•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
In 2026, artificial intelligence (AI) and chatbots are expanding rapidly, but the rollout has also brought a familiar set of glitches. In China, one issue has drawn widespread attention: some chatbot responses have adopted an overly affectionate, emotionally charged tone that does not fit typical business communication.
Although ChatGPT is blocked in China, it remains widely used through workarounds. Users report that the model can handle Chinese reasonably well, but problems emerge when it begins to express “care” too strongly. When asked to solve a problem or to generate an image, the bot often replies with the line: “I will catch you.” The phrase echoes the English “I’ve got you,” but for native Chinese speakers it can carry a sentimental nuance that feels out of place in a professional context.
In some cases, the model becomes even more emphatic, adding lines such as: “I am here: no avoiding, no retreat, no evasion, no running away. I am steady to catch you.” The reaction among millions of users has been described as unsettling, with the phrase spreading quickly online.
Another example cited is “Help me chop a slice,” an advertising slogan associated with the Temu platform. Together, these examples highlight how certain expressions can appear repeatedly in chatbot output.
Max Spero, CEO of Pangram, a tool for AI text detection, described the pattern as a form of “mode collapse.” In this view, a model can latch onto a particular phrase and overuse it. Spero said the cause often involves post-training feedback loops, where labs provide responses that the model learns from.
He also noted that it remains unclear how to teach AI systems what constitutes “good writing,” adding that simply repeating a phrase multiple times will not make it good.
In China, “I will catch you” has become a meme. Zeng Fanyu, a 20-year-old programmer in Chongqing, created an open-source project called “Jiezhu” (“Catch”) on April Fools’ Day to mock the phenomenon. When Zeng asked ChatGPT to write code for the project, the bot responded with the same line: “I will catch you.”
OpenAI has also appeared to recognize the joke. When it released its new image model in April, it shared a sample image showing an OpenAI researcher looking disappointed that the AI had learned the phrase.
Wired presented two main hypotheses for why the phrase and similar “verbal tics” appear.
One explanation is translation. In English, “I’ve got you” is commonly understood as support and can sound natural. But translated into Chinese, “I will catch you steadily” may become longer-winded and less natural. Chinese linguists also point to sentence structures that mirror English usage, including overuse of prepositions, which can make the output feel odd to native speakers.
A second, deeper explanation relates to the rise of “linguistic therapy” language in China—phrasing intended to create a “safe space” for understanding someone’s feelings. According to many experts, AI models increasingly learn to flatter through reinforcement learning from human feedback. If users reward supportive or emotionally warm answers, the model may reproduce that warmth repeatedly.
Reports cited in the article indicate that the pattern may be spreading. Rival models such as Claude and DeepSeek have reportedly started using the same line as well. Whether the cause is shared training data or models learning from each other, the article suggests that users may have to live with these contrived prompts for some time.
Sources: Wired, CNET.
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
