A prompt is a recipe, and the context you give the AI is the groceries. Those viral lists of secret prompts sell recipes to people with empty refrigerators.
A post went around Facebook this week with a siren emoji at the top. Ten “forbidden” ChatGPT prompts. It promised they weren’t illegal, just insanely powerful. Save this.
The comments ran the usual way, with a string of great-infos, thank-yous and nice-I’ll-try-thems. Somebody pasted the whole list back into the comments. Somebody else ran it through ChatGPT to clean it up and posted that version too. One person called it stupidity, in Romanian.
I’ve seen this post in many forms. The ten prompts that will change your life. The secret prompts the AI companies don’t want you to know. The one prompt that replaced my marketing team. They’re crap. The people who fall for them are trying to learn something real, so they deserve to know why.
What were the ten “forbidden” ChatGPT prompts?
For the record, the list went like this, trimmed of the emojis. Ask what you’re missing that could change the outcome. Start from your goal and map the fastest path. Be brutally honest about why your plan will fail. Find the actions that create most of the results. Name a completely different strategy if this one fails. Find the opportunities you’re overlooking. Find the simplest way to reach the goal.
Then it got grander. Name the trends to watch in your industry. Show how to create an unfair advantage. Name the single most important action to take right now.
None of these are forbidden. None are secret. They’re planning questions any consultant, coach or decent manager has been asking for decades. Putting quotes around them and calling them prompts doesn’t change what they are.
Do viral AI prompt lists work?
Some of the prompts do. I’d sort this list into three piles.
Two are useful. “Be brutally honest. Why will this fail, and what am I underestimating?” and “What am I missing that could completely change this outcome?” Both work against the AI’s built-in urge to agree with you. Chatbots are designed to please the person typing. I wrote about that in AI echo chambers. Any question that tells the machine to look for problems gets you further than one that invites it to cheer.
One commenter put his finger on why the “what am I missing” version works. Asking what you’re missing finds gaps. Asking what you got wrong mostly gets the machine to confirm what you already believe.
Three are fine. The 80/20 question, the different-strategy question and the simplest-way question are ordinary planning questions. You’d get much the same answer asking in plain English, without the arrow and the label.
The other five are mostly filler. “What trends should I pay attention to in my industry?” “How can I create an unfair advantage?” “What is the single most important action I should take right now?” Ask those with no detail and you get confident, generic lists that could apply to anyone in any business. The single most important action comes back as one plausible pick out of a dozen equally plausible ones. It sounds decisive, and nothing backs it up.
A prompt is a recipe. Context is the groceries. Most viral prompt lists sell recipes to people with empty refrigerators. – Richard LoweShare on X
Why is context more important than the prompt?
Because the AI can only work with what you give it.
Type “Be brutally honest, why will my book fail?” with nothing else, and you’ll get a list of reasons books fail. Weak marketing, crowded genre, no platform. It’s true of every book, so it tells you nothing about yours.
Now paste in your book description, your first chapter, your target reader and your launch plan, then ask the same question. The answer changes completely. It can point at the description that buries the hook, the chapter that opens with backstory, the launch plan that depends on a newsletter you don’t have yet. The prompt didn’t change. The groceries did.
When I use AI in my writing work, the material goes in first. For editing, I give the AI an expert role and one narrow job, and I have it work through a manuscript line by line. None of it would make a good Facebook post. All of it works better than any list with a siren emoji on top.
Almost any plain question works with good context. Even the best prompt returns generic answers without it.
What should writers ask AI about a manuscript?
The same rule applies, with the manuscript as the groceries.
Paste in a chapter and tell the AI what the book is and who it’s for. Then ask a narrow question. Where would a reader in this genre lose interest, and why? Which scenes repeat information the reader already has? Where does the dialogue explain what the scene already showed? Which paragraphs could go without the reader missing anything?
For a book description, paste the description, the genre and two or three books like yours, then ask what a browsing reader would think the book is about after the first two lines. If the answer doesn’t match your book, the description needs work. For a nonfiction outline, paste the outline and describe the reader’s problem, then ask which chapters a reader with that problem could skip, and what question they’d still have at the end. That’s “what am I missing” with something real to chew on.
None of those are secret. They’re the questions a good editor asks. The difference is that the AI answers in seconds, and you still have to decide whether it’s right.
Why do “secret prompt” posts keep going viral?
Because they’re built to be saved and shared. Nobody designs them to be used.
“Save this” gives the game away. The post asks for the action that pleases the algorithm. A save signals value to the platform, so the post spreads to more people, so the account grows. Whether anybody runs the prompts doesn’t matter to the poster. Many of these accounts sell a course, a prompt pack or a paid group, and the free list is the bait.
The framing does the rest. “Forbidden” and “secret” make an ordinary question feel like insider knowledge. “Insanely powerful” promises a shortcut. People who are new to AI, or nervous about it, want to believe there’s a magic phrase that makes it work, because then they don’t have to learn how to use it. The comment threads show the effect. Plenty of thank-yous, plenty of “I’ll try these,” and almost nobody saying what happened when they did.
How do you write AI prompts that work?
Start with the material. The magic words can wait. Paste in the draft, the data, the plan or the document you want help with.
Say who you are and who it’s for. “I’m a first-time memoir writer, and this is for my grandchildren” gets a different answer from “I’m a business coach writing a book to attract clients,” and it should.
Say exactly what you want. Not “improve this,” but “find the three places where the pacing drags and tell me why.” The more specific the job, the more useful the answer. Ask it to argue with you. The two good prompts from the list belong here. Once the context is in, ask what you’re missing and why it will fail.
Then judge the answer yourself. The AI doesn’t know your readers, your market or your life the way you do. I covered more examples in AI writing prompts that work, and for fiction in prompt engineering for fiction writers.
Should you save prompt lists you see online?
Save the questions if you like them. Just don’t expect them to do the work.
A good question is a good question whether you ask it of a chatbot, a coach or yourself at two in the morning. “What am I missing?” was useful long before ChatGPT existed. But nobody’s life ever changed because they copied ten questions into a note on their phone. The change comes from doing the work the questions point at.
The AI and writing hub has more on using AI well without handing it your judgment. If you want help using AI on your own book, with your own material and your own voice, my book coaching covers that. Skip the recipe cards with the siren emojis. Stock the refrigerator, then ask a plain question.
