Hank Green says his ChatGPT habit got 'not healthy.' Now other science communicators are drawing their own lines.

Hank Green told his YouTube audience that the line in question was his. The four words were "I appreciate the pushback." A lot of viewers didn't believe him. By the end of last week, the SciShow and Crash Course host had walked that back in public, admitting he had leaned on ChatGPT for research on the script and acknowledging, on Reddit, that the habit had become "not healthy."

The AI did not write any of Green's sentences. He used it to generate research notes, and the resulting script ended up with what he later called "an AI feel," a compression tic, a smoothing of phrasing that survived into the line he read on camera. The line that got him caught was the line he actually wrote, and the assist that produced the catch happened upstream, in the part of the work the audience never sees.

Green's apology on X landed earlier this month after a wave of comments accusing him of using a synthetic voice. In a longer post on Reddit, he said he had been "moving so fast that my own process isn't actually clear to me," and he committed to making videos at a slower pace. His production company, Complexly, has published a short response acknowledging the episode. Coverage in TechCrunch on August 1 and Fortune on August 4 framed it as a notable moment in the creator-and-AI conversation.

Science communicators interviewed by Scientific American drew different lines. Kanta Dihal, an associate professor of science communication at Imperial College London, argues AI can help communicators keep up with the volume of new papers, "because, at a human pace, it's almost impossible to keep up with science." Raven Baxter, a molecular biologist who presents as Raven the Science Maven, uses AI for coding and for parsing her audience's questions, but she still goes to other scientists, not a chatbot, when she needs to understand a paper's actual claim. The split is not between AI users and AI refuseniks. It is between AI as a reading aid and AI as a thinking aid.

That split is also the falsifier. The "AI feel" fans caught in Green's video could be the compression tic of any fast explainer working at speed. Green was producing against a deadline, and the smoothing of phrasing he later identified in his own script could simply be what tight deadlines do to science writing. Generalizing from a uniquely pressured Green process to a craft rule about AI-assisted research is the move a careful reader should resist.

But the timing makes the question unavoidable. ChatGPT and its peers have moved from a writer's tool into a researcher's tool. The risk Scientific American identifies is that the assist can shape a communicator's understanding before the explaining even begins. Green's case is a public surface for a quieter change that has been happening to the field for at least a year.

Green has said he will slow down. The next few videos will be the test, not of whether he can stay off ChatGPT, but of whether the AI feel he noticed in his own script was really the chatbot, or just the speed. Either answer is useful, and only one of them lets the rest of the field draw a line that actually holds.