ChatGPT secret commands: do /x10think and /human work?

Short answer

Most "secret commands" that go viral, like /human, /x10think, /ELI5 and /TLDR, are not built-in. The AI simply reads them as words and guesses what you mean. They can work, but for the same reason that "explain it simply" works: it is a plain request, not a command. Real built-in commands, like /compact and /clear in Claude Code, are handled by the program itself. If you need a reliable result, a full sentence beats a shortcut.

First, two different things

A built-in command is something the program recognizes and runs. In Claude Code, typing /clear empties the conversation. The program does it, whatever the AI thinks.

A prompt shortcut is different. The program does not interpret it. The characters are passed to the AI as part of your message, and the AI guesses what you want. So the same /human can work in one chat, behave differently in another, or be ignored.

Table 1 lists things confirmed in official documentation or help pages. Table 2 lists viral shortcuts that are not confirmed as features. This page was checked on 2026-11-05, and product features change often.

Table 1: What is real

We could not confirm from official help whether typing / in the ChatGPT message box opens a command list. ChatGPT does let you pick features (add files, create images, web search and so on) from the + menu. The menu layout can differ by app, account and date, so the best check is to open it yourself.

Some posts list /resume, /fork or /compact as ChatGPT commands. Those often come from OpenAI's coding tool, Codex CLI, which is a different program from regular ChatGPT chat.

Gemini's @ feature has changed over time. According to a report from October 2025, public services such as Maps and YouTube no longer need @. You can just ask in a sentence. Google's current help page tells you to enter @ and select the service, or name the service in your prompt, for Google Workspace apps like Gmail. Whether you can use it also depends on your account type and settings.

Claude Code
Command or shortcut
/compact
What it does
Summarizes the conversation to free up context (how much the AI can keep in mind at once)
Verification
Confirmed in official docs
Claude Code
Command or shortcut
/clear
What it does
Starts a new conversation with empty context
Verification
Confirmed in official docs
Claude Code
Command or shortcut
/model
What it does
Switches the AI model
Verification
Confirmed in official docs
Claude Code
Command or shortcut
/effort
What it does
Sets how hard the AI thinks (low up to max)
Verification
Confirmed in official docs
Claude Code
Command or shortcut
/help, /resume, /rewind
What it does
Help, continue a past conversation, go back to an earlier point
Verification
Confirmed in official docs
Claude Code
Command or shortcut
ultrathink
What it does
A keyword that asks for deeper thinking on that one turn
Verification
Confirmed in official docs
Claude Code
Command or shortcut
think, think hard
What it does
Not keywords. Passed through as ordinary text
Verification
Confirmed in official docs
Claude (web and apps)
Command or shortcut
Skills
What it does
Not called with a slash. Applied automatically when relevant (you can also ask for one by name)
Verification
Confirmed in official help
Gemini
Command or shortcut
@ then pick a Google app
What it does
Point Gemini at Gmail, Drive and other Google services. Naming the app in your sentence also works
Verification
Confirmed in official help
ChatGPT
Command or shortcut
Temporary Chat
What it does
A separate chat that does not use memory or custom instructions
Verification
Confirmed in official help (earlier check)
ChatGPT
Command or shortcut
"Remember this" / "Forget this"
What it does
Ask in words to add or delete saved memory
Verification
Confirmed in official help (earlier check)
ChatGPT
Command or shortcut
Custom instructions
What it does
Applies the instructions you wrote to every chat
Verification
Confirmed in official help (earlier check)
ChatGPT
Command or shortcut
Ctrl+/ (Cmd+/ on Mac)
What it does
Opens the keyboard shortcut list
Verification
Several guides agree; the official page was not accessible, so not directly verified

Product	Command or shortcut	What it does	Verification
Claude Code	/compact	Summarizes the conversation to free up context (how much the AI can keep in mind at once)	Confirmed in official docs
Claude Code	/clear	Starts a new conversation with empty context	Confirmed in official docs
Claude Code	/model	Switches the AI model	Confirmed in official docs
Claude Code	/effort	Sets how hard the AI thinks (low up to max)	Confirmed in official docs
Claude Code	/help, /resume, /rewind	Help, continue a past conversation, go back to an earlier point	Confirmed in official docs
Claude Code	ultrathink	A keyword that asks for deeper thinking on that one turn	Confirmed in official docs
Claude Code	think, think hard	Not keywords. Passed through as ordinary text	Confirmed in official docs
Claude (web and apps)	Skills	Not called with a slash. Applied automatically when relevant (you can also ask for one by name)	Confirmed in official help
Gemini	@ then pick a Google app	Point Gemini at Gmail, Drive and other Google services. Naming the app in your sentence also works	Confirmed in official help
ChatGPT	Temporary Chat	A separate chat that does not use memory or custom instructions	Confirmed in official help (earlier check)
ChatGPT	"Remember this" / "Forget this"	Ask in words to add or delete saved memory	Confirmed in official help (earlier check)
ChatGPT	Custom instructions	Applies the instructions you wrote to every chat	Confirmed in official help (earlier check)
ChatGPT	Ctrl+/ (Cmd+/ on Mac)	Opens the keyboard shortcut list	Several guides agree; the official page was not accessible, so not directly verified

Table 2: Viral shortcuts

We could not find where these shortcuts first spread. They circulate in blog posts and newsletters as lists of "secret ChatGPT codes." The lists differ from each other, and many of them say these are community habits, not official features. This page only calls them "viral shortcuts."

That does not mean they never work. AI models saw phrases like TLDR during training, so they usually understand them. But there is no guarantee they work the same way every time, so for anything important, use the full sentence.

