I have a graveyard. It’s inside my AI tool, and it’s absolutely full of dead content ideas.

The half-written caption I got interrupted on. The carousel I brainstormed at 11pm because it felt genius at the time, then never touched again. The reel hook I was SO excited about that’s now buried three weeks deep in a chat I will realistically never scroll back to find. All of it, just sitting there, quietly decomposing.

And here’s the thing… if you’ve been using ChatGPT, Claude, or any AI tool to help with your content for more than a few weeks, you’ve got one too. You’ve probably got a bigger one than you think.

This week I dug mine back up. One prompt, and my AI handed me a list of every content idea I’d started and never finished – stuff I’d genuinely forgotten I’d ever come up with. Weeks of ideas, all still good, all still usable, all previously written off as gone. I’m going to show you exactly how to do the same thing, because it might be the least sexy and most useful AI trick I share all year.

Why your ideas disappear in the first place

Let’s be clear about what’s actually happening here, because it’s not a you problem.

When you work with an AI tool day to day, every conversation becomes its own little island. You start a chat about Monday’s email, you get what you need, you close it. Tuesday you open a fresh one about a carousel. Wednesday, another. By the end of the month you’ve had thirty conversations, and each one is holding a couple of ideas you started developing and then abandoned the second something more urgent landed in your lap.

Nobody’s filing these away. There’s no folder called “good ideas I’ll come back to.” They just live wherever the conversation happened to be, and conversations scroll off into the void faster than you’d believe.

So it feels like you’re constantly starting from zero. You sit down to plan content, stare at a blank screen, and think “I’ve got nothing” – when the truth is you’ve got dozens of half-built ideas, you just can’t see them. That’s not a creativity problem. That’s a retrieval problem. Completely different thing, completely fixable.

The feature almost nobody’s using: search your past chats

Both Claude and ChatGPT can now search back through your previous conversations. Not just the one you’re in – the whole history.

Most people have no idea this exists, or they know it’s there and have never thought to point it at their own content. They use AI to make the next thing, never to go back and rescue the last thing. Which is a bit like having a filing cabinet full of your own best work and only ever using the top of it as a desk.

This is the shift I bang on about constantly – the difference between AI as a fancy party trick and AI as an actual business tool. A party trick answers the question in front of it and forgets you existed. A business tool remembers everything you’ve ever told it and hands it back to you when you need it. Searching your own history is you finally using the second one.

The exact prompt to run

Here’s the one I used. Steal it word for word:

“Search our previous chats and give me a list of all the content ideas I’ve started but never finished.”

That’s genuinely it. No clever prompt engineering, no 400-word setup. It goes back through your conversations, finds every idea you opened and didn’t close, and lays them out in a list.

Want it to work harder? A few variations worth trying:

• “…and tell me roughly how finished each one is” – so you can see the quick wins versus the ones that need real work
• “…and group them by platform” – reels here, carousels there, emails over there
• “…and flag which ones are still relevant versus time-sensitive stuff that’s passed” – because some ideas have a use-by date and you don’t want to resurrect a dead promo

Run the basic version first. I promise the list alone will surprise you.

The one catch: projects are their own walled garden

This is the bit that trips people up, so read this part twice.

If you organise your chats inside a project – which, if you’re using AI seriously for your business, you absolutely should be – then the search only sees what’s inside that project. Ask the question from a random new chat floating around outside, and it’ll come back with nothing, and you’ll assume the whole feature is broken.

It’s not broken. You’re just standing in the wrong room.

So if all your content lives in a project called, say, “Content” or “Marketing,” you have to open that project and ask the question from inside it. That’s where your ideas are, that’s where the search needs to look. Ask it anywhere else and you’re searching an empty room and wondering why the shelves are bare.

The flip side is handy too – you can run the same search across your recent chats generally, not just inside a project. Great for catching the stray ideas you had before you got organised, the ones living in loose conversations that never made it into a proper home.

What to actually do with the list

Finding the ideas is step one. Here’s how I turn a graveyard into a content plan without it becoming another overwhelming to-do list.

First, I skim the whole thing and delete anything that was rubbish. Not every abandoned idea deserves resurrecting – some of them got abandoned for good reason, and that’s fine.

Then I pick the two or three that make me go “oh, that’s still really good,” and I get the AI to finish them right there. It already has the context from when I started, so I’m not explaining from scratch. I’m just saying “great, let’s finish this one,” and we pick up where I left off weeks ago.

The rest go into a holding list for when I next need to fill a content gap. Instead of a blank screen and a rising sense of panic, I’ve got a running bank of pre-started ideas to pull from. That’s the whole game, really – you stop creating from nothing and start creating from what you’ve already got.

This is what building a real system looks like

None of this is complicated. That’s exactly why I love it.

You don’t need a new tool, a subscription, or a tech degree. You need to know the feature exists, know the prompt, and know the one rule about projects that stops most people before they start. That’s it. Ten minutes and you’ve clawed back weeks of thinking you thought you’d lost.

This is the kind of thing we build together, live, inside my next masterclass – the small, sharp moves that turn AI from a thing you occasionally poke at into a system that actually runs alongside you and remembers your business. Not theory. The actual buttons, the actual prompts, the actual workflow.

Go and dig up your graveyard first, though. Run the prompt, have a look at what’s been quietly waiting for you in there, and I’d put money on you finding at least one thing worth reviving today.