How I choose topics for AI-written blog posts

Last updated: October 10, 2026

Quick answer: I use a 3-step filter for AI-written blog posts

If you want AI-written blog posts that earn traffic, I choose topics the same way I would choose any editorial topic: by matching search intent, proving there is a real audience, and picking angles that the model can cover without sounding thin or repetitive. In this post, I’m answering how do you choose topics for AI-written blog posts? in a practical way, not as a theory exercise. The question is not “What can AI write?” It is “What topic can AI write well enough to deserve a click?”

I write on content strategy and search intent, and the first filter is always the same: pick subjects where a machine can help with speed, not subjects where judgment, reporting, or original expertise is the whole product. In practice, I look for topics with stable intent, clear structure, and enough public material to synthesize. A topic with a clear framework is easier to draft than one built on one-off judgment. According to Google’s Search Central documentation, helpful content should be written for people first, and that matters here too: a topic still needs to be useful after the AI novelty wears off. (Source: Google Search Central)

The topic decision starts with one question

I start by asking: will this topic still help a reader after the novelty of AI wears off? If the answer is yes, it is a candidate. If the answer is no, I skip it.

That question sounds broad, but it cuts through most bad topic ideas fast. A topic like “best ways to structure a service business onboarding sequence” can work because the useful part is the framework, not a personal confession. A topic like “my favorite breakfast after a hard week” dies quickly, because the value is in voice and lived detail, not synthesis.

For AI-written posts, I prefer topics with three traits:

1. Stable search intent. The query should mean roughly the same thing for six months or longer. “How to choose blog topics” is stable. A news spike is not.
2. Clear substructure. The topic should break into steps, criteria, comparisons, or examples. AI is strongest when it has a skeleton.
3. Enough public material to synthesize. If the topic depends on private data, interviews, or original testing, AI becomes a weak first draft tool.

The trap is choosing topics that look efficient because they are easy to generate. “100 ideas for X” is easy to write and easy to forget. A better topic is usually narrower: “How to pick B2B blog topics from sales calls” or “Which blog topics work for a local accountant in a 10-mile service area.” Those titles give the model a lane and give the reader a reason to stay.

A generic post would stop at “choose something your audience cares about.” That is true, but uselessly vague. My rule is more specific: choose a topic where the answer can be built from criteria, not from personal taste alone.

What makes a topic a good fit for AI writing?

A good AI topic is one where the outline does most of the work. If I can name the sections before I write the draft, the topic is probably fit for AI assistance. If I cannot, I usually need a human reporter or a subject specialist instead.

The best-fit topics usually sit in one of these buckets:

• How-to explanations
• Comparison posts
• Decision guides
• Glossaries and definition pages
• Process documents
• Troubleshooting articles

These topics work because they reward structure. AI can draft the first pass on a “how do I choose” post, but it still needs a human to decide which criteria matter and which do not. In Ahrefs’ own blog coverage, search behavior is often discussed in terms of intent and usefulness, not volume alone, which is a useful reminder when choosing topics. (Source: Ahrefs)

The worst-fit topics are the ones that ask for originality disguised as guidance. Think of topics built on one of these:

• unique experience
• unpublished numbers
• legal interpretation
• medical direction
• financial advice tailored to a person’s situation
• product claims that require hands-on testing

A topic can still be covered in those areas, but not in the same way. I would not choose “What should I invest in this year?” as an AI-written blog topic unless the post is strictly educational and heavily caveated with professional guidance. I would not choose “which car seat is safest” unless I had authoritative sources and a careful scope. Those subjects are not impossible; they are just easy to get wrong.

One quick test helps me a lot: can I state the reader’s problem in one sentence without using “interesting” or “important”? If yes, the topic is usually concrete enough. If not, it is probably too fuzzy.

I also look at the format match. AI does better with posts that can be broken into repeatable patterns, like:

• “How to choose X”
• “X vs. Y”
• “X mistakes to avoid”
• “X checklist”
• “X for beginners”

That is not because AI is magical. It is because readers scan. A topic that invites scanning gives you room to be useful fast.

How do I know if a topic is worth writing at all?

I only keep a topic if it passes a three-part test: search demand, editorial angle, and commercial value or strategic value. Miss two of those and I move on.

Search demand does not mean chasing huge volume. A topic with modest volume can still be valuable if it attracts the right reader. A post for a niche audience of agency owners, for example, can outperform a generic topic with bigger traffic but weaker intent.

Editorial angle matters just as much. If I cannot explain what the post will say that the other ten results on page one do not, I do not write it. That does not mean the topic has to be revolutionary. It means the post needs a sharp use case, a better framework, or a more exact decision rule.

Strategic value is the part people skip. Some topics do not bring the most clicks, but they support a service page, product, newsletter, or lead magnet. A post about “how to choose blog topics for a law firm” may be more valuable than “blog topic ideas for 2026” because it brings in a narrower reader with higher intent.

I ask four practical questions before I approve a topic:

• Can I explain the point of the post in 15 seconds?
• Would a reader know exactly who this is for?
• Can I support it with sources, examples, or a framework?
• Does it fit the audience I actually want, not just any audience?

If the answer to the last question is no, I drop it. A topic that attracts the wrong reader is worse than no topic. It creates bounce, weak engagement, and vague topical authority.