/human
Claimed effect
Sounds like a person wrote it
What it really is
Not a built-in feature. Often understood from the words
A sentence that does the same
Rewrite this so it sounds natural, like explaining to a friend. Cut stiff phrasing.
/x10think
Claimed effect
Thinks 10 times harder
What it really is
Not a built-in feature. Posts disagree on what it means (deeper reasoning, or "explain to a 10-year-old"). No official measurement of "10x" found
A sentence that does the same
Work through this step by step, check the opposite case, then answer.
/ELI5
Claimed effect
Explain so a 5-year-old gets it
What it really is
Internet slang for "Explain Like I'm 5". The AI understands the phrase, but it is not a command
A sentence that does the same
Explain this with no jargon, so anyone could follow it.
/TLDR
Claimed effect
Short summary
What it really is
Slang for "too long; didn't read". Not a command
A sentence that does the same
Summarize the key points in 3 lines.
/brief
Claimed effect
Short answer
What it really is
Not a built-in command
A sentence that does the same
Give the conclusion first, in 5 sentences or fewer.
/jargon
Claimed effect
Explain technical terms
What it really is
Not a built-in command
A sentence that does the same
Add a plain-English explanation in parentheses after any technical term.
/devil
Claimed effect
Argue the other side
What it really is
Not a built-in command (from "devil's advocate")
A sentence that does the same
Give me 3 weaknesses of this plan and the strongest objections.

Shortcut	Claimed effect	What it really is	A sentence that does the same
/human	Sounds like a person wrote it	Not a built-in feature. Often understood from the words	Rewrite this so it sounds natural, like explaining to a friend. Cut stiff phrasing.
/x10think	Thinks 10 times harder	Not a built-in feature. Posts disagree on what it means (deeper reasoning, or "explain to a 10-year-old"). No official measurement of "10x" found	Work through this step by step, check the opposite case, then answer.
/ELI5	Explain so a 5-year-old gets it	Internet slang for "Explain Like I'm 5". The AI understands the phrase, but it is not a command	Explain this with no jargon, so anyone could follow it.
/TLDR	Short summary	Slang for "too long; didn't read". Not a command	Summarize the key points in 3 lines.
/brief	Short answer	Not a built-in command	Give the conclusion first, in 5 sentences or fewer.
/jargon	Explain technical terms	Not a built-in command	Add a plain-English explanation in parentheses after any technical term.
/devil	Argue the other side	Not a built-in command (from "devil's advocate")	Give me 3 weaknesses of this plan and the strongest objections.

Sentences worth copying

Spelling out what you want works about the same in any AI. Copy these and add them to the end of your question.

Explain simply (instead of /ELI5)

Explain this without jargon, so someone hearing it for the first time can follow. Include one simple analogy.

Make it think carefully (instead of /x10think)

Answer carefully

Do not jump to a conclusion. First list the possible options, weigh the pros and cons of each, then recommend one. If you do not know something, say so.

In Claude Code, putting ultrathink in your prompt officially asks for deeper thinking on that turn. In regular chat apps, use the thinking or reasoning option in the model picker if there is one. That is more reliable than a made-up keyword.

Hear the other side (instead of /devil)

Find the weak spots

From the opposing point of view, point out the 3 weakest parts of my plan. Skip the praise, and add one line on how to fix each.

How to make your own shortcuts that work

If you like shortcuts, tell the AI once what they mean. This does not add a built-in command. It saves an instruction that says "when you see this, do that."

In ChatGPT, put it in or in a project's instructions. In Claude, use a project's instructions or a style. Menu names and locations can change with app updates. Once saved, you do not have to explain it in every new chat.

Paste into custom instructions or project instructions

When I use one of these shortcuts in a message, respond as described:
- /simple: explain with no jargon, so a beginner can follow
- /sum: summarize in 3 key lines
- /careful: weigh the pros and cons of each option, then recommend; say so if you do not know
- /oppose: point out 3 weaknesses in my view from the opposite side
If there is no shortcut, answer as usual.

You can name the shortcuts anything. Test once, and if it does not behave as you expect, make the instruction more specific.

If you want real commands

The tools with real command lists are coding tools that run in a terminal. Claude Code, for example, has dozens of commands that start with /, and /help shows the list. Unlike chat apps, these commands are handled by the program, not by the AI.

Related terms

Slash command (/): Typing / before a word so the program runs it. It only works if the program supports it.
Custom instructions: A ChatGPT setting that applies your tone, background and rules to every chat.
Context: How much conversation and material the AI can keep in mind and use at once.

FAQ

Does /x10think really make the AI think 10 times harder?

We found no source that confirms it. It is not a built-in feature, and posts describe it in different ways. If you want deeper thinking, write out the process you want, like "weigh the pros and cons of each option before answering."

Does /human get past AI detectors?

There is no evidence for that. /human is usually presented as a request for a more natural tone, which is separate from detection. If you use AI for school or work, check the rules on AI use first.

Does typing / in ChatGPT show a command list?

We could not confirm that from official help. It may vary by app, account and date, so try it yourself. Choosing features from the + menu is confirmed.

Can I use /compact in the Claude web app?

/compact is a Claude Code command, not a feature of regular Claude chat. In the web app or phone app, start a new chat or ask for a summary instead.

My own shortcut only works sometimes. Why?

Your shortcut is an instruction the AI interprets, so it is not 100% fixed. Add an example line to the instruction, or use the full sentence for important work.

Keep going

Claude Code command reference

Checked against sources on 2026-11-05. Sources: Claude Code Docs, Commands, Claude Code Docs, Model configuration (ultrathink, effort), Claude Help Center, Using skills in Claude, OpenAI Help Center, Memory FAQ, Google Gemini Help, Connected Apps, 9to5Google, Gemini @ mention change (2025-10). How we write