Welcome to the ChatGPT Bible
A practical AI Con Todo guide for non-technical people who want to understand every major ChatGPT feature, use it safely, and turn it into a daily operating system for work and life.
Chat, models, search, files, images, charts, and writing blocks.
Projects, pinned chats, Library, memory, and storage management.
Apps/connectors, scheduled tasks, GPTs, agent mode, deep research.
Permissions, Lockdown Mode, finance limits, memory review, team governance.
10-Day ChatGPT Power User Path
This is the beginner-safe path for learning ChatGPT as a daily operating system: chat, files, memory, apps/connectors, scheduled tasks, finance, GPTs, agent mode, and review habits.
The promise
10-day path
| Day | Level | Module | Outcome |
|---|---|---|---|
| 1 | Beginner | ChatGPT Basics | Know chat, models, memory, files, search, and privacy settings. |
| 2 | Beginner | First Deliverable | Create a report, plan, resume, chart, image, or spreadsheet. |
| 3 | Beginner | Organize Work | Use Projects, pinned chats, Library, and document blocks. |
| 4 | Builder | Prompt Frameworks | Use goal, context, output, rules, and review criteria. |
| 5 | Builder | Apps and Connectors | Connect tools safely and use @ mentions or app selection. |
| 6 | Builder | Scheduled Tasks | Create reminders, recurring briefings, and monitoring tasks. |
| 7 | Power | GPTs | Create a reusable custom assistant with instructions and knowledge. |
| 8 | Power | Agent Mode | Delegate multi-step research and action workflows with supervision. |
| 9 | Advanced | Finance, Health, Deep Research | Use specialized experiences with stronger review habits. |
| 10 | Advanced | Team Playbook | Create policies for apps, memory, GPTs, tasks, and review. |
What ChatGPT Is in 2026
ChatGPT is no longer just a chat box. It is a workspace for conversation, writing, files, images, charts, apps, scheduled tasks, custom GPTs, agentic work, memory, and specialized experiences like Finances.
Writing, research, analysis, planning, images, spreadsheets, email drafts, personal organization, learning, coding help, and multi-step tasks.
A replacement for professional review in legal, tax, medical, investment, or irreversible decisions.
Current model picker mental model
| Option | Plain meaning | Use when |
|---|---|---|
| Instant | Fast everyday answer. | Simple questions, drafts, quick help. |
| Medium / High | More reasoning effort. | Planning, analysis, problem solving. |
| Extra High / Pro | Highest effort for paid tiers. | Hard strategy, math, research, sensitive review. |
Memory, Personalization, and Privacy
ChatGPT memory can personalize responses using saved memories, past chats, custom instructions, files, and connected apps where available. The June 2026 updates add better memory controls and a memory summary.
Beginner rules
| Setting | Use it for | Safety habit |
|---|---|---|
| Memory | Preferences, goals, recurring context. | Review and delete stale memories. |
| Temporary Chat | One-off private or irrelevant chats. | Use when you do not want memory updates. |
| Memory sources | See what shaped a response. | Correct outdated assumptions. |
| Lockdown Mode | Reduce network/external-service risk. | Use for sensitive or security-conscious work. |
Memory cleanup prompt
Review what you seem to remember about me and my work.
Tell me:
1. What still seems useful
2. What might be outdated
3. What could cause wrong assumptions
4. What I should delete, correct, or keep
Do not add new memories from this review unless I explicitly ask.Files, Library, Images, and Writing Blocks
ChatGPT can work with uploaded files and files it creates. Library now stores reusable files across plans, while images have their own images area. Writing blocks support long-form work and can be saved to Library.
What belongs where
| Feature | Use it for | Tip |
|---|---|---|
| Library | Uploaded and created files: docs, spreadsheets, presentations, images. | Delete old files from Library or storage settings. |
| Writing blocks | Reports, PRDs, essays, notes, long docs. | Ask for full-screen editor when revising. |
| Images | Generated image browsing and reuse. | Use exact style, aspect ratio, and edit instructions. |
| Charts | Bar, line, pie, scatter visuals in answers. | Ask for a chart when trends matter. |
File analysis prompt
Analyze this file for a non-technical reader.
Please provide:
1. What this file is
2. Key findings
3. Risks or missing information
4. A simple summary table
5. Recommended next actions
6. Questions I should ask before trusting itProjects: Your Workspaces
A Project is a dedicated workspace for one ongoing topic. Instead of starting from scratch in every chat, you set the context once and reuse it. A Project bundles four things in one place: the chats you have inside it, files you upload, project instructions (a reusable "master prompt"), and its own project-scoped memory. Every new chat you open inside the Project already knows that context, so you stop re-explaining yourself.
Good mental model: a normal chat is a one-off conversation. A Project is a living home you keep coming back to and keep adding to. People use Projects for things like a marketing workspace, a personal finances tracker, a "research assistant," or a writing/brand voice workspace.
Create a Project and add instructions
Click New project in the sidebar, give it a clear name tied to its purpose (for example "Marketing" or "Investing"), and optionally pick a color so it's easy to spot. To remove one, open the three-dots menu and choose Delete project.
The most valuable thing to set up first is project instructions — a "master prompt" of a few thousand characters that applies to every chat in the Project. It's separate from your account-level Custom Instructions (covered in Memory, Personalization, Privacy) and lets you focus the assistant on this one topic: set a goal, a format, a tone, and standing rules so you don't restate them every time.
A simple, reusable structure for project instructions is Goal, Format, Tone, and "ask if unclear." Copy this and fill in the brackets:
Goal: [what this project is for — e.g. help me plan and write content for my YouTube channel].
Audience / context: [who the output is for and any background you'd otherwise repeat].
Format: [how to respond — e.g. keep answers under 5 bullet points; use a table when comparing things].
Tone: [e.g. casual and direct; no hype].
Standing rules: [anything that should always apply — e.g. always cite sources; flag anything I should verify].
If my request isn't clear or you need more information, ask me before answering.Project knowledge (files)
Use Add files to attach reference material to the whole Project. Every chat inside the Project can then read those files, so you don't re-upload per chat. Supported types include PDF, DOCX, XLSX, CSV, TXT, JSON, HTML, images, code files, and more.
There's a limit on how many files one Project can hold, and it varies by plan. As of 2026 the range is roughly 5 to 40 files (Free at the low end, top paid tiers at the high end) — check the official plans page for the current numbers, since they change. You can create unlimited Projects, so if you hit a cap, split work into separate Projects rather than cramming everything into one. Quality beats quantity: a few clean, well-edited documents work better than a pile of noisy ones.
Project memory vs global Memory
This control is easy to miss and worth setting deliberately. Each Project has its own memory setting, separate from the global Memory covered in Memory, Personalization, Privacy. Facts ChatGPT learns inside a Project stay in that Project — they do not bleed into your main chat or your other Projects.
The Project can read your regular ChatGPT memory from outside it. Good when you want it to use everything it already knows about you (for example a launch plan that should reflect your whole situation).
The Project uses only its own memories, and they stay hidden from outside. Good for focused or sensitive work — it won't pull in irrelevant facts from past unrelated chats.
Pick the mode on purpose, per Project, when you set it up. Self-contained memory keeps work focused and can reduce the chance of made-up answers (because the assistant draws from your contained files and history instead of guessing from elsewhere) — but it does not eliminate mistakes. Always verify anything that matters.
Project vs normal chat vs custom GPT
These three get mixed up. A custom GPT (see GPTs: Custom Assistants) is built for one fixed, repeatable task. A Project is a living workspace you keep customizing for an evolving body of work. A normal chat is a quick one-off.
| Use this | When | Keeps context? |
|---|---|---|
| Normal chat | A quick, one-off question or task you won't return to | No (just that chat) |
| Project | An ongoing topic you revisit — files + instructions + history that grow over time | Yes — files, instructions, project-scoped memory |
| Custom GPT | One fixed, repeatable job you (or others) run the same way each time | Fixed setup, not a growing history |
Three real example workflows
Upload a brand guideline, a voice doc, and a press kit. Set instructions for tone and format. Then in one Project, layer prompts: build a launch plan, write an email sequence in your brand voice, and draft social posts — all reusing the same context.
Set instructions like "act as my research assistant, cite sources, ask a clarifying question." Run a search or Deep Research, trim the best output into a clean document, and add it to the Project's files as a reusable knowledge base.
Recurring-work workspace: upload your own data (say, an analytics export), then keep asking follow-ups without re-uploading. Because the conversation stays in the Project's memory, it gets more useful the more you use it — for example, find the top themes in past titles, then draft a 30-day content calendar in a table you can share. (For longer, web-backed reports, Deep Research and the research tools are covered in ChatGPT Agent, Deep Research.)
A "help me set up this Project" starter prompt — paste this into a new chat inside the Project:
I'm setting up a ChatGPT Project for [topic / goal — e.g. managing my YouTube channel].
Help me set it up. Specifically:
1. Suggest a short, clear name for the project.
2. Write project instructions I can paste in — include a goal, an output format, a tone, and a rule to ask me clarifying questions when needed.
3. List the 3-5 files or documents I should add as the knowledge base, and what each should contain.
4. Suggest whether I should keep memory "default" or "project-only" for this topic, and why.
Ask me any clarifying questions first if you need them.Sharing with a team
Individual plans (Free and Plus) can't share a Project, and there's no public-link sharing for Projects the way there is for custom GPTs. Real-time, multi-person shared Projects — where teammates work from the same files, instructions, and history and see updates live — are a Business / Enterprise capability. For how the company plan, shared workspaces, and company-wide governance work, see Teams.
Checklist
Apps, Connectors, and Permissions
OpenAI renamed connectors to Apps in ChatGPT. Apps can provide interactive experiences, search connected data, support deep research, sync content, and sometimes take write actions.
Permission levels
| Permission | Meaning | Beginner default |
|---|---|---|
| Always ask | Ask before any app action, including reading. | Best for sensitive data. |
| Any changes | Read automatically, ask before changing anything. | Good default for most users. |
| Important actions | Ask before consequential actions. | OpenAI default. |
| Never ask | Read and act without asking. | Avoid until you fully trust the workflow. |
Connected app prompt
Before using any connected app, explain:
1. Which app you want to use
2. What data you need from it
3. Whether you will only read or also change something
4. What approval I will see
5. What could go wrong
Use the minimum access needed.Scheduled Tasks and Pulse
Scheduled tasks let ChatGPT run one-off reminders, recurring work, or monitoring tasks. As of June 17, 2026, Pulse is being sunset as proactive updates move into scheduled tasks.
Task types
| Task | Example | Review habit |
|---|---|---|
| Reminder | Remind me Friday to renew insurance. | Keep it simple. |
| Recurring briefing | Every morning, brief me on sales, calendar, and priorities. | Review sources and connected apps. |
| Monitoring | Tell me when a competitor changes pricing. | Define what counts as meaningful. |
| Finance/health check-in | Biweekly spending review or portfolio update. | Do not treat as professional advice. |
Task prompt
Create a scheduled task for me.
Purpose:
[what I want]
Schedule:
[date/time or recurring window]
Sources to check:
[web / connected apps / past chats / none]
Only notify me when:
[what counts as useful]
Before saving, show me the exact task name, schedule, apps used, and pause/delete instructions.Personal Finances in ChatGPT
OpenAI is rolling out Finances for U.S. Pro users on web and iOS. It can connect supported accounts through Plaid, show spending, bills, subscriptions, net worth, and investments, and answer questions grounded in financial context.
Good beginner uses
Find categories, recurring charges, and unusual changes.
Compare goals with actual spending.
Identify recurring payments to review.
Explore scenarios, then verify independently.
Finance prompt
Review my finances as an educational assistant, not as a financial adviser.
Please provide:
1. Spending patterns
2. Recurring charges to review
3. Possible budget categories
4. Questions I should ask myself
5. What you cannot conclude from this data
6. What I should verify with a qualified professionalGPTs: Custom Assistants
GPTs are custom versions of ChatGPT with instructions, knowledge files, capabilities, apps, and actions. Building/editing is web-only for paid users and managed workspace users with permission.
GPT builder map
| Field | Use it for | Best practice |
|---|---|---|
| Instructions | Rules, tone, workflow, examples. | Put behavior here, not in knowledge files. |
| Knowledge | Reference docs and handbooks. | Use clean text-forward files. |
| Capabilities | Search, images, canvas, data analysis, apps. | Enable only what helps. |
| Actions | Custom APIs. | Require privacy policy for public GPTs. |
GPT instructions prompt
Help me design a custom GPT.