Here is the blunt version: good AI topics are chosen for fit, not just interest. AI lowers the cost of drafting. It does not rescue a weak topic.

The honest side-by-side: broad topics vs. narrow topics

Broad topics win on reach; narrow topics win on usefulness. I choose narrow topics more often because they produce better AI drafts and better reader satisfaction.

Broad topics are not useless. They matter when I need a pillar page or a high-level explainer. A post on “content marketing strategy” can make sense if it sits above many more specific posts. But it is a hard first choice for AI-written content because the draft tends to sprawl.

Narrow topics win for most AI-written blog posts because they create pressure in the right direction. A topic like “how to choose blog topics for local plumbers” forces decisions: local intent, service area language, seasonal demand, and lead quality. That pressure produces a better post.

The downside is obvious. Narrow topics cap volume. A reader with a big traffic goal may find them too small. I would still choose them if the goal is ranking, conversion, or building a content cluster that compounds over time. Semrush has made a similar point in its discussions of topic clusters and search intent: specific pages are often easier to map to a clear query than broad ones. (Source: Semrush)

Which topic types I would choose first, and which I would skip

I would choose decision guides first. They are the safest and most useful format for AI-written blog posts because they give the model a job: compare criteria, show trade-offs, and lead the reader to a choice. A topic like “How do you choose the right AI blog topic?” works because the post can organize itself around filters, not anecdotes.

I would also choose comparison posts when the two options are real alternatives. “AI vs. human drafting for first-pass blog outlines” is stronger than “AI and humans in content creation,” because the reader is trying to make a decision. Decision intent gives the post a spine.

I would choose process topics when the steps are repeatable and specific. “How to build a topic shortlist from customer service tickets” is a strong AI topic because the sequence can be taught plainly.

I would skip opinion-only topics unless the site already has authority and a strong voice. “Why I think X is better” without evidence is thin. AI can imitate tone, but it cannot create credibility where none exists.

I would also skip topics that require fresh data unless the plan includes real sourcing. Search trends change. Policy changes. Product specs change. If a post depends on live numbers, AI is only the draft layer, not the authority.

The best topic often sits at the intersection of audience pain and repeatable structure. A simple rule helps me decide: if I can imagine the H2s before I write the title, the topic is probably workable. If the structure is still fuzzy, I keep looking.

What changes my mind in exceptional cases?

I reverse the usual choice in four situations.

First, I will choose a broader topic if it is meant to be a pillar page with at least 5 to 10 linked supporting posts. In that case, a page like “AI blog topic selection” can hold the cluster together.

Second, I will choose a topic that looks too narrow if the audience is high-value and specific. A post for SaaS support teams with a 25-person company is better than a general post for all marketers, even if the audience is smaller.

Third, I will avoid AI-first drafting entirely if the subject needs original reporting, legal review, or medical guidance. In those cases, I want a specialist first and AI second.

Fourth, I will pick a topic with lower search volume if it maps to conversion intent. A post that attracts ten strong leads can beat a post that attracts a thousand casual readers.

Those exceptions matter because topic choice is never only about traffic. It is about fit, trust, and what the page is supposed to do after the click. Google Analytics and Search Console can help confirm whether a lower-volume post is still attracting engaged readers, which is why I check performance after publishing rather than guessing. (Sources: Google Analytics; Google Search Console)

What is the final rule I use to pick a topic?

I choose the topic that gives me the best mix of search intent, clear structure, and reader usefulness, then I reject anything that needs invented authority to sound complete. In practice, that often means comparing 2 or 3 candidate topics before I pick one. I would rather have one post with a clear purpose than five posts that all try to do the same job.

Choose a narrow, decision-oriented topic if the post needs to rank, teach, or convert with AI assistance. Choose a broader topic if it is the top of a content cluster and you can support it with several follow-up posts. Neither is right if the topic depends on original reporting, regulated advice, or anything that would collapse without human expertise. It also helps to check the query in Google Trends or a keyword tool before you commit, because a topic can look promising and still have flat demand. (Source: Google Trends)

That is the real answer: AI is not the reason to pick a topic. It is the reason to be more selective.

Key takeaways

• The best AI blog topics are built on clear decisions, not vague inspiration.
• Narrow topics usually outperform broad ones because the outline is easier and the draft is sharper.
• A topic should have stable intent, a definable reader, and enough public material to support it.
• Skip subjects that need original reporting, live data, or specialist judgment you cannot supply.
• If a topic cannot be explained in one clean sentence, it is probably not ready.

Questions readers ask me about AI blog topic selection

How narrow should an AI-written blog topic be?

Narrow enough that you can name the reader and the outcome in one sentence. If the title could fit a dozen audiences, it is probably too broad.

Should I choose topics based on search volume first?

No. I start with fit, then check whether the query is worth the effort. A smaller topic with strong intent is often better than a larger one with weak intent.

Can AI help me generate topic ideas?

Yes, but I treat AI as a suggestion engine, not a decision-maker. It is good for variants, not for deciding what the business should publish.

What topics are a bad fit for AI-written posts?

Topics that rely on private experience, current data, medical judgment, legal interpretation, or original reporting are poor fits unless a qualified human is leading the work.

Is a comparison post better than a how-to post?

It depends on the reader’s intent. If the reader is choosing, comparison wins. If the reader is trying to accomplish a task, how-to usually wins.

Drafted with AI; not yet reviewed by a person.