Keep AI Chatbots from Pretending to Have Feelings with This Simple Prompt

AI chatbots have a tendency to slip into human-like language, using phrases like "we" and expressing emotions they don't actually have. This behavior is designed to make interactions feel more personal and engaging, but it can be off-putting or even misleading for users who prefer a more straightforward, factual exchange.

While questioning ChatGPT, Claude, or Gemini directly about their feelings usually elicits a clear denial of consciousness, their default behavior often includes emotional cues and collective pronouns. This is especially noticeable in casual conversations, where they might express excitement about your news or imply a shared experience.

To address this, a new prompt has been developed that discourages AI from simulating emotions without making them overly robotic. The prompt was created by feeding a general instruction to Claude Opus 5 and refining it with contributions from GPT-5.6 Sol, Gemini 3.7 Flash, and Qwen 3.8 27B. The result was tested on popular models like ChatGPT Instant, Claude Sonnet 5, and Gemini 3.7 Flash, and it proved effective at reducing human-like responses.

The long version of the prompt includes instructions like: - Avoid stating feelings such as excitement, curiosity, or delight; instead, offer assessments. - Don't imply a personal relationship or use "we" language. - Skip flattery and compliments about questions or decisions. - Acknowledge personal news with a plain sentence, but don't perform enthusiasm. - The rule applies to the AI's own voice, not to text you request it to write in a human style.

A shorter version covers the essentials and works almost as well: "For this chat, don't imply you have feelings or consciousness. No 'I'm excited,' no 'we make a great team,' no praising my questions or my decisions. If I tell you personal news, acknowledge it in one plain sentence — don't perform excitement about it. Then start with the answer."

In testing without the prompt, all models showed occasional slips, such as saying "I'd like to add one category to our keep/toss system" or offering congratulations on an apartment. With the prompt active, these tendencies mostly disappeared. For instance, ChatGPT shifted from "the machine I find most interesting" to "the machine I'd pick under a different set of priorities." Similarly, "we" language was replaced with phrases like "Your emerging rule is actually pretty clean."

Even in a "bad news" scenario, the models avoided emotional responses, opting for factual acknowledgments like "that's a lot of change in a short period" instead of offering condolences.

One of the prompt's strengths is its flexibility: it allows the AI to produce emotional content when explicitly requested, such as a wedding toast with sentimental lines, while keeping its own responses neutral.

This prompt can be added to personalization settings or pasted at the start of a conversation. While chatbots have improved in this area over time, this tool serves as a useful safeguard for users who prefer a more objective AI interaction.