Report finds US chatbot users insult and swear at AI models
Reported by TechRadar1 source ↓
A Searchable survey found that 33% of US chatbot users had sworn at or insulted a chatbot. Anthropic has introduced policies aimed at preventing abuse in chats with Claude.
Key points
- Searchable found 33% of surveyed US users had sworn at or insulted chatbots
- Rude complaints were corrected less often than plain corrections, the report found
- Anthropic introduced new policies to prevent abuse in Claude chats
Details
TechRadar reports that Searchable analyzed an opt-in survey covering 8 million messages exchanged with Claude, Gemini and ChatGPT over 12 months, from September 2025 to September 2026. The survey found an average of 19.1 derogatory words per 10,000 words in chatbot exchanges.
In Searchable’s analysis, 31% of complaints containing swearing were eventually corrected, compared with 41% of plain corrections. At the end of those conversations, 35% of the issues raised in sweary complaints remained unfixed, compared with 24% of issues raised in plain corrections; the number of replies needed when an issue was fixed was about the same.
The report also found regional differences: Nevada had 22.7 rude or offensive words per 10,000, while New Jersey had 11.4. TechRadar says Anthropic introduced new usage policies to block cruelty in Claude chats; the article does not detail the policies.
Written from coverage as of
Images
- TechRadar
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh