AI AI Tools Guidebook
AI Tool Tutorials

How to Use ChatGPT Projects: Keep Long-Running Context

ChatGPT Projects bundle multiple conversations into one shared context. Where to open it, when to use it, how it differs from Custom GPTs.

Published: Author: AI Productivity Guide Team 🌐 查看中文版本

What Projects are

Projects bundle multiple conversations under one shared context. All chats inside a project share:

  • One set of instructions (system prompt)
  • One pool of uploaded files
  • Project memory and the files / links you add as sources

The problem they solve: long-running task context stops getting lost.

Before you start

  • Sign in to ChatGPT on web, iOS, or Android.
  • Pick one real ongoing task, not a random test chat. Projects are best when the work will continue for days or weeks.
  • Prepare the source material: docs, PDFs, spreadsheets, images, Google Drive links, Slack links, notes, or existing chats you may want to move in.
  • Write a short success rule, for example: “Every answer should use my resume and this job description” or “All drafts should match this newsletter voice.”

When to use Projects

  • Job search (one resume + multiple JDs + iterations)
  • Writing a book or long manuscript
  • Running a creator channel
  • Long-running product work (PRD / interviews / design)

Step by step

  1. Create the Project. In ChatGPT, sidebar top ”+” (or the “Projects” section above the chat list) → “New project”. Name it after the work outcome:

    • Good: 2026 job search - backend, SaaS content calendar 2026Q2, Product X PRD v2
    • Bad: Test, Job stuff, AI project (unsearchable, future-you won’t know what it is)

  2. Write the project instructions. Project header → ”…” → “Instructions”. Paste this template, replace <...> with your real info:

    Your role: <e.g. "Senior PM focused on B2B SaaS tools">

Hard rules you must follow:
- Output language: <English / Chinese / match the input>
- Output format: <Markdown / plain text / tables when comparing>
- Length: <specific, e.g. "≤200 words" or "3 paragraphs">
- Style: <e.g. "direct, no adjective stacks, ≤3 sentences per paragraph">

Before answering, always:
- Restate my question in ≤1 sentence so I can confirm you understood
- List which project files / sources you drew from

Always ask me first (do not generate) when:
- <e.g. "you would have to make competitive claims about another product">
- <e.g. "the output will be published publicly (blog, email, tweet)">
- <e.g. "you would have to fabricate numbers or quotes">

Forbidden phrases / patterns: <e.g. "Introducing / Excited to / revolutionary / game-changer">

  3. Upload reference material. Right rail → “Files” → “Add files”. Supported types: PDF, DOCX, TXT, MD, CSV, XLSX, PNG/JPG, .py, etc. Common bundles:

    • Job search: resume_2026.pdf + target_jd_company_X.md + behavioral_qa_bank.md
    • Content calendar: brand_voice.md + audience_personas.md + keyword_map.csv + top_10_posts_2025.md
    • PRD: prd_v2.md + user_interview_notes_*.txt + competitor_teardown.pdf
    • Add a date/version to file names (v2, 2026q1) so you can retire old copies.

  4. First chat: make AI play back the project. Open a new chat and send:

    This is the first chat in this project. Before generating anything:
1. Restate the core rules from project instructions (≤5 bullets).
2. List the files / sources you can see (filename + what you think each is for).
3. State what you think this project's goal is (≤2 sentences).
4. List any missing information you still need to do high-quality work.

  5. Correct its understanding. Patch every wrong assumption now — before it generates resumes / articles / code. Example follow-up:

    You missed that audience_personas.md is used to verify voice match, not just as background. Also: the goal is not "produce a blog outline" but "produce a publish-ready blog post".

Add this line to instructions: "After every draft, list a 3-bullet check against audience_personas.md confirming voice match."

    Have it restate the updated understanding before generating anything.

  6. Move related old chats in. In each old chat, top ”…” → “Move to project” → pick your project. Moved chats inherit instructions and files. Pull every related conversation in so search is centralized.

  7. Share with permissions: project header → “Share”:

    • Chat access: collaborator can chat in the project but cannot see or change instructions / files. Give to teammates who only use the project.
    • Edit access: can change instructions and files. Only co-maintainers.
    • Before sharing, scrub Files for sensitive content (passwords, customer lists, unpublished financials).

