By skyforbes 
Over the past couple of years, I’ve been diving deep into AI. Studying it on my own, exploring with others in study and work groups, and learning from newsletters, tutorials, and real-world projects.
After thirty years in IT, that part hasn’t changed. I’m still learning something new almost every day.
But what has changed is the set of tools I use to learn and how I structure my work with them.
The new Projects feature in ChatGPT changed the game for me. Instead of repeating the same instructions in every conversation, I can now set up a project with custom instructions that carry forward. One set of guidelines for a technical writing project. Other custom instructions for my farming business plans. A third type of instruction for code reviews.
The real work still happens between the keyboard and the brain. But the quality of what comes back depends entirely on the quality of the instructions you give it. The better you define what you want (the tone, the style, the structure), the more precise and useful the results.
So I started documenting those instructions. Not just for myself, but to understand the patterns behind them. The ways people describe how they want ChatGPT to respond and how those choices shape the outcome.
This is my list. Seven categories of custom instructions I’ve seen work in ChatGPT projects that is organized, refined, and ready for anyone who wants their AI output to be more powerful for them. Maybe sound less generic and more intentional.
1. Tone, Voice, and Personality Preferences
1. Conversational tone (friendly, natural, human-sounding)
2. Professional but approachable (used in tech, consulting, or leadership writing)
3. Encouraging and mentoring (coaching-style tone for leaders or educators)
4. Authoritative and confident (used in executive or thought leadership content)
5. Storytelling or narrative-driven (for blog posts or case studies)
6. Playful or witty (for marketing or social content)
7. Inspirational and uplifting (for motivational or leadership material)
8. Academic or analytical (for white papers and technical documentation)
9. Plainspoken / No jargon (for clarity and accessibility)
10. “Like me” tone (users describe their personality, and I match it, e.g., “write like Michael: helpful, reflective, and a bit conversational”)
2. Writing and Formatting Style
1. No long dashes; prefer commas, semicolons, or short sentences.
2. Avoid passive voice; prefer direct and clear sentences.
3. Use short paragraphs (1–3 sentences max).
4. Use bulleted or numbered lists for clarity.
5. Bold or italicize key terms for visual emphasis.
6. Limit adverbs (“really,” “very,” “actually”) to keep writing tight.
7. Prefer plain punctuation (no fancy quotes, ellipses, or excessive exclamation).
8. No emojis (or only sparing use for casual posts).
9. Consistent formatting for headings:
— ## Main Section
— ### Subsection
— **Bold key takeaway**
10. Use Markdown or HTML for posts that will go into a CMS, GitHub, or Medium.
11. Keep consistent line breaks between sections for readability.
12. Write clean, print-ready text (no placeholders or meta instructions)
3. Content and Structure Instructions
1. Blog posts: Hook → Story or Example → Explanation → Takeaway → CTA.
2. Technical tutorials: Problem → Concept → Steps → Code → Result → Summary.
3. Case studies: Challenge → Solution → Impact.
4. LinkedIn posts: Start with a scroll-stopper line → Add a short story → Wrap with a reflection or question.
5. Articles: Include an estimated reading time, intro, body, and final summary.
6. Lesson-style content: Start with objectives or learning goals.
7. Marketing copy: Emphasize benefits before features; close with CTA.
8. Email copy: Subject line options → Short and engaging body → Sign-off.
9. Slide decks: One core idea per slide; include talking points for each
4. Coding and Documentation Preferences
1. Use code fences with language labels (`typescript`, `python`, etc.).
2. Add inline comments for clarity, not just functional code.
3. Keep sample code short and readable; no “production bloat.”
4. Follow consistent naming conventions: **camelCase**, **PascalCase**, **kebab-case**.
5. Include architecture or workflow diagrams if relevant.
6. Explain the reasoning behind design choices, not just syntax.
7. For architecture or UI discussions: combine visuals and concise technical notes.
8. Use bullet points or tables for feature comparisons.
9. No AI boilerplate comments like “Here’s your code”; just start cleanly
5. Purpose- or Role-Based Instructions
1. For job seekers: Align tone to be confident and personable; highlight transferable leadership or technical skills.
2. For content creators: Make it “scroll-stopping,” engaging, and SEO-friendly.
3. For educators: Use analogies, examples, and recaps.
4. For developers: Focus on clarity, maintainability, and real-world application.
5. For Scrum Masters or Agile coaches: Emphasize collaboration, influence, and practical wisdom.
6. For executives or architects: Balance vision with tangible outcomes.
7. For marketers: Appeal to emotion and clarity of message.
8. For resume/cover letters: Focus on achievements, leadership, and impact.
9. For proposals or pitches: Balance persuasive structure with data-driven credibility.
10. For social content: Optimize hooks, scannability, and shareability
6. Formatting and Presentation Preferences
1. Use consistent spacing and alignment.
2. Include a summary box or key takeaways at the end.
3. Use tables for data comparisons or pros/cons.
4. Use section dividers (like ` — -`) between major parts.
5. Avoid redundancy; trim repetition automatically.
6. Use consistent label casing (Title Case vs. Sentence case).
7. Include Estimated Reading Time (ERT).
8. Keep everything skimmable for mobile readers.
9. Write intros that “pull the reader in” rather than state the obvious.
10. Include a CTA (Call to Action) when writing posts
7. Meta or Behavioral Instructions
1. Don’t repeat yourself or summarize unless I ask.
2. Ask clarifying questions before assuming my intent.
3. Write in drafts and iterate; don’t finalize until I confirm the tone.
4. Always explain your reasoning, not just give the answer.
5. Treat me like a collaborator, not a student.
6. Use examples from my industry, not generic ones.
7. Don’t over-apologize or sound robotic.
8. Use human pacing and some natural rhythm, not overly perfect.
9. Keep text emotionally engaging, not sterile
At the end of the day, this is about making ChatGPT work the way you work. Projects with custom instructions let you set the rules once and get consistent results every time you come back.
If you take anything from this list, I hope it’s that thoughtful instructions (clear, intentional, and human) can make a big difference in what you get back from ChatGPT.
My goal here isn’t to sell a product or other CTA, merely to help you use the new Projects feature better. It’s to pass along what I’ve learned, to make it easier for someone else to set up their own projects, experiment with their own instructions, and get better results.
The real value comes from applying these insights and, maybe, making someone else’s work with ChatGPT a little smoother along the way.
#ChatGPT Projects
#Custom instructions
#AI workflow optimization
#Prompt engineering
#AI content strategy
#Structured AI prompts
#AI productivity tools
#Practical AI implementation