AI keyword research: find and group keywords with AI tools
Short answer Feed a chatbot your real words (Search Console queries, customer emails, your service list) and ask for phrases and groups, not numbers. Group by one test: could one page fully answer both searches? Paid tools cluster keywords automatically: Semrush from $139 a month, Ahrefs from $249. Check every group against Google before building pages.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callAI keyword research in three steps
AI does two jobs well here: finding the many ways people phrase a need, and sorting a long list into pages. It does a third job badly: counting searches. So the work splits like this.
| Step | Who does it | Output |
|---|---|---|
| 1. Find phrases | A chatbot, fed your own data | 50–300 phrases customers might type |
| 2. Group them | A chatbot or a paid tool's clustering | A handful of groups, one per page |
| 3. Check them | You, with Google's free tools | Groups worth building pages for |
Google is fine with AI for research. Its guidance says:
"Generative AI can be particularly useful when researching a topic, and to add structure to original content." (Google Search Central: Guidance on using generative AI content)
Step 3 is covered in AI keyword research: where to check the numbers. This page is about steps 1 and 2.
Which AI tools can do keyword research, and what they cost
You can do all of it with free tools. Paid tools save time on big lists and add their own volume estimates.
| Tool | AI job it does | Cost (November 2026) |
|---|---|---|
| ChatGPT or another chatbot | Brainstorms phrases, labels intent, groups lists, reads spreadsheets you upload | Free plan has "Limited" file uploads and data analysis; paid plans have them in full (ChatGPT pricing page) |
| Google Search Console | Not AI; gives the real searches your site already shows for, to feed the chatbot | Free |
| Google Keyword Planner | Not AI-branded; suggests related keywords