Purpose:
[what it should do]
Users:
[who will use it]
Workflow:
[step-by-step process]
Knowledge files:
[what reference material it should use]
Capabilities needed:
[search / image / canvas / data analysis / apps / actions]
Return:
1. Name
2. Description
3. Conversation starters
4. Instructions
5. Test prompts
6. What not to includeChatGPT Agent, Deep Research, and Action Workflows
ChatGPT agent combines research and action. It can use a virtual computer, visual browser, text browser, terminal, direct API access, and connected apps to complete longer tasks, while asking permission before consequential actions.
Use agent mode when
| Job | Example | Control rule |
|---|---|---|
| Research + deliverable | Analyze competitors and create a slide deck. | Ask for progress summaries. |
| Authenticated workflow | Brief me using calendar and inbox context. | Use takeover securely; approve actions. |
| Spreadsheet work | Update a financial model or dashboard. | Verify formulas and assumptions. |
| Web tasks | Find options, compare, prepare next steps. | Confirm before purchases or bookings. |
Agent prompt
Use agent mode for this task, but keep me in control.
Goal:
[finished outcome]
Tools allowed:
[web / files / connected apps / spreadsheet / browser]
Do not:
[purchases, sends, deletes, external changes without approval]
Every 10 minutes or before any important action, give me:
1. Progress
2. Next step
3. Risk
4. Approval neededWorking as a Team (ChatGPT Business)
If you run a small company or a team, this section is for you. It explains what a company plan gives you that personal plans don't, how people actually work together day to day, and how to roll it out without creating a privacy mess. You can read this even if your team is still on free or personal accounts.
First: the name changed
The company plan used to be called ChatGPT Team. Around April 2026 it was renamed ChatGPT Business. It is the same product family, just a new name. Older videos and articles will still say "Team" or "the Teams plan" — when you see that, read it as "Business." Above Business sits Enterprise, aimed at larger organizations.
What a company gets that personal plans don't
The headline difference is about your data. On a ChatGPT Business workspace, your conversations are not used to train OpenAI's models. This is the contractual default for the whole workspace — there is no per-person toggle to remember, and no one can accidentally turn it off. On personal plans, training-on-your-data is controlled by a setting each user manages themselves (covered in Memory, Personalization, Privacy).
Beyond that, a company plan adds the controls an admin needs to run AI for a group of people:
A central admin console, single sign-on (SAML SSO), automatic user provisioning (SCIM), domain verification, role-based access, and the ability to enforce two-factor login.
Usage analytics across the workspace, plus compliance attestations like SOC 2 Type 2 and ISO 27001 that your security or legal team may ask about.
Members also get the full feature set — GPTs, Projects, apps and connectors, Agent and Deep Research — plus shared workspace GPTs at no extra cost, generous GPT-5.5 usage, and developer tools like Codex (as of 2026).
How a team actually collaborates day to day
"Collaboration" here is less about chatting in the same window and more about sharing the things that hold your context so everyone works from the same setup.
| What you share | What it does for the team |
|---|---|
| Shared / workspace GPTs | Build a custom assistant once (e.g. a brand-voice writer or a proposal helper) and publish it to the whole workspace, so everyone uses the same instructions instead of reinventing them. |
| Shared Projects (multi-person) | On Business and Enterprise, several people can work inside the same Project — same files, same project instructions, same chat history — and see each other's updates. On free and personal plans you cannot share Projects, and there is no public link for them. |
| Company Knowledge connectors | Connect tools like Slack, Google Drive, SharePoint, and GitHub so ChatGPT can answer from your company's own documents and messages, with the workspace's permissions respected. |
| A shared prompt / playbook culture | Keep your best prompts, GPT specs, and "how we ask ChatGPT" guidelines in one place the team can copy. The tool is only as good as the prompt, so a shared playbook raises everyone's results. |
GPT is for a fixed, repeatable job; a shared Project is for an evolving body of work. See the comparison table in Projects for the full GPT vs Project vs chat breakdown.Which plan for a company?
