.svg)
Disclosure isn't the only trust factor. Here's why AI trying too hard to sound convincingly human can undermine trust even when properly disclosed.
TL;DR
Disclosure gets treated as the finish line for AI trust: tell people they're talking to AI, and the trust question is settled. It's a necessary step, but it's not the whole picture. There's a second, quieter issue that shows up even after disclosure happens correctly: an AI that's working overtime to sound convincingly, performatively human can undermine trust in a different way entirely.
Not knowing you're talking to AI is one kind of trust failure. Knowing you're talking to AI, and having it behave like it's trying hard to seem like it isn't, is a different kind. The first is about honesty. The second is about whether the interaction feels genuine or feels like a performance aimed at you specifically. A properly disclosed AI can still fail the second test, and that failure is worth understanding on its own terms rather than assuming disclosure alone covers it.
Trust in AI-generated content isn't starting from a neutral place. A Quinnipiac poll found that only 21% of Americans trust AI-generated information most or all of the time, while 76% trust it hardly ever or only some of the time. That's a genuinely skeptical starting point, and it means any signal that reads as an attempt to influence or charm someone artificially lands on ground that's already primed for suspicion, not neutral curiosity.
There's a meaningful difference between an AI that communicates clearly and naturally, and one that performs humanness as a strategy. Inserting first-person emotional language that doesn't track with the actual conversation, faux personality quirks meant to seem endearing, or a level of casual warmth that feels disproportionate to a routine transaction, these tend to read as effortful rather than genuine, precisely because the AI has already been disclosed as AI. The performance draws attention to itself instead of disappearing into a smooth interaction.
The goal isn't to sound as human as possible, it's to communicate clearly and helpfully without drawing attention to the fact that it's trying to seem like something it's already been honest about not being.
An AI that communicates plainly, answers accurately, and doesn't oversell its own personality tends to earn more trust than one straining for warmth it hasn't organically built. Clarity and competence read as trustworthy on their own. Performed humanness, layered on top of an already-disclosed AI, tends to read as a tactic, and against a genuinely skeptical baseline, tactics are exactly what people are primed to notice.
Should a business AI try to sound as human as possible?Not necessarily. Sounding clear, helpful, and natural matters more than performing humanness. An AI that tries too hard to seem convincingly human, especially after it's already disclosed itself as AI, risks feeling like a tactic rather than a genuine interaction.
How skeptical are people of AI-generated content in general?Quite skeptical. According to a Quinnipiac poll, only 21% of Americans trust AI-generated information most or all of the time, while 76% trust it hardly ever or only some of the time. Any interaction that feels like it's trying to manipulate that skepticism tends to backfire.
What's the difference between an AI sounding natural and an AI sounding artificially human?Natural communication is clear and appropriately toned for the situation. Artificially performed humanness includes things like unearned emotional language, forced personality quirks, or warmth disproportionate to a routine interaction, details that draw attention to themselves rather than disappearing into a smooth conversation.
Does this mean AI should sound robotic instead?No, that's not the alternative being suggested. The goal is clear, competent, appropriately warm communication, not a swing to the opposite extreme. The issue is specifically with performance that oversells personality, not with an AI simply being pleasant and easy to talk to.