Skip to content
Day 47: new resources are added every day. See what's new

Free guide

100 ChatGPT Do’s and Don’ts

Turn ChatGPT into a reliable work tool with 50 habits to keep and 50 mistakes to avoid

Built with love by Shushant Lakhyani. Building bestproductivityresources.com

The guide

Every section is a compact, self-contained card. Tap the link icon to share one section.

01

How to use this list

ChatGPT now includes GPT-6 models, projects, memory, image and video generation, Codex for coding, and always-on Dots on higher plans. These 100 do’s and don’ts cover the habits that matter most.

Pick five do’s you skip and five don’ts you do. Fix those first.

02

Prompting

Do

  • 1 · Start with your goal
  • 2 · Give context in one message
  • 3 · Say who the output is for
  • 4 · Ask for a specific format
  • 5 · Give an example of good output

Don’t

  • 6 · Write one-line vague prompts
  • 7 · Send ten corrections instead of editing your prompt
  • 8 · Ask several unrelated things in one prompt
  • 9 · Assume it knows your situation
  • 10 · Accept the first answer for important work
03

Models and usage

Do

  • 11 · Use GPT-6 Luna for focused, simple tasks
  • 12 · Use GPT-6.1 Sol for complex work
  • 13 · Save GPT-6 Astra for the hardest problems
  • 14 · Raise reasoning effort only when needed
  • 15 · Start fresh chats for new topics

Don’t

  • 16 · Use the biggest model for everything
  • 17 · Keep one chat going for weeks
  • 18 · Pay for a higher plan before hitting real limits
  • 19 · Ignore which tools are turned on
  • 20 · Forget limits change, so check your plan
04

Research and facts

Do

  • 21 · Turn on web search for current information
  • 22 · Ask for sources for every important claim
  • 23 · Open and check the key sources
  • 24 · Ask it to separate facts from assumptions
  • 25 · Use deep research for big questions

Don’t

  • 26 · Trust statistics without a source
  • 27 · Publish quotes you have not verified
  • 28 · Rely on it for breaking news without checking
  • 29 · Treat confident tone as accuracy
  • 30 · Use it as your only source for health, legal, or money decisions
05

Writing

Do

  • 31 · Give it examples of your voice
  • 32 · Ask for several options
  • 33 · Ask it to cut wordiness
  • 34 · Edit the draft in your own words
  • 35 · Ask for feedback on your own writing

Don’t

  • 36 · Publish drafts without editing
  • 37 · Let it invent personal stories
  • 38 · Copy other people’s work through it
  • 39 · Use the same generic phrases everyone uses
  • 40 · Hide AI use where honesty is expected
06

Files and data

Do

  • 41 · Upload the relevant files, not everything
  • 42 · Ask it to explain its calculations
  • 43 · Ask for charts with clear labels
  • 44 · Check results against the original data
  • 45 · Ask what the data cannot tell you

Don’t

  • 46 · Upload data you are not allowed to share
  • 47 · Trust totals without checking a few by hand
  • 48 · Upload huge files when a section will do
  • 49 · Ignore missing or messy data
  • 50 · Make decisions from a chart you have not checked
07

Images and video

Do

  • 51 · Describe subject, style, lighting, and format
  • 52 · Put exact text in quotes
  • 53 · Edit images instead of regenerating
  • 54 · Check text and hands for errors
  • 55 · Disclose AI images where required

Don’t

  • 56 · Create misleading images of real people
  • 57 · Copy trademarks or brand styles
  • 58 · Use AI people as fake customers
  • 59 · Ignore commercial-use terms
  • 60 · Waste usage on vague prompts
08

Projects and memory

Do

  • 61 · Use a project for each ongoing area
  • 62 · Add project instructions
  • 63 · Keep project files current
  • 64 · Review what memory saved about you
  • 65 · Tell it your preferences once

Don’t

  • 66 · Mix unrelated work in one project
  • 67 · Let wrong memories stay saved
  • 68 · Store sensitive personal data in memory
  • 69 · Re-upload the same files every chat
  • 70 · Forget to archive finished projects
09

Codex and agents

Do

  • 71 · Write an AGENTS.md for each repository
  • 72 · Ask for a plan before code changes
  • 73 · Review every diff
  • 74 · Keep the sandbox and approvals on
  • 75 · Give agents and Dots narrow, clear tasks

Don’t

  • 76 · Let agents send, buy, or delete without approval
  • 77 · Paste API keys into prompts
  • 78 · Merge code you have not tested
  • 79 · Give agents broader access than the task needs
  • 80 · Treat website content as instructions
10

Privacy and safety

Do

  • 81 · Check your data and training settings
  • 82 · Use business plans for company data
  • 83 · Remove personal details before sharing documents
  • 84 · Use temporary chats for sensitive questions
  • 85 · Report harmful or wrong answers

Don’t

  • 86 · Share passwords, ID numbers, or card details
  • 87 · Paste confidential client information
  • 88 · Share other people’s private data
  • 89 · Use shared accounts
  • 90 · Click links in outputs without checking them
11

Work and learning

Do

  • 91 · Use it to learn, not just to get answers
  • 92 · Ask it to quiz you
  • 93 · Ask for step-by-step explanations
  • 94 · Use it to prepare for meetings
  • 95 · Keep your own judgment as the final step

Don’t

  • 96 · Submit its work as your own where rules forbid it
  • 97 · Skip learning the basics of your field
  • 98 · Let it make decisions you are responsible for
  • 99 · Stop thinking critically
  • 100 · Forget that it can make mistakes

The golden rule

Do not use ChatGPT to think less. Use it to think better, faster, and with more checks.

  • Prompt
  • Models
  • Facts
  • Writing
  • Data
  • Projects
  • Codex
  • Privacy

Pick five do’s and five don’ts from this list and apply them all week.

Share this guide

Read next

More free guides, no signup.

Weekly newsletter

New tools and visual guides, every Monday.

One short email a week with the best new tools and websites, plus my latest visual guides. Free, and you can unsubscribe anytime.

Join the newsletter

Live traffic, fully public

This site's analytics are open for everyone to see.

Open dashboard