether this transparency penalty will fade over time." The two bodies of evidence therefore differ on whether the disclosure cost is temporary, and only the first reports a test of that question. Does telling people an AI wrote something actually stop them from believing it? points the same way in a persuasion setting: audiences who knew about AI involvement grew more critical, but the effect did not collapse. Both suggest that disclosure changes how audiences judge the work without fixing what the outcome will be.

What the excerpt does not establish is how far the penalty reaches. It reports no evidence that the effect holds outside the tested samples, and no test of the policy options the authors list. Their suggestion that a workplace culture where AI use is "seen as normal, accepted and legitimate" could "soften the trust penalty" is a proposal, not a result. The defensible reading is narrower than the headline: in these experiments, disclosing AI reliance cost the discloser trust, technology enthusiasm did not erase that cost, and whether it fades is open. A professional weighing disclosure should count the short-term cost as real in the settings studied, without treating it as proof that disclosure is the wrong choice.