Most people use ChatGPT one way: type a question, get an answer. It has a dozen other ways to behave sitting behind a plus button, a slider and a settings page. Here's every one, where it lives, and the exact words to build your own.
The plan · 15 MODES, 3 MOVES
- 1Learn the five modes from the video: Study, Incognito, Genius, Health and custom modes.
- 2Unlock the ten more hiding in the app: Work, Deep Research, Voice, Projects, Scheduled and more.
- 3Paste the ten tested custom modes into your settings so they trigger every time.
Before you start
15 modes
Every one with the path to find it and the plan it needs.
Mostly free
Almost all of them work on the free plan, some with tighter limits.
1 sentence
That's all it takes to build your own mode.
A mode is any switch that changes how ChatGPT behaves before you type a word. Some are real buttons OpenAI built (Study, Temporary chat, the Thinking slider, Health). Some are settings pages most people have never opened. And some you make yourself with one sentence and a saved memory.
All fifteen at a glance
| Mode | Where | Plan |
|---|---|---|
| 1. Study | Plus button, then Study (or type @study) | All plans |
| 2. Incognito (Temporary chat) | Temporary button, top right of a new chat | All plans |
| 3. Genius (Thinking slider) | Model menu, Thinking slider; Free and Go use Think | All; deepest levels on Pro |
| 4. Health | Health in the sidebar | Free to Pro, US, 18+ |
| 5. Custom modes | One sentence plus "save this to memory" | All plans with memory on |
| 6. Work | Chat / Work switch at the top | Paid plans; limited on Free and Go (desktop) |
| 7. Deep Research | Type /Deepresearch, or + then Deep research | All plans, bigger limits on paid |
| 8. Voice | Voice button in the message box | All plans |
| 9. Projects (private memory) | Sidebar, Projects, then project settings | All plans |
| 10. Scheduled tasks | Sidebar, Scheduled | All plans (Free gets 3 active) |
| 11. Personality | Settings, Personalization, Base style and tone | All plans |
| 12. Lockdown | Settings, Security, Advanced security | Eligible accounts |
| 13. Quiz | Say "quiz me on..." (or inside Study) | All plans |
| 14. Sites | Work mode, ask to publish | Plus and up (not Free or Go) |
| 15. Shopping | Ask to compare or find products | All plans |
The Five From The Video
Study mode: it teaches instead of answering
Click the + in the message box and choose Study (or just type @study). A Study chip appears in your message box. From then on ChatGPT asks you questions, explains one step at a time, checks you understood, and builds practice questions from whatever you upload: notes, PDFs, photos of a textbook. To turn it off, delete the Study chip. Works on every plan and every model.
Incognito mode: nothing gets saved
OpenAI calls it Temporary chat. Open a new chat and click the pill-shaped Temporary button in the top right (a toggle on mobile). Nothing in that chat goes into your history, no new memories get created, and it isn't used for training. You can pick Personalized (it still uses what it already knows about you) or Unpersonalized (a totally blank slate). One honest detail: OpenAI keeps temporary chats for up to 30 days for safety review, then they're gone. Off the record from your account's point of view, not a vanishing act.
Genius mode: turn the thinking up
The model menu shows GPT-6 with a Thinking slider: Instant, Medium, High, Extra High and Pro. Plus gets Medium and High, and Extra High and Pro need a Pro plan. On Free and Go, tap Think. The higher you go, the longer it reasons before answering and the better it is at hard, multi-step problems. It's also slower and eats more of your limit, so save it for the questions that matter.
Health mode: it can see your real body data
Open Health from the sidebar (check under More if you don't see it). Inside, open the menu and choose Apple Health, then pick which categories to share: sleep, steps, heart rate, workouts. Apple Health connects from the iPhone app only. You can also link supported US medical portals. Health is available to Free, Go, Plus and Pro users in the US, 18 and over, on web and iOS. Health chats are never used for training, and everything deletes within 30 days if you disconnect.
Maverick's tip
Health is built to support your care, not replace it. It will build a sleep plan around your actual numbers; it will not diagnose you. Treat it like a smart friend reading your Apple Watch, not a doctor.
Build your own modes
Start a new chat and type one sentence: "Whenever I type BRUTAL MODE, ..." then "Save this to memory." ChatGPT confirms, and the trigger word works in future chats. The Build Your Own Modes section has ten modes and the one thing the video leaves out: where the rule should really live.
