How to Come Up With Content Ideas Using AI (Prompts That Work)

Monday morning, blank page. You need a blog post, three social media posts and a newsletter for Thursday, and no ideas are coming. Anyone who creates content regularly knows that moment. The good news: coming up with content ideas using AI is now one of the simplest and most rewarding ways to use ChatGPT and its competitors.

You still need to know how to go about it. “Give me some blog post ideas” produces a bland list, the same one your competitors get. In this guide, I show you how to generate content ideas with AI methodically: the information to provide, the brainstorming techniques that work, ready-to-copy prompts, and above all how to sort through them so you keep only the ideas worth producing.

Contents

1. Why AI is useful for finding ideas
2. Step 1: Give it the right context
3. Step 2: Generate lots of ideas
4. Step 3: Use brainstorming techniques
5. Step 4: Sort and critique the ideas
6. Step 5: Turn an idea into an outline
7. Prompts by type of content
8. Which AI should you use to find ideas?
9. Mistakes to avoid and limits
10. My checklist before a brainstorming session
11. Frequently asked questions
12. The bottom line

Why AI is useful for finding ideas

Running out of ideas rarely comes from a lack of knowledge. It usually comes from fatigue, from always looking at your topic from the same angle, or from the struggle to go from a vague theme (“gardening,” “accounting”) to a specific, publishable topic.

That’s exactly where AI helps. It never gets tired, it can produce fifty variations in seconds, and it’s good at combining a topic with a format, an audience or a season. In my own use, I never ask it for “the” right idea: I ask it for food for thought, then I choose.

What it does well:

• Breaking down a theme into dozens of specific topics.
• Switching angles: beginner, expert, skeptic, short on time, tight budget.
• Adapting an idea to several formats: article, short video, carousel, newsletter, podcast.
• Organizing: grouping by theme, ranking by priority, building a calendar.

What it does poorly, on the other hand: knowing what really interests your readers, knowing the latest news in your field without searching, or guessing what makes you different. That information is yours to bring.

Step 1: Give it the right context

The first mistake is asking for ideas without explaining anything. The AI then answers with the most obvious topics. Before any brainstorming session, I prepare a short context paragraph that I reuse every time.

I create content for [type of website / account / newsletter]. My main topic is [theme]. My audience: [who, level, age, situation, main problems]. My goal: [drive traffic, build loyalty, sell a service, inform]. My tone: [simple, expert, funny, warm]. I've already published on: [list of 5 to 10 topics]. I want to avoid: [overdone topics, topics outside my expertise]. First, ask me 3 questions if anything is missing.

That last sentence matters a lot: it pushes the AI to have you clarify what’s vague, rather than making things up. If you often use ChatGPT for the same project, you can save this context in custom instructions or group your chats in a ChatGPT project, so you don’t have to paste it every time.

A concrete example

Take Maya, who runs a hiking blog about coastal Maine for families with young kids. Her context looks like this: an audience of parents aged 30 to 45, kids aged 3 to 10, not much hiking experience, a goal of building loyalty through a newsletter, and a warm, practical tone. With that context, the AI no longer suggests “10 hiking tips” but topics like “what to pack for a 5-year-old on a 2-hour hike” or “coastal trails with no dangerous cliffs.” The difference is immediate.

Step 2: Generate lots of ideas

Once the context is set, ask for volume. The good ideas often show up at the end of the list, once the obvious ones are used up.

Based on this context, suggest 30 content ideas. For each one: a working title, the specific reader problem it solves, and the best format for it. Avoid the generic ideas you see everywhere.

Then follow up to get off the beaten path:

Give me 15 more ideas that are more original and less obvious. Include contrarian ideas, ideas based on common mistakes and ideas for very specific situations.

Step 3: Use brainstorming techniques

Beyond a simple list, some techniques produce much richer results. Here are the ones I use most.

The audience × problem matrix

List your types of readers and their problems, then have the AI cross the two.

Here are my 3 reader profiles: [profile 1], [profile 2], [profile 3]. For each one, list 5 concrete problems related to [theme]. Then suggest one content idea for each profile × problem combination, as a table.

Real questions from real people

The best ideas answer real questions. If you have customer questions, comments or messages you’ve received, paste them in (with names removed).

Here are questions and comments I've received from my readers: [list]. Group them by theme, identify the questions that come up most often, then suggest one content topic for each group.

Without that material, you can also ask the AI to imagine a beginner’s questions, but keep in mind these are hypotheses, not data.

Multiple angles

The same topic can turn into ten different pieces of content depending on the angle.

Take the topic "[topic]" and break it into 10 different angles: for beginners, for experts, mistakes to avoid, comparison, case study, myth to debunk, checklist, tight budget, quick 5-minute version, personal point of view.

The calendar and the seasons

Suggest a 12-month content calendar for [theme], taking into account the seasons, holidays, back-to-school and the times when my audience has specific needs. One main idea per month, plus two secondary ideas.

Repurposing existing content

Your past content is a gold mine of ideas. Paste an article or an outline and ask for follow-ups.

Here's an article I published: [text]. Suggest 8 spin-off pieces: topics to go deeper on, questions left unanswered, short formats for social media, and an idea for a sequel.

Step 4: Sort and critique the ideas

A list of 45 ideas is useless if you don’t know which one to produce. It’s the step many people forget, and yet it’s where AI helps me the most.

Here's my list of ideas: [list]. Rate each one from 1 to 5 on: likely interest for my audience, originality, ease of production, fit with my goal. Present the result as a sorted table, then tell me honestly which ones are too bland.

