The short version: A useful prompt book is not a list of clever sentences, it is a small filing system built around the actual jobs you do every week, tagged by outcome rather than topic, and rewritten in your own voice so it works without you standing over it. Copying someone else’s “50 best ChatGPT prompts” PDF will not save you time. Building your own, from prompts you have already tested on your own work, will.
Why I started keeping mine in the first place
I rebuilt my consulting business almost entirely with ChatGPT running alongside me for the last two years, and by month four I had a folder on my desktop called “prompts” with 340 files in it. Screenshots, half-finished Notion pages, a Google Doc called “prompts NEW NEW final”. It was a mess, and worse, it was useless, because I could never find the prompt I needed at the moment I needed it. I’d spend ten minutes hunting for the LinkedIn hook prompt that worked in March, give up, and rewrite a worse version from scratch.
That’s the bit nobody tells you about prompt collecting: hoarding prompts feels productive and isn’t. It’s the digital equivalent of buying a gym membership. The value only shows up when you sit down, structure the thing, and use it under pressure, on a Tuesday, with a client waiting on an email.
The contrarian bit: most prompts don’t survive being copied
Here’s what almost every “prompt bible” article skips over. A prompt that worked brilliantly for you last week rarely works for someone else, and often doesn’t even work for you again next month. Why? Because a good prompt is really three things bundled together: the instruction, the context you fed it (your tone, your niche, your past examples), and the model version you ran it on at the time. Strip out the context and the instruction alone is often just a nice-sounding sentence with nothing behind it.
This is why those free “500 ChatGPT prompts” lead magnets are mostly filler. They’re generic on purpose, because generic is shareable. Your book should be the opposite: narrow, specific, and slightly boring to anyone but you. A prompt like “Write a follow-up email for a warm lead who went quiet after a proposal, referencing the exact objection they raised, in my direct-but-friendly tone, under 120 words” is worth ten times more to you than “Write a persuasive follow-up email.” The second one is a title. The first one is a tool.
Step one: pull everything into one place before you organise anything
Don’t try to sort as you go, you’ll stall out. Spend one sitting, 45 to 60 minutes, and just dump. Go through:
- Your ChatGPT chat history (search for keywords like “write”, “draft”, “rewrite”, “summarise”)
- Any notes app entries, sticky notes, or half-remembered prompts in Slack or WhatsApp to yourself
- Screenshots you’ve saved but never actioned
Paste every single one into a single running document, no formatting, no judgement, just capture. When I did this for the first time I ended up with 118 prompts. After I removed duplicates and dead ends, I was down to 41 that I’d used more than once. That’s a normal ratio, expect to keep roughly a third of what you find.
Step two: sort by job, not by topic
This is the part that decides whether your book gets used in six months or joins the folder of shame. Don’t organise by “marketing prompts”, “writing prompts”, “business prompts”. Organise by the actual task you sit down to do. My categories ended up as:
- Client emails (chasing, onboarding, awkward conversations)
- First drafts (blog intros, LinkedIn posts, proposal openers)
- Editing and tightening (cutting waffle, checking tone, shortening for length)
- Research and summarising (long PDFs, call transcripts, competitor pages)
- Thinking out loud (when I want ChatGPT to challenge an idea, not just execute it)
Five categories, not fifteen. If you can’t remember your own categories, they’re too fine-grained. A book you have to search is a book that will sit unopened.
Step three: rewrite each prompt so it doesn’t need you to remember anything
Every prompt in your book should include the three missing pieces most people leave out: the format you want back, the tone or voice reference, and a placeholder for the thing that changes. Here’s a real one from my “Client emails” section, exactly as it sits in my book:
“Write a short follow-up email to [client name] about [specific project or proposal], referencing their exact concern which was [insert their words], in a direct but warm tone, no corporate padding, under 100 words, one clear next step at the end.”
Notice the brackets. That’s the whole trick. A prompt with brackets is a template you can fill in thirty seconds. A prompt without brackets is a story you have to reconstruct from memory every time.
Step four: pick the actual format for the book
You don’t need fancy software. Three options work, in order of how much upkeep they need:
- A single Google Doc with headers, one heading per category, prompts listed underneath with a one-line note on when to use each. Fastest to set up, fine if you’re one person working alone.
- Notion or a similar wiki tool, with tags so you can filter by category and by “how often I use this” (weekly, monthly, rarely but valuable). Better if you share prompts with a team or virtual assistant.
- A physical notebook, which sounds old-fashioned but works surprisingly well if you’re someone who thinks better on paper and only wants your 15 to 20 absolute favourites, not a growing archive.
I use Notion now, with three columns: Prompt, Category, Last used. That last column matters more than people expect, because it forces an honest audit. Anything untouched for four months gets deleted or merged. A book that only grows and never gets edited turns back into the messy folder you started with.
Step five: add the “why” next to each prompt, briefly
One line per prompt explaining what it’s for and what a good output looks like. Not a paragraph, one sentence. Mine reads something like: “Use this when a lead has gone cold after a proposal call, not for cold outreach. Good output sounds like me on a Tuesday morning, not a sales script.” Six months from now you won’t remember the nuance, and this line is what stops you misusing a prompt in the wrong context.
How often to update it
Set a recurring 20-minute slot once a month, phone reminder, whatever works. Review what you used, add anything new that worked twice, delete anything you didn’t touch. I do mine on the first Monday of the month, straight after I check invoices, so it’s tied to a habit I already have rather than a fresh thing I need to remember. A book that’s reviewed monthly stays lean. A book that’s never reviewed becomes the 340-file folder again within a year, guaranteed.
What to leave out
Skip anything that only worked because of a specific mood, a specific client, or a one-off event. Skip prompts you copied from someone else’s newsletter and never tested, they clutter the book with false confidence. And skip the temptation to make it “shareable” or pretty before it’s useful. A prompt book with ugly formatting that you use every day beats a beautifully designed one you built instead of doing your actual work. I’ve watched people spend a whole weekend designing a Notion template for their prompts and then never fill it in, because the designing felt like progress and wasn’t.
If you want to go further: turning it into something others can use
Some of my clients have taken their internal prompt book and turned it into a lead magnet or a paid mini-guide, and it can work, but only once you’ve used every prompt in it for real, for months, on real client work. A prompt book built to sell before it’s been tested reads exactly like the generic ones I warned you about earlier, and readers can tell. If you’re at the point where you want to build a proper prompt and workflow system for your business rather than a personal list, that’s usually where working with someone who does this daily saves you the months of trial and error, which is the whole reason I set up AI implementation coaching for business owners who want the shortcut without the folder of shame.
Frequently asked questions
How many prompts should a personal prompt book have?
Somewhere between 20 and 50 is realistic and useful. Beyond that, you’re maintaining an archive instead of a working tool, and you’ll stop reviewing it, which is when it goes stale.
Should I use ChatGPT itself to help organise the book?
Yes, it’s a good use of it. Paste your raw dump of prompts in and ask it to group them by task and flag duplicates, then do the final sort yourself, because it doesn’t know which ones worked for you.
Is it worth buying a pre-made prompt pack instead of building my own?
Pre-made packs are fine as inspiration for structure, but the actual prompts inside them are usually too generic to save you real time, since the good ones depend on your specific tone, clients, and context, not someone else’s.
What’s the biggest mistake people make when compiling one?
Organising by topic instead of by the actual task they sit down to do each week, which means they still can’t find the right prompt fast when they’re under time pressure, defeating the whole point of building the book.
