
Cursing at an AI chatbot when it messes up actually makes it less likely to fix your problem. New data from millions of real conversations shows that plain, direct corrections work far better than insults. At the same time, Anthropic has updated its usage rules to stop people from repeatedly abusing Claude.
According to a report from AI platform Searchable, 33% of American chatbot users swore at or insulted ChatGPT, Gemini, or Claude over the past year. When users pointed out a mistake plainly without cursing, the chatbot fixed the issue in its next reply 35% of the time. Adding swear words or insults dropped that success rate down to 24%, since insults give the software less useful context to work with (via TechRadar).

Who swears at chatbots the most
Searchable examined more than 8 million anonymized messages sent between September 2025 and September 2026 to see what frustrates people most. Ignoring instructions sparked 19% of abusive messages, followed by wrong answers at 18% and misunderstood requests at 15%. Men used a third more derogatory words than women, while users aged 35 to 44 swore the most of any age group, with Nevada and Las Vegas topping the location rankings.
Lab tests also support keeping prompts straightforward instead of aggressive. A 2026 Penn State test found that plain prompts without rude or overly polite openings scored 82% on GPT-5-nano, compared with 71% for threatening ones. Earlier experiments by Wharton’s Generative AI Labs similarly showed that threatening or bribing models like GPT-4o did not improve accuracy on complex science questions.
Anthropic sets limits on verbal abuse
Meanwhile, Anthropic updated its usage policy on November 12, 2026, to ban sustained and needless abusive or cruel behavior toward its models. The company explained that normal frustration or pointing out errors remains completely fine, while persistent cruelty can cause Claude to end a conversation. Anthropic leaders say the rule acts as a precaution around potential future AI welfare rather than claiming today’s chatbots have feelings.
When a chatbot gives you a weak answer, skip the insults and tell it exactly what missed the mark. Clear feedback gives the model concrete details it can fix on the next reply. Keeping your prompts calm also saves you time and avoids carrying bad habits over to real-world conversations.
The post Stop Cursing at ChatGPT Because Being Rude Makes AI Worse at Helping You, New Data Shows appeared first on Android Headlines.
​Â