We instinctively distrust ads because someone wants something from us, but what's the equivalent gut check for AI-written text?

This explores how readers can apply the automatic skepticism we give to ads and other interested speech to AI-generated text, when no cultural habit for doing so exists yet. The corpus suggests we can't borrow an existing discount. Advertising comes with a built-in posture (someone wants something from me), but AI text arrived too recently and changes too fast for a posture like that to settle, so it spreads without one How do we learn to read AI-generated text critically?. What the collection does offer is a set of parts for building the discount on purpose.

Start with why the default is no discount at all. Text acts on readers as a social force whatever produced it, so we run the same interpretive machinery on AI prose as on human prose Does AI text affect readers the same way human text does?. We also add work of our own. AI output carries the surface markers of speech without the event of someone actually speaking, and readers fill in the missing speaker themselves, creating an exchange that exists only on the human side Does AI generate genuine utterances or just text patterns?. Oral cultures had a knowledge style that was performative, additive and situational, but a person in the room always anchored it. AI reproduces the style without the anchor Does AI-generated content mirror oral culture's knowledge patterns?. So the first discount is to notice your own contribution: the feeling that someone is talking to you is something you supplied.

Second, discount in a known direction. A study of 2,939 writers and 11,091 readers found that AI assistance pushed all 29 tested dimensions of writer persona the same way, toward more extreme, confident, high-quality, agreeable and privileged Does AI writing assistance change how readers perceive the writer?. Writers edited AI paragraphs only 23% of the time, and the edits kept 96% similarity to the original, so that tilt reaches readers nearly intact Do writers actually edit AI-generated text before publishing?. That gives a practical rule: mark down text that is polished, sure of itself and agreeable. People in every language follow confident-sounding AI answers even when they're wrong Do users worldwide trust confident AI outputs even when wrong?. Confidence is the signal our instincts track, and the one least worth trusting.

Third, learn the tells that aren't about style, since style is easy to edit away. Discourse-level choices like character agency and chronological structure separated AI fiction from human fiction with 93.2% accuracy, and those choices survive surface polishing because fixing them takes a rewrite Can AI stories be detected without analyzing writing style?. Human posts make an implicit bid for your attention. AI posts inherit visibility without making that bid, which readers feel as aloofness Does AI writing lack the internal appeal to attention that humans use?. Artificial text also loses four properties of natural writing: two-way dialogue, continuity of context, an embodied author, and political situatedness. AI hotel reviews are caught over 80% of the time because they claim personal experience that never happened Does AI-generated text lose core properties of human writing?. The useful question isn't 'does this sound like a robot?' but 'is there a stake or a body behind this?'

The limit is that individual vigilance doesn't scale. AI can generate knowledge faster than people can judge it, and the tools we'd use to check are increasingly AI-made too Can AI generate knowledge faster than humans can evaluate it?. Ad skepticism works because we can name an interested party, and the event-residue idea suggests AI text has no speaker to name in that way. So the discount probably has to become a shared cultural habit, the way ad skepticism did, and not just a personal skill.