How to Use AI to Brainstorm Blog Post Ideas That Aren’t Generic

1. Start with the generic list — then throw it away

This sounds backwards, but it works. Ask the AI for twenty blog post ideas on your topic and let it be as boring as it wants. Don’t judge them. You’re not looking for a winner; you’re looking for the map. The generic list shows you what everyone else is already writing, which is exactly the territory you need to avoid.

Read through it once and cross out anything you could have guessed before you asked. What’s left — usually two or three items near the bottom, where the model ran out of obvious answers and started reaching — is where the interesting stuff hides. I keep a running document of these leftovers. Half my better posts started as the idea at number 17 that I almost didn’t read.

2. Feed it something real first

AI brainstorming fails when the prompt is just a topic. “Give me blog ideas about gardening” will always return the same list, because the model has nothing of yours to work with. It’s guessing what a generic gardening blog would publish, and it’s right — that’s the problem.

So give it raw material before you ask for ideas. Paste in three reader emails. Drop in the last ten comments on your site. Copy a paragraph from your notebook — the half-formed thought you jotted down on Tuesday. Then ask: “Based on these, what would my readers actually want to read next?”

The difference is immediate. Instead of “10 Gardening Tips for Beginners,” you get something like “Why your tomatoes split after rain — and the one watering habit that stops it.” That’s a post someone searches for, because it came from a real question, not from the model’s idea of a topic.

3. Ask for angles, not topics

A topic is a subject. An angle is a point of view on that subject — and angles are where originality lives. “Productivity” is a topic. “Why productivity advice fails people who work night shifts” is an angle. The first one has ten thousand competing posts. The second has almost none, because it starts from a specific person’s real situation.

When I brainstorm, I ask the AI for angles, not topics. The prompt is usually something like: “Give me five angles on [topic] that assume the reader already knows the basics.” Or: “What would a sceptic say about this topic, and what post would answer them?” Or my favourite: “What’s the advice everyone gives on this topic that doesn’t work for beginners?”

Each of those prompts forces the model away from the obvious. You’re not asking it to summarise the internet. You’re asking it to argue with the internet, and arguments make better posts than summaries.

4. Ban the formats that make everything sound the same

If you let the AI choose the format, it will choose a listicle. Every time. “10 ways to…” is the model’s comfort food, and it will serve it until you tell it to stop.

So tell it to stop. Add one line to your prompt: “No listicles, no ‘ultimate guides,’ no numbered tips.” Then watch what happens — the model has to reach for other shapes. A story. A comparison. A before-and-after. A mistake and what it taught you. A question your readers keep asking.

One of my better-performing posts started this way: I banned the listicle, and the model suggested a post built around a single failed experiment instead. It was more work to write. It also got three times the comments of my usual list posts, because a story about something going wrong is more interesting than ten tips nobody asked for.

5. Use constraints — they’re the real creativity tool

Unlimited choice is the enemy of a good idea. “Give me blog ideas” is too open, and the model fills the vacuum with the safest answers it has. Constraints do the opposite: they force it into corners where the generic answers don’t fit.

Try constraining by audience: “Ideas for people who’ve tried this and failed.” By format: “Ideas that work as a single story, not a list.” By stance: “Ideas that argue against the standard advice.” By time: “Ideas about what changed in this field in the last year.”

My go-to constraint is the contrarian one. “What does everyone in this niche believe that might be wrong?” It doesn’t always produce a usable post — sometimes the consensus is right, and you should say so. But when it lands, it lands hard, because a post that questions the standard advice is a post people share, bookmark, and argue about in the comments.

6. Steal from your own inbox, not from the AI

The best blog ideas don’t come from a brainstorming session. They come from the questions people already ask you — in comments, in emails, in messages from friends who know what you write about. Those questions are proof that someone wants the answer, which is more than you can say for any idea the AI invents.

Once a month, I go through my inbox and my post comments and pull out every genuine question. I paste them into a document and feed the lot to the AI with one instruction: “Turn each question into a post title, and tell me which three are most worth writing.” The model is good at this job — it’s sorting and sharpening, not inventing. The ideas come from real people; the AI just helps me see which ones are strongest.

If you have no comments yet, use your own search history. What did you Google last week that gave you a bad answer? That bad answer is a post waiting to happen.

The honest truth

AI can’t give you original ideas, because originality isn’t something you generate — it’s something you notice. The model can only remix what already exists, and the remix of the average is the average. What it can do, very well, is take the specific things you already know — your readers’ questions, your failed experiments, your half-formed opinions — and turn them into angles worth writing about.

So stop asking it for ideas from nothing. Bring it something real, constrain it hard, ban the listicle, and let it argue with the consensus. The ideas that come back won’t all be good. But the good ones will be yours in a way no generic list ever is — because they started with something only you could have given the machine.

Frequently Asked Questions