How to do keyword research with AI

Keyword research starts with a more important question than “what do people search for?”: what type of person do you want to attract and what do you want them to do when they arrive?

Before opening any tool, give your AI this context:

“I have [type of website or channel], aimed at [audience]. I want to attract people searching for [type of content] so they’ll then [goal: hire, buy, subscribe]. Give me 20 topic ideas that audience might search for on Google.”

The key is being specific. “I have a productivity blog for Latin American freelancers who want to organize their time better” gives much better results than “I have a productivity blog.”

Save the list the AI returns. Those topics are your starting point, not your final list.

A topic is broad (“productivity”). A keyword is the exact phrase someone types into Google (“how to organize my time as a freelancer”). The difference matters for SEO.

For each topic on your list, ask the AI to generate the real phrases:

“For the topic ‘[your topic]’, give me 10 search phrases a real person would type into Google. Include variants as questions, with ‘how’, ‘what is’, ‘best’ and comparisons. Put them in a list with the exact format, in lowercase.”

Then go to Google Trends, type some of those phrases and see which ones have sustained or growing interest. Those with zero interest in Google Trends almost always have very low actual volume.

Repeat this process for your 5 to 10 main topics. You’ll end up with a list of 50 to 100 candidate keywords.

Now you need two numbers per keyword: search volume (how many people search for it monthly) and difficulty (how hard it is to appear in the top 10).

The most accessible tool for this is Ubersuggest (free plan: 3 daily searches). For each keyword:

1. Go to neilpatel.com/ubersuggest and type the phrase.
2. Look at the monthly volume and SEO Difficulty (SD). An SD below 30 is a good sign.
3. Scroll to the “Keyword Ideas” section for related variants you might not have considered.

If you want more data without limits, the Ahrefs Free Keyword Generator (ahrefs.com/keyword-generator) also gives volume and difficulty without creating an account.

Ask the AI to help you interpret the results:

“I have this list of keywords with their volumes and difficulties [paste it]. Help me identify which ones have the best balance between real demand and ranking potential with a new site.”

Not every keyword with good volume is right for you. The selection criteria combines three factors:

1. Relevance: the keyword describes something your audience searches for and you can answer well.
2. Manageable difficulty: for a new site, target keywords with SD below 30. As you grow, you can go after harder ones.
3. Clear search intent: the person searching that phrase, what exactly do they want? To learn something, compare options, or buy? Choose keywords whose intent aligns with the content you can create.

Use AI to classify your final list:

“I have these keywords [paste them with volume and SD]. Classify them into three groups: (a) those I can work on now as a new site, (b) those for 6 months from now when I have more authority, and (c) those that don’t apply to my case. Briefly explain the criteria for each group.”

With that, your list of 50 keywords becomes 10 to 15 clear priorities for the next three months.

To turn those keywords into content, check our tutorial on how to create a content strategy with AI.

With your prioritized keywords, the next step is turning them into a plan of concrete articles or posts. One keyword = one article (or video, or post, depending on your channel).

For each main keyword, ask the AI to build the content outline:

“The main keyword for this article is ‘[keyword]’. Give me: (1) an SEO-optimized title that includes the exact keyword, (2) a 155-character meta description, (3) the 5 main subtopics the article should cover and (4) 3 related secondary keywords I can mention naturally.”

Also think about how your articles will connect to each other: this is called a topic cluster. A main article (the most complete one on the topic) links to more specific secondary articles, and all of them link back to the main one. AI can help you map that network:

“I have these 12 planned articles [paste them]. Which should be the central pillar article and how should the others link together?”

If you want to go deeper into semantic search and how Perplexity can complement your research, check our tutorial on how to use Perplexity AI to research.

SEO is not static. What works today can change with new AI tools, market trends or Google updates. Spend 30 minutes a month reviewing your list:

• Ask the AI to generate new ideas based on what happened in your industry that month.
• Check in Google Trends what terms are rising.
• Look at which of your articles generated the most traffic and find related keywords to cover in the coming months.

A useful trick: when you publish each article, ask the AI to suggest 5 related keywords that article could also help rank for. That way you build topical authority systematically, not randomly.

Remember that volume data can change. Always confirm on the official page of the tool you use.