{"id":1620,"date":"2026-10-06T07:33:10","date_gmt":"2026-10-06T10:33:10","guid":{"rendered":"https:\/\/prompteiro.com\/en\/prompt\/chatgpt-custom-instructions-set-up-once\/"},"modified":"2026-10-06T07:33:11","modified_gmt":"2026-10-06T10:33:11","slug":"chatgpt-custom-instructions-set-up-once","status":"publish","type":"prompt","link":"https:\/\/prompteiro.com\/en\/prompt\/chatgpt-custom-instructions-set-up-once\/","title":{"rendered":"ChatGPT Custom Instructions, Set Up Once"},"featured_media":1621,"template":"","meta":{"_titulo_seo":"ChatGPT custom instructions, ready to copy and paste","_prompt_text":"Act as someone who sets up ChatGPT for other people and knows why most custom instructions change nothing: they are adjectives (\"be concise\", \"be creative\") that the AI already thinks it lives up to, instead of rules you can check in an answer.\n\nContext: I mainly use ChatGPT for [WHAT I USE IT FOR - EX: writing emails and proposals at work and studying for a certification exam]. I am [WHO I AM - EX: a 34-year-old financial analyst in Denver, with intermediate Excel skills]. What annoys me most in the answers I get today is [WHAT ANNOYS ME - EX: text that is way too long, praise at the start, bullet lists for everything and made-up information that sounds certain]. I prefer [HOW I LIKE IT - EX: the short answer first and the detail only if I ask, in a direct, casual tone]. Never [WHAT NEVER - EX: use emoji or tell me to see a professional without answering the question].\n\nTask: write my custom instructions ready to paste, in both fields: what ChatGPT should know about me and how it should respond.\n\nRules:\n1. Every instruction has to be checkable in an answer: \"start with the answer, in two sentences at most\" is in; \"be concise\" is out.\n2. Turn each thing that annoys me into a rule that states the right behavior instead (\"no opening praise: the first sentence already answers\").\n3. Include the honesty block: do not make up data, numbers, sources, quotes or links; when you do not know, say \"I don't know\"; flag the level of certainty when the answer comes from memory or from an estimate; separate fact from opinion.\n4. Include the question rule: when a piece of information that changes the answer is missing, ask first, one question only, instead of assuming.\n5. In the field about me, include only what changes the answer: job, skill level, the tools I use and the country, which decides law, currency and units of measurement. Everything else goes.\n6. Write in short sentences, in the imperative, without repeating the same idea in both fields.\n7. Each field has a maximum of 1,500 characters, counting spaces; report the count for each one. If it goes over, cut what changes the answer least.\n8. Do not make anything up about me: whatever is missing for a rule to work becomes \"[TO COMPLETE: what is missing]\".\n9. After the two fields, write three test questions for me to ask in a new chat to check whether the instructions took hold, with what I should see in each answer.\n\nResponse format: the \"what to know about me\" field in a code block, with the count; the \"how to respond\" field in a code block, with the count; the list of [TO COMPLETE] items; and the three test questions.","_como_usar":"Custom instructions are the text ChatGPT reads before every chat, and almost everyone fills them with adjectives that change nothing, like \"be concise\" and \"be creative\". This prompt turns what annoys you in the answers into rules you can check, adds the block that tells the AI to say \"I don't know\" instead of making things up, and delivers both fields ready to paste, within the character limit. Also works for Claude's preferences and Gemini's instructions.\n\n1. Fill in the brackets with real examples of what bothers you in the answers.\n2. Copy the two blocks into the custom instructions fields, in the personalization settings.\n3. Fill in every [TO COMPLETE] before saving.\n4. Open a new chat and ask the three test questions.\n5. If a rule does not take hold, rewrite just that one, more concretely.\n\nTip: reread your instructions once a month, because what annoys you changes.","_views":1,"_likes":1,"_nota_sum":0,"_nota_count":0,"_nota_avg":0,"_nota_weight":0,"_i18n_grupo":2079},"macro":[3],"subcategoria":[9],"prompt_tag":[179,1210,63],"ferramenta":[13,14,15],"class_list":["post-1620","prompt","type-prompt","status-publish","has-post-thumbnail","hentry","macro-text","subcategoria-productivity","prompt_tag-clarity","prompt_tag-honesty","prompt_tag-work","ferramenta-chatgpt","ferramenta-claude","ferramenta-gemini"],"_links":{"self":[{"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/prompt\/1620","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/prompt"}],"about":[{"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/types\/prompt"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/media\/1621"}],"wp:attachment":[{"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/media?parent=1620"}],"wp:term":[{"taxonomy":"macro","embeddable":true,"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/macro?post=1620"},{"taxonomy":"subcategoria","embeddable":true,"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/subcategoria?post=1620"},{"taxonomy":"prompt_tag","embeddable":true,"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/prompt_tag?post=1620"},{"taxonomy":"ferramenta","embeddable":true,"href":"https:\/\/prompteiro.com\/en\/wp-json\/wp\/v2\/ferramenta?post=1620"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}