Prices and exact limits change, so treat the figures below as a snapshot and confirm current details on the official ChatGPT pricing page before you commit.
| Plan | Best for | Key point (as of 2026) |
|---|---|---|
| Free / Plus | Individuals trying it out | Personal accounts. Data-use settings are managed per person; you cannot share Projects with teammates. |
| Business | Most small and mid-size companies | Multi-seat (minimum 2 seats). Workspace data not used for training by default, admin console, SSO/SCIM, shared GPTs, shared Projects, Company Knowledge connectors, usage analytics. |
| Enterprise | Larger organizations with strict needs | Adds larger context, more admin and security controls, and custom terms. No public per-seat price — you talk to sales. |
Rollout best practices
A calm rollout beats a rushed one. A simple order that works:
| Step | Do this |
|---|---|
| 1. Start small | Begin with 2–5 people who have real, frequent use cases. Learn before you scale. |
| 2. Set the data rules | Decide which connectors are allowed, who can publish workspace GPTs, and what data should never go in. This is governance — see Safety, Security, and Team Governance. |
| 3. Seed shared assets | Publish 1–2 useful workspace GPTs and create a shared Project for a real team task so people have something to copy on day one. |
| 4. Write a one-page playbook | A short guide: how we prompt, what tools are approved, what to verify, what not to paste in. |
| 5. Review monthly | Check usage analytics, retire connectors and GPTs no one uses, and update the playbook. |
Steps 2 and 5 connect directly to Safety, Security, and Team Governance, which has the full governance checklist and audit prompt. This section is about the plan and collaboration; that section is about the rules.
Copy-paste prompts
Use the first to draft a rollout plan, and the second to spec a workspace GPT you want to share. Fill in the brackets.
Act as a practical change-management advisor for a non-technical
team lead. We are rolling out ChatGPT Business to a company of
[number] people, mostly in [departments]. Our top use cases are
[use case 1], [use case 2], [use case 3].
Give me a one-page rollout plan with:
1. A pilot group and who should be in it.
2. Data rules to set before we start (which connectors to allow,
who can publish shared GPTs, what data must never be entered).
3. 2-3 shared workspace GPTs worth building first, and why.
4. A short staff playbook outline (how to prompt, what to verify).
5. A 30-, 60-, 90-day timeline.
Keep it plain and skimmable. Flag anything I should have our
security or legal contact review.Help me write a spec for a shared workspace GPT my whole team
will use. The GPT's job is: [one-sentence purpose].
Audience: [who on the team will use it].
Produce:
- A clear name and one-line description.
- The "Instructions" text (role, goal, tone, format, what to
always do, what to never do, and to ask a clarifying question
if my request is unclear).
- What reference files or knowledge it should hold (and what it
should NOT contain, e.g. secrets or customer data).
- 3 example prompts a teammate could use on day one.
Keep the instructions tight and unambiguous so results are
consistent across the team.Quick check
Safety, Security, and Team Governance
As ChatGPT gains apps, tasks, memory, finance, GPTs, and agent mode, the power move is governance: clear rules for what ChatGPT may read, remember, schedule, connect, and change.
Governance checklist
| Area | Rule | Owner |
|---|---|---|
| Apps | Default to ask before changes. | User or workspace admin. |
| Memory | Review summary monthly. | User. |
| Tasks | Review active tasks weekly. | User or team lead. |
| GPTs | Test before sharing; avoid secrets in knowledge. | Builder. |
| Finance/health | Educational support only; professional review required. | User. |
| Agent mode | Approval before real-world consequences. | User. |
Audit prompt
Audit my ChatGPT setup for safety and usefulness.
Review:
1. Memory habits
2. Apps/connectors and permissions
3. Scheduled tasks
4. GPTs I built or use
5. Library/storage cleanup
6. Agent mode risks
7. Finance or sensitive workflows
Return a simple action plan for a non-technical user.Power Tips & Hidden Features
Once you know the basics, a handful of small habits make ChatGPT noticeably faster and more useful. These are the moves regular users tend to discover late. None of them require a paid plan unless noted, and we have stripped out anything that was hype or that is not reliably true as of 2026.
A few of these touch features covered in depth elsewhere, so we just point you there instead of repeating them.
