How to Find Blog Topics with AI Keyword Research Prompts

This page may contain affiliate links.

Every blogger hits the same wall. You want to write, but you have no idea what to write about. You open a keyword tool, get buried under thousands of rows of data, close the tab, and make a cup of tea instead.

The good news is that AI has changed this job completely. Not because it magically knows what your readers want, but because it is very good at turning a vague idea into a structured list of angles, questions and long-tail phrases you can actually check and use. The trick is knowing what to ask it.

Below is the workflow I use to go from "I fancy writing something about home offices" to a shortlist of titles I am confident will get traffic. All prompts are copy-and-paste ready.

Start with a seed, not a blank prompt

AI is a mirror. Ask it something vague and you get something vague back. So always lead with a seed: your niche, your audience, and a rough topic. The more context you give, the less generic the output.

Here is a prompt that works well as a starting point:

"You are a UK-focused content strategist. My blog is about budget home tech for renters. My readers are 25–40, mostly in flats, and they care about things that do not need drilling into walls. Give me 20 blog topic ideas that would attract search traffic. For each one, explain in one sentence who would search for it and why."

The key moves here: you named the audience, the constraint (renters, no drilling), and the format (20 ideas with a reason). That last part matters. Asking for a reason forces the model to justify each idea rather than dumping filler.

Run this three or four times with slightly different seeds and you will quickly see which ideas keep reappearing. Those repeats are usually your strongest candidates.

Turn ideas into real search phrases

Topic ideas are not keywords. "Smart home gadgets for renters" is a theme. "Best smart plugs that do not need a hub UK" is something a person actually types into Google.

Use a second prompt to convert your shortlist into search-shaped phrases:

"For each of the 20 topics above, give me 5 long-tail search phrases that a real person might type into Google. Use natural British English. Include question formats (how, what, which, is, can) and include phrases with 'UK' where it makes sense. Do not include brand names."

You will get things like "do smart plugs work without wifi", "cheapest way to automate lights in a rented flat" and "smart bulbs vs smart plugs which is better". These are the phrases worth chasing. They are specific, they have clear intent, and big publishers often ignore them because the volume looks small.

Small volume is fine. Fifty searches a month from people who are ready to buy beats five thousand from people who are just browsing.

Use AI to find the gaps your competitors missed

This is where AI earns its keep. Most blogs in a niche cover the same ten topics. You want the eleventh.

Paste in the titles of the top-ranking posts for a keyword, then ask:

"Here are the titles and headings from the top 5 ranking articles for the keyword 'smart plugs for renters UK'. Identify the questions these articles do not answer, the assumptions they make, and any sub-topics they skip. Then suggest 5 blog posts that would fill those gaps."

You can grab those headings by hand in a few minutes, or paste in the visible text from each page. The output is usually a list of angles like "what happens to your smart plugs when you move out" or "smart plugs that work with Alexa and Google without an app". Those are the posts that pick up the long-tail searches nobody else is targeting.

One caution: do not let AI invent statistics or product specs. Treat its output as a list of questions to investigate, not as facts to publish.

If you want the shortcut version of all this — the prompts already written, tested and organised by task — that is exactly what our Content & Blogging Pack is. It is 30 prompts covering blog writing, SEO, newsletters and social media, from £9, and it saves you the afternoon of building your own prompt library from scratch.

Validate before you write a single word

AI cannot tell you search volume. It has no live data. So the final step is always manual, and it is quick if you use free tools.

Take your top 10 phrases and check them against:

Google's own autocomplete and "People also ask" boxes. Type your seed phrase and see what Google suggests. Those suggestions are real queries from real people, and they cost nothing to harvest.

Google Trends, filtered to the UK and the last 12 months. You are not looking for huge numbers. You are looking for a line that is flat or rising rather than falling off a cliff.

A free keyword tool such as Google Keyword Planner or a freemium option like Ubersuggest. You only need rough ranges here.

Then apply one simple filter: can you write something genuinely more useful than what already ranks? If the top results are thin, generic or clearly written for search engines rather than humans, that is your opening. If they are excellent, move on.

Build a repeatable monthly routine

The bloggers who never run out of ideas are not more creative. They have a system. Here is one that takes about an hour a month.

Week one: run the seed prompt with three different angles in your niche. Save every idea to a single document, even the bad ones.

Week two: convert the best 20 into long-tail phrases, then check them against autocomplete and Trends. Delete anything that fails.

Week three: for your top five, run the gap-analysis prompt against the current top results. Pick two to write.

Week four: write. Do not research further. You already did the research.

Keep a running "parked ideas" list too. Some phrases will not be worth writing about yet but will become relevant when a product launches or a season comes round. A note in October about "best budget space heater for a small flat" is worth ten times more in November than it is in June.

One more thing worth doing: after a post has been live for a couple of months, feed its title and the search terms it is picking up back into the AI and ask for five follow-up posts that would naturally link to it. That is how a scattering of posts turns into a topic cluster that Google actually understands.

The bottom line

AI will not hand you a winning blog on a plate, and anyone who tells you otherwise is selling something. What it will do is take the tedious part of keyword research — brainstorming, phrasing, spotting gaps — and compress it from an afternoon into twenty minutes. You still have to check the numbers, judge whether a topic is worth your time, and write something a human wants to read.

Start with one seed prompt today. Pick your niche, name your reader, and ask for twenty ideas with reasons. Then run the long-tail prompt on the best five. You will have a month of posts mapped out before your tea goes cold.