Things that make us go hmmm…
Too good to be human?
Nu U Staff | April 2026
The infosphere is starting to feel a little like the “real or cake?” game: It’s harder to tell what’s human and what’s AI. It used to be that "AI-powered" was easy to spot: images with wonky spelling, humanoid figures with too many fingers, tell-tale phrases (“delve into”), rote bot responses ("sorry you are having trouble"). But if a piece read well, a voice sounded human, or a social media account engaged like a person, more often than not, you could safely bet there was a genuine somebody behind it. Lately, signs suggest that even those most tuned in to AI are starting to second-guess their instincts, with good reason.
Reddit’s latest anti-bot move was one telling signal. The platform announced that it is introducing a clear App label for approved automated accounts, while some accounts showing automated or “fishy” behavior may be asked to verify that there is a human behind them. Reddit’s CEO wrote that “If something suggests an account isn’t human … we may ask it to confirm there’s a person behind it.” In other words, Reddit is calling the bluff on the old “if it looks like a duck and quacks like a duck” assumption.
To be fair, asking for proof of humanity on digital systems is not new. Well before the generative AI surge, users were already being asked to prove they were human by clicking “I’m not a robot” on apps and websites to filter out spam, bots, scraping, and other suspicious traffic. But when a platform as human-driven and anonymity-forward as Reddit recognizes that AI bots are now so good at mimicking human engagement habits and conversational patterns that it needs to take action to sustain true human-to-human interaction (and protect what makes it distinct), that’s a hmmm… all by itself.
But this is more than a bot-improvement story. It’s a human one, too. As AI tools and technologies have improved, people (even experts) are also getting worse at spotting what is synthetic. New research from UNSW found that people are overconfident in their ability to distinguish AI-generated faces from real ones and that many of us—even among those with exceptional face-recognition skills—can be fooled by today’s improved AI. (If you doubt it, try UNSW’s short self-test yourself.) Research in Communications of the ACM points in a similar direction, suggesting that human detection of AI-generated content can be weak enough to feel close to chance in some settings and that over time, as AI gets better, humans are likely to get worse at discerning the subtle differences. The rule of thumb now? Don’t look for the mistakes or obvious glitches; look for what’s “too right” or “too good to be true.” AI has improved to the point that imperfection is starting to look human again and we’ve come almost full circle: the typo that once hurt credibility can now signal proof of authenticity.
In the human + AI-powered world we live and work in, there’s more complexity and nuance added daily. So much content now sits somewhere in between: drafted by AI, edited by a person; designed by AI, approved by a person; written by a person, expanded by AI, fact-checked by a team of humans and machines. In the near term, there may be more pressure for people proof (more labels, more verification, more demands) to show there’s a human in the loop and that the work is not purely machine-generated. But the more mainstream AI use becomes, the harder it may be to treat every assist as suspicious or every disclosure as meaningful. Major platforms are already experimenting with that middle ground through policies that require disclosure of realistic altered or synthetic content rather than treating every AI touchpoint the same. We could even get to a point where the distinction matters less—where human and AI teammates are both accepted parts of the workflow, and we stop asking “real or cake?” questions and focus more on the value of the outputs (is the work accurate, reliable, useful?).
If you’re leading people, teams, or learning, a few gut-check questions are worth sitting with now: Do we have policies in place to support the human + AI balance we want on our teams? What guidelines and guardrails clarify where AI should and should not be used and how humans stay in the loop? What learning experiences are we creating to build judgment and foresight? What kinds of AI use need disclosure in our context and which do not? And as AI evolves, what do we want credibility to rest on—proof of humanity, proof of process, or proof of value? That’s the “hmmm…” we’re watching.