Explicitly asking for criticism is important: by default, AI tends to find every idea “excellent.” Here’s an example of the kind of grid you might get for Maya’s blog:

Idea	Audience interest	Originality	Effort	Keep it?
What a 5-year-old needs for a 2-hour hike	High	Medium	Low	Yes
10 hiking tips	Medium	Low	Low	No, overdone
Coastal trails with no dangerous cliffs	High	High	Medium	Yes, check on the ground
Hiking in the rain with kids	Medium	High	Low	Yes

The scores in the table are still just the AI’s opinion of the ideas: they don’t replace your site analytics or your knowledge of your readers. It’s a thinking aid, not a verdict.

Step 5: Turn an idea into an outline

Once you’ve picked the idea, AI can help you move to a detailed outline.

For the topic "[title]", suggest a detailed outline with the questions the reader wants answered, the sections, the examples to include and what would make this piece more useful than what's already out there. Flag the points I'll need to verify or illustrate with my own experience.

For the writing itself, I explain my method in my guide to writing an article with AI.

Prompts by type of content

Needs vary depending on the channel. Here are a few tailored prompts.

For a blog

Suggest 15 blog post topics for [audience] about [theme]. Each topic should answer a specific question someone might type into a search engine. For each one, note the reader's intent: learn, compare, find out how to do something, choose.

For social media

Suggest 20 short post ideas for [social network] about [theme], split between practical tips, behind the scenes, questions for the community, before/after and common misconceptions. For each idea, suggest a catchy opening line.

For a newsletter

I publish a [frequency] newsletter for [audience]. Suggest 10 issue ideas, each with a main topic, a short recurring section and a question to ask readers to get them to respond.

For videos

Suggest 10 ideas for videos under one minute about [theme] for [audience], with the hook for the first 3 seconds and what we see on screen.

And to fit these ideas into a broader strategy, read my guide to using ChatGPT for marketing.

Which AI should you use to find ideas?

For brainstorming, the major assistants are fairly evenly matched, and your prompt matters most. ChatGPT is the most versatile and its free version is enough to get started; ChatGPT Plus, at $20 a month, gives you more access and tools if you use it every day. Claude is appreciated for how nuanced its rewording is. Gemini is handy if you already work in Google Docs and Drive.

In my experience, it’s useful to give the same request to two different assistants: the lists partly overlap, but each one brings a few ideas the other missed. To compare the tools, read my article on which AI to choose.

Mistakes to avoid and limits

Publishing ideas without validating them

AI suggests plausible topics, not guaranteed ones. It doesn’t know what your readers are really looking for, or what has already worked well for you. Always cross-check its suggestions with your analytics, the questions you receive and your gut. I won’t promise you that AI can find “the best SEO ideas” on its own: it can’t, and no generated list replaces knowing your audience.

Trusting the trends it claims

If the AI tells you about a “current trend” or a statistic, check it. AI tools can be wrong, and OpenAI itself points out that responses aren’t always accurate and should be evaluated before use. If you need up-to-date information, use the tool’s web search and look at the cited sources.

Mass-producing weak content

Finding 50 ideas in 5 minutes doesn’t mean you should publish 50 pieces. Google says its criterion is content quality, not how it was produced. However, using automation or AI primarily to manipulate search rankings goes against its spam policies. A few truly useful pieces beat an avalanche of interchangeable text.

Losing your own voice

AI ideas are often decent but lack personality. What makes the difference is your examples, your mistakes, your opinions. For every idea you keep, ask yourself: what can only I bring to this topic?

Sharing sensitive data

To enrich the context, it’s tempting to paste customer messages or internal data. Remove names, addresses and anything that could identify someone, and check your company’s rules before sharing internal documents with an outside service.

My checklist before a brainstorming session

• I’ve written my context paragraph: audience, goal, tone, topics already covered.
• I’ve asked for at least 30 ideas, then a second, more original batch.
• I’ve used at least one technique: matrix, real questions, angles, calendar or repurposing.
• I’ve had the list scored and critiqued.
• I’ve checked the ideas I kept against my analytics and my readers’ questions.
• I’ve identified what I personally bring to each chosen topic.

Frequently asked questions

Can you find content ideas with ChatGPT for free?

Yes. Brainstorming works really well on the free version of ChatGPT. A paid subscription is only useful if you use it heavily or need extra tools. See my article on ChatGPT Free.

Can AI find the topics that will perform best?

No, not on its own. It can suggest and organize ideas, but it doesn’t know your readers or your past results. Validating them is up to you, with your data and your experience.

Does Google penalize AI-generated content?

Google says it judges the quality of content, not how it was produced. The problem is using AI primarily to manipulate rankings, for example by mass-publishing pages with no value.

How do I avoid bland ideas?

Give precise context, ask for lots of ideas, insist on original angles and ask the AI to flag the overdone ideas itself. Also bring in real material: customer questions, comments, past content.

How many ideas should I ask for?

I usually ask for 30 ideas, then a second batch of 15 more original ones. Beyond that, the lists get repetitive. It’s better to switch angles or techniques than to ask for 100 ideas at once.

Can you build a full content calendar with AI?

Yes, it’s a great use case. Ask for a monthly or weekly calendar in a table, then adjust it to your real production pace.

The bottom line

To come up with content ideas using AI: give precise context, ask for volume, use brainstorming techniques (angles, matrix, real questions, calendar), have the list critiqued, then validate it yourself with what you know about your audience. AI is a fantastic idea machine, but you’re the one who picks the ideas that deserve to exist.

To go further, check out my guide to writing a good prompt.