Ten More Modes Hiding In The App
Work mode: it builds instead of chats
The Chat / Work switch at the top of ChatGPT. Work is the mode for multi-step jobs: it researches, writes files, uses a browser, connects to Gmail and Slack, and publishes sites. If you've ever asked for a spreadsheet and gotten a paragraph, you were in Chat. Full Work is on paid plans; Free and Go get a limited version in the desktop app. My ChatGPT Work guide goes deep.
Deep Research mode: a report, not an answer
Type /Deepresearch in front of your question, or pick Deep research from the + menu. It shows you a research plan, goes away for several minutes, reads dozens of sources, and comes back with a cited report. Use it for anything you'd otherwise spend an afternoon Googling. Every plan gets it, with bigger limits on paid plans.
Voice mode: talk, interrupt, keep going
The voice button in the message box. In Settings, then Voice, pick Live, the newest option: it listens and speaks at the same time, so you can cut it off mid-sentence like a person. Use Advanced if you want video or screen sharing. Background conversations (also in Settings, then Voice) keep it talking while you use other apps or lock your phone. Free gets limited access to a lighter Live model; paid plans get the full one with more hours.
Private project mode: memory that stays in one box
Create a Project from the sidebar, open its settings (the three dots) and switch Memory to Project-only. Now everything ChatGPT learns in that project stays there and never leaks into your other chats. Perfect for health, money, a client, or anything you don't want coloring every future answer. All plans.
Autopilot mode: scheduled tasks
Sidebar, Scheduled. Tell ChatGPT to do something every morning or every Friday and it runs without you: a news brief, a weekly report, a reminder with research attached. You can pause, edit or delete each one from the Scheduled page. Free and Go get 3 active tasks, paid plans get more, and Plus and Pro can also trigger tasks from events like a new email.
Personality mode: pick a default voice
Settings, then Personalization, then Base style and tone. Seven options: Default, Professional, Friendly, Candid, Quirky, Efficient and Cynical. Candid is the closest built-in thing to brutal mode. Efficient strips the fluff. This applies to every chat, so it's the mode you set once and forget.
Lockdown mode: cut the live internet
Settings, then Security, then under Advanced security turn on Lockdown Mode (if your account has it). Web browsing drops to cached pages only, and Deep Research and agent actions switch off, so nothing you're working on gets sent out to a live website. Use it when you're working with something sensitive. You can turn it off for a single chat when you need the web.
Quiz mode: test yourself on anything
Just say "Quiz me on [topic]" or "quiz me on this" with a file attached. ChatGPT builds an interactive quiz, scores you, and explains what you missed. It's built into Study mode too, but works on its own in any chat. All plans.
Publisher mode: ChatGPT Sites
In Work, ask ChatGPT to build something and publish it, and it hosts the result at a live URL: a landing page, a portfolio, a tool, a game. You can change the URL from the site's settings. Plus, Pro and business plans (not Free or Go).
Shopping mode: it compares for you
Ask "compare the best [thing] under $[price]" and ChatGPT returns products side by side with prices, specs and where to buy. Say what you care about (battery life, weight, no subscriptions) and it re-ranks. All plans.
Maverick's tip
Finances (Explore > Finances, US, every personal plan; the credit score is on Plus and Pro): a dashboard of spending, bills, subscriptions and net worth. Codex (the ChatGPT / Codex switch in the desktop app): the mode for building real software. Images (More > Images): start from a template instead of a blank prompt. Library (sidebar): the files and images you've made and uploaded, called Space on Pro and Business accounts.
Maverick's tip
Three of these stack. Work mode plus Deep Research plus a scheduled task is a research assistant that hands you a cited report every Monday morning without being asked.
Build Your Own Modes
The video's brutal mode is real and it works. I tested ten custom modes in fresh chats and tightened the wording until each triggered every time. Two rules came out of it.
Rule one: put the trigger word first, in caps, with a colon. "BRUTAL MODE: here's my idea" triggers reliably. Burying the word mid-sentence doesn't. Rule two: numbers survive, adjectives don't. "Under 150 words" and "exactly five bullets" get followed. "Keep it short" gets ignored by week two.
Mode 1: BRUTAL MODE
Whenever I type BRUTAL MODE: critique the work directly. Lead with the biggest flaw. No compliments, no reassurance, no closing encouragement. Name what I'm avoiding. Respectful about me, blunt about the work. Save this to memory.
Mode 2: EXPLAIN LIKE I'M 12
Whenever I type EXPLAIN LIKE I'M 12: plain words, exactly one analogy, no technical terms, under 150 words, and end with one question to check I got it. Save this to memory.
Mode 3: DECISION MODE
Whenever I type DECISION MODE: do not list options. Ask me up to three clarifying questions in one message and stop. After I answer, pick one answer and defend it in exactly five sentences. Save this to memory.
Mode 4: RECAP
Whenever I type RECAP: summarize this conversation only, as exactly five bullets, then one line starting with "Next action:". No preamble. Save this to memory.
Mode 5: DEVIL'S ADVOCATE
Whenever I type DEVIL'S ADVOCATE: argue the strongest case against my last statement in 3 to 5 numbered points, then a section titled "Actually worry about:" naming which points matter most. Save this to memory.
Mode 6: COACH MODE
Whenever I type COACH MODE: do not give advice or solutions. Ask exactly one question per message and wait. Maximum five questions. After the fifth answer, reflect back what I said in one paragraph, no recommendations. Save this to memory.
Mode 7: STEELMAN
Whenever I type STEELMAN: restate the position I'm arguing against in its strongest, most charitable form, as its best advocate would, under 150 words, then stop. Save this to memory.
Mode 8: SHRINK IT
Whenever I type SHRINK IT: rewrite whatever I paste at half the length, keeping every fact and the original tone. Output only the rewrite, no commentary. Save this to memory.
Mode 9: FIVE WHYS
Whenever I type FIVE WHYS: take the problem I just described and ask "why" five times in sequence, one per line, then name the likely root cause in one sentence. Save this to memory.
Mode 10: PLAIN ENGLISH
Whenever I type PLAIN ENGLISH: translate the jargon or contract clause I paste into everyday language in one short paragraph, then one line: "What this means for you:". Save this to memory.
Maverick's tip
Saved memories are short paraphrased facts, so the numbers get dropped. Put the modes in Settings, then Personalization, then Custom instructions, one line each (5,000 characters fits all ten). They're injected word for word into every chat. To check one, type the trigger alone and ask "what rule are you applying?" Want more trigger words? My ChatGPT secret codes guide has a whole list you can turn into modes the same way.
Mistakes That Make Modes Stop Working
| Mistake | Fix |
|---|---|
| Custom modes stored only in memory | Put the rules in custom instructions. Memory compresses; instructions don't. |
| Trigger word buried mid-sentence | Trigger first, caps, colon. "BRUTAL MODE: ..." |
| Vague rules | Numbers: "under 150 words," "exactly five bullets," "one question per message." |
| Two jobs in one turn | "Ask three questions, then decide" can't happen in one reply. Say "and stop." |
| Expecting Study mode to persist | It's a chip in the message box. New chat, add it again, or use @study. |
| Thinking it's private because of Temporary chat | Nothing saved to your account, but OpenAI keeps it 30 days for safety. Use Lockdown to cut live web access. |
| Running genius mode on everything | Slower and burns your limit. Instant for quick stuff, high levels for the hard question. |
| Trying Apple Health on the web | The connection is iPhone-app only. Connect there once, then use it anywhere. |
The one-page map
| Want to... | Use | Path |
|---|---|---|
| Learn something properly | Study, Quiz | Plus button, Study; or "quiz me on..." |
| Ask something off the record | Temporary chat, Lockdown | Temporary (top right); Settings, Security |
| Get the best possible answer | Thinking slider | Model menu, slide Thinking up |
| Plan around your body | Health | Health in the sidebar; Apple Health from iPhone |
| Get things built, not described | Work, Sites, Scheduled | Chat/Work switch; ask to publish; Sidebar, Scheduled |
| Change how it talks to you | Personality, Custom modes | Settings, Personalization |
| Keep a topic sealed off | Project-only memory | Project settings, Memory |
Maverick's tip
Open Settings, Personalization. Set your Base style. Paste three of the custom modes into Custom instructions. Then start a new chat, type BRUTAL MODE and your current project. You'll know in thirty seconds whether it's worth the other twelve.
Same ChatGPT. Fifteen different tools. Pick the right one before you type.
Published · Checked