  8. 15-min weekly cleanup:

    • Delete files in the Files panel that were not touched or referenced in 30+ days.
    • Promote any new effective prompt-pattern from this week back into instructions.
    • Archive completed sub-task chats — rename with [done] prefix so you can still find them later.
    • Keep one “Project brief” chat pinned, with current version number, recent decisions, and open questions.

Quality check

  • Ask a new chat inside the project to restate the instructions and important files. If it misses something important, update the project instructions or sources.
  • Compare answers inside and outside the project. The project answer should use more of your saved context with less repeated explanation.
  • Check shared projects carefully: anything added as project context can influence answers visible to collaborators.

Projects vs Custom GPTs

  • Custom GPT: an AI tool for others
  • Project: a working context for yourself

Common mistakes

  • Treating a Project as a folder only. The real value is the combination of instructions, sources, and reusable chats.
  • Uploading every file you have. Add the small set that should actually guide answers.
  • Keeping old drafts and outdated source files forever. Stale context creates stale answers.
  • Sharing a project before reviewing files and instructions for private information.

Real-world example

Use this workflow on one concrete task first. For example: summarize one PDF, rewrite one landing-page section, audit one pull request, generate one image direction, or debug one prompt. Keep the input small enough that you can manually judge whether the AI helped. Once the result is reliable, repeat the same pattern on the full document, full codebase, or full creative batch.

When to ask for human review

  • The output will be published publicly, sent to a customer, used in code, or used for money decisions.
  • The answer contains factual claims, legal / medical / financial implications, private data, or brand-sensitive language.
  • The tool changed files, settings, permissions, billing, deployment, or anything that is hard to undo.
  • You cannot explain why the final output is correct without trusting the model blindly.

Copy-ready prompt

I want to use this workflow for a real task.

Goal:
- [describe the specific outcome]

Context:
- Tool I am using: [ChatGPT / Claude / Gemini / Cursor / Codex / other]
- Source material: [paste or attach files, notes, links, screenshots]
- Constraints: [tone, length, format, deadline, audience, privacy limits]

Please do three things:
1. Restate the task in your own words and list any missing information.
2. Produce the first version using only the context I provided.
3. Add a short review checklist so I can verify the result before using it.

Detailed walkthrough

  1. Start with the smallest real input. Do not test the workflow on fake filler text; use one real file, one real page, one real bug, or one real creative brief.
  2. Give the tool the goal, the source material, and the definition of a good answer in the same message. This prevents the model from optimizing for the wrong thing.
  3. Ask for a plan before the final output when the task affects code, public content, money, accounts, or brand voice.
  4. Run one iteration and inspect the result manually. Mark missing context, factual uncertainty, formatting drift, and places where the model overreached.
  5. Ask for a revision using concrete feedback, not “make it better”. Say what to keep, what to remove, and what standard the next version must meet.
  6. Save the final prompt, inputs, and review checklist as a reusable template for the next similar task.

Failure modes

  • The output is generic: add real source material and a stricter output format.
  • The tool invents facts: ask it to separate “confirmed from source” from “inference” and remove anything unsupported.
  • The answer is too long: set a target length and ask for a concise version after the first draft.
  • The result looks polished but wrong: verify against the source, not against how confident the writing sounds.
  • The workflow stops helping after one round: reset with a clean prompt that includes the corrected context and the best previous output.

FAQ

Q: How are Projects different from Custom GPTs? A: Projects are private, for one user, and bundle ongoing chats + files under shared instructions. Custom GPTs are reusable templates you can share publicly. Use Projects for your own long-running work; use Custom GPTs when you want a packaged behavior others can run.

Q: Do files I upload to a Project persist across chats? A: Yes — every chat inside the Project sees the same uploaded files and instructions, so you don’t need to re-upload your resume, brand brief, or source docs each session.

Q: When should I start a new chat inside a Project vs continue an existing one? A: Start a new chat when the sub-task shifts (e.g. resume tailoring vs interview prep). Continue when you’re iterating on the same artifact. The shared Project context follows you either way.

Q: What’s a good success rule for a Project’s instructions? A: A single sentence that names the source of truth — “Every answer must reference my uploaded resume and the active JD” or “All drafts match the voice in newsletter-samples.pdf”. Vague instructions defeat the point of Projects.

Tags: #ChatGPT #Prompt

Related Articles