AI0 views

One in Three US Users Insults AI Chatbots—And It Usually Backfires

A new study analyzing over 8 million messages sent to ChatGPT, Gemini, and Claude between September 2025 and September 2026 reveals that 33% of American users have insulted or cursed at AI chatbots at least once. Men were roughly one-third more likely to use offensive language than women, while people aged 35 to 44 showed the highest rate of hostile exchanges. Interestingly, younger users aged 18 to 24 were the least likely to resort to insults.

The research points to three primary frustrations driving these outbursts: ignored instructions (19%), incorrect responses (18%), and misunderstood requests (15%). However, insults appear to be counterproductive. When users sent hostile messages, the AI resolved their issue in only 24% of cases—compared to 35% success when no insult was involved. The data suggests that maintaining a respectful tone with chatbots may actually improve outcomes.