Keyword research with AI tools

Keyword research means finding the words people type into search engines when they are looking for what you offer. For a membership site, it decides which free articles, guides and landing pages are worth writing, because those public pages are how strangers find you before they ever see your sales page. Guess wrong and you spend months writing for searches nobody makes.

AI assistants are excellent brainstorming partners for this work. They can produce hundreds of phrasing variations, group them into topics and suggest what each searcher wants. What they cannot do on their own is tell you how many people actually search for any of it. This article shows you how to use AI for the thinking and real search data for the checking.

Know what AI can and cannot tell you

A general assistant such as ChatGPT, Claude or Gemini has learned how people talk about almost every topic, so it is good at suggesting the words and questions your audience might use. But unless it is connected to a real keyword data source, any search volume it gives you is an estimate at best and a guess at worst. Treat every number from an AI chat as unverified until you have checked it against real search data. Some dedicated SEO tools now include AI features on top of their own data; those numbers are still estimates, but they are grounded in something.

The safest way to think about it: AI widens your list of ideas, and search data narrows it down.

Start with member problems, not seed keywords

Traditional keyword research starts with a few obvious terms and expands from there. You get better results by starting with the problems your future members are trying to solve, because that is how they search. Harriet runs a sourdough baking membership. Her obvious keyword is “sourdough,” which is far too broad to win. Her members' real searches look more like “why is my sourdough flat” or “sourdough starter not rising in the fridge.”

I run a membership for [describe your audience]. Before they join, they are usually struggling with [list three to five problems]. For each problem, list 15 ways a real person might type it into a search engine, including questions, short phrases and the words a beginner would use rather than an expert. Mark any phrase that suggests the person is ready to pay for help, such as searches for courses, classes, coaching or communities.

Add the questions you already hear in emails, webinars and forums. Those are the most valuable phrases of all, because you know they are real.

Group phrases by topic and intent

Search intent is the reason behind a search. Someone typing “what is a sourdough discard” wants a quick explanation. Someone typing “online sourdough class” is comparing options and may buy. One page can usually serve a whole group of similar phrases with the same intent, so grouping stops you writing five thin articles that compete with each other.

Here is my list of search phrases: [paste list]. Group them into topics, where each topic could be answered by one page. For each group, suggest a working title, state the main intent (learning something, comparing options, or ready to buy), and name the single phrase that best represents the group. Flag any phrases that do not fit my membership about [topic] so I can drop them.

A useful result looks like a set of tidy clusters: “Flat or dense loaves” with eight related phrases and a learning intent; “Choosing a sourdough class” with four phrases and a comparing intent. If the AI produces one giant list, ask it again with a limit on how many phrases each group may contain.

Check every idea against real search data

This is the step that turns a brainstorm into research. Use at least one real source:

• Google Search Console. If your site already has public pages, this free tool from Google shows the searches you already appear for, how often and how many clicks you get. It is the most trustworthy data you have, because it is about your own site.
• Google's autocomplete and related questions. Type your main phrases into Google and note what it suggests and which questions appear. These reflect real searches, although without numbers.
• A keyword tool. Google's keyword planner in an ads account, or a dedicated SEO tool, gives volume ranges and a sense of competition.

Go through your AI-generated groups and mark each one as confirmed, weak or not found. Many owners find that some of the AI's most confident suggestions barely register in real data, while a few of the plain, beginner-style phrases turn out to be steady searches. Remember that niche membership topics often have low volumes; a phrase searched by a small number of the right people can still be well worth a page. For methods that need no paid tools, see keyword research for membership sites without expensive tools.

Choose what to write, and a person makes the call

Once your list is verified, prioritize it. Weigh three things for each topic: how closely it relates to what your membership teaches, whether the searcher could become a member, and whether you can realistically write something better than what already ranks. Look at the current top results yourself; AI summaries of search results can be out of date or simply wrong.

You can ask AI to propose a ranking, but you make the final decision, because you know which topics lead naturally to your paid content and which you are genuinely qualified to write about. If you are still deciding which assistant to use for this kind of work, choosing your first AI assistant covers how to trial them on a real task.

Your first keyword session

1. Write down three to five problems your future members have before they join.
2. Use the first prompt to generate search phrases, and add real questions from your inbox.
3. Group the phrases by topic and intent with the second prompt.
4. Check every group in Google Search Console, Google's own suggestions or a keyword tool.
5. Drop anything unconfirmed or unrelated, and pick three topics to write first.
6. Keep the verified list in a spreadsheet and revisit it every few months.