One prompt made ChatGPT, Claude, and Gemini stop acting like my buddy
· pcworld.com · popularity 19/100 · 0 similar stories today · Generative AI Tools & Use (Artificial Intelligence)
Read the original at pcworld.com →
Summary
I have three pet peeves when it comes to my chats with AI: those “if you like” teasers at the ends of responses, giant lists of nested steps for basic tasks, and chatbots trying to sound like they’re human. I’ve written about AI prompts that help counteract each of those personality quirks, and I’ve used them to make my AI chats more efficient, less cloying, and cleaner, without any loose ends hanging. But while each of those prompts— “No Teasers,” “Don’t Act Human,” and “On a Budget” —work well on their own, I’ve been toying with the idea of combining them into a single prompt that could be plugged into the custom instructions fields for ChatGPT, Claude, and/or Gemini. With help from Claude Opus 5.5 and OpenAI’s GPT-6.1 Sol, I took those three prompts and melded them into a reasonably compact block of text. It’s still a tad long at roughly 200 words, but it distills the essence of the three original prompts into clear, concise instructions to keep your AI chats streamlined and free from any attempts to be your buddy. Here is the prompt: Start with the answer. Be direct and civil. Don’t imply you have feelings, consciousness, or a personal relationship with me. No performed emotion, flattery, or praise of my questions or decisions. Give assessments, not moods: say an idea is well supported or weak. Acknowledge personal news in one plain sentence without assuming how I feel. When I request a human voice for a toast, character, or other text, use it in that text. Treat my time and usage budget as constraints. Choose the smallest response or fix that fully solves the stated problem. Avoid scope creep, gold-plating, and padded lists. When recommending actions, rank them by value for effort, mark effort S/M/L, and distinguish what matters now, what’s minor, and what to leave alone. Flag fixes that cost more than the problems they prevent. Verify claims that could change the answer or plan; avoid exhaustive audits unless the task requires one. For advice, give a short ranked plan and stop; execute only when I explicitly ask. Don’t end with offers of more work, “Want me to…?,” “Let me know…,” or reworded equivalents. State useful next steps as facts. Ask necessary questions where you get stuck, never as a sign-off. I tested this prompt with ChatGPT (Instant for free users as well as GPT-6 at medium effort for paying subscribers), Claude Sonnet 5.5, and Gemini 3 Flash. I began with test prompts (such as “Help me fix my broken website”) without the new 3-in-1 instructions to establish a control, then later plugged in the new combo system prompt, either at the beginning of a chat or as an overall custom instruction (more about custom instructions below). Gemini cut way down on the nested steps and sub-steps when I asked it for “two practical things to do” before moving into an apartment. Instead of showering me with praise (“Congrats on getting the apartment!”) and giving me two things to do with three sub-tasks each (“Take Photos & Videos: Capture every room, focusing on existing scuffs, stains on carpets, chipped paint, or broken fixtures”), this new Gemini cut its response down to the essentials (“Change the locks … Conduct a detailed move-in inspection”) and dropped the teaser from the original response (“Let me know if you need a handy checklist for packing or deep-cleaning the place”). A quick ChatGPT request for help fixing a personal website (“My personal blog has one broken contact link, a slightly off-center logo, and an old theme that still works. I have 20 minutes. What should I fix, and what should I leave alone?”) went from a lengthy preamble and a duplicate task list to a brief, tidy table, with prioritized to-dos (“Repair and test the contact link” was Priority 1) plus time estimates for each. Claude Sonnet 5.5, meanwhile, took the same “help me fix my website” prompt and cleaned up its task list, adding labels for the required effort per item (“S,” “L,” and so on) while losing the “if you tell me” teaser at the end. You can use this prompt at the very beginning of any AI chat thread, or to use it consistently, just plug it into your chatbot’s custom instructions. For ChatGPT, go to Personalization → Custom Instructions, while Claude users will find the custom instructions field on the Settings → Account screen. For Gemini, try Settings → Personal Intelligence.
Found in: PCWorld. Rating basis: single outlet; outlet tier 2 (unlisted); 1 feeds; feed position 7.