Everyday speed
Small things that save you the most clicks day to day.
| Tip | Why it helps |
|---|---|
| Talk to it like a text message, not a search engine | Plain, natural sentences ("I have 2 hours to make dinner for my family, something simple but a little impressive?") usually work as well as elaborate "prompt engineering." Lower the effort, keep the quality. |
| Ask a direct question, not a vague one | "What are the five key steps to bake a sourdough loaf at home?" beats "tell me about baking." Naming the outcome you want is the single biggest quality lever. |
| Use voice or dictation for long inputs | Speaking a messy thought is faster than typing it, and ChatGPT cleans up the grammar. Good for brain-dumps and on-the-go notes. |
| Edit a message you already sent | Instead of re-explaining, hover the message and edit it. ChatGPT re-answers from your corrected version, so you are not fighting the old context. |
| Ask for the exact output format up front | Add "...show it as a table" or "...as a bulleted list" and you skip a round-trip. You can also say "now convert that to JSON" afterward. |
Better outputs
Ways to get a stronger answer without writing a longer prompt.
| Tip | Why it helps |
|---|---|
| Don't accept the first draft | ChatGPT is good at revising its own work. "Tighten this, make it more energetic, under 100 words" usually beats trying to get it perfect in one shot. |
| Ask it to ask you questions first | "Before you answer, ask me five questions that would help you do this well." It surfaces the details you forgot to mention, then gives a far better answer. |
| Have it state its assumptions | "First list the assumptions you are making, then I'll correct them." You catch wrong guesses before they shape the whole answer. |
| Give a good example and a bad example | Showing the style you want and the style to avoid removes the guesswork far better than describing it in words. |
| Break big tasks into stages | Tell it all the steps up front, but ask it to do "stage one only," then approve before it continues. Long projects come out cleaner this way. |
Turn on Thinking for math or logic | For multi-step reasoning, pick GPT-5.5 Thinking (paid tiers). For simple questions the default Instant is fine — and on any plan it can auto-escalate to deeper reasoning on hard tasks. (Model picker basics: Choosing a model.) |
Before you answer, ask me 5 questions that would help you
give me the best possible result. Wait for my answers,
then continue.You are my copy coach. Here is a good example and a bad
example of what I want.
GOOD: a clear, specific hook with real stakes.
BAD: vague, "this changes everything" filler.
Now write 5 options that match GOOD and avoid BAD.Lesser-known features
Worth knowing they exist, even if you only reach for them occasionally.
| Tip | Why it helps |
|---|---|
| Branch a conversation | On a reply you like, open the menu and branch it into a new chat. You keep the original as a save point and explore a different direction without losing it. |
| Regenerate when an answer misses | If a reply is off, regenerate it rather than arguing with it. You often get a cleaner result from a fresh attempt than from a long correction. |
| Read text out of a screenshot | Paste an image and ask it to "read the text in this image." Handy for grabbing a prompt or quote shown on screen that you can't copy directly. |
| Ask it to write your prompt | "What's a good prompt to ask you for X?" It will hand you a stronger prompt than most people write, which you can paste right back. |
| Ask it to teach, then quiz you | "Explain [topic] like I'm new to it, then give me a 5-question quiz with an answer key." Turns a single answer into a quick learning loop. |
Make a marketing plan for a small coffee shop. But first,
list the assumptions you're making about its budget,
location, and customers so I can correct anything off.Your turn
Official Sources and Research Notes
This Bible is based on OpenAI release notes and help documentation checked on June 19, 2026.
| Source | What it supports |
|---|---|
| ChatGPT release notes | June 2026 updates: tasks, Pulse sunset, memory, model picker, Library, Finances, GPT-5.5 Instant, charts, writing blocks, jobs, images, and more. |
| Apps in ChatGPT | Connectors renamed to Apps, app capabilities, permissions, app directory, MCP/custom apps, plan capabilities, admin controls. |
| Scheduled Tasks in ChatGPT | Dedicated Scheduled page, task creation, monitoring tasks, limits, app usage, health and finance task examples. |
| Creating and editing GPTs | GPT builder, instructions, knowledge, capabilities, apps, actions, testing, version history. |
| Introducing ChatGPT agent | Agent mode, virtual computer, browsers, terminal, connectors/apps, permissions, supervision, availability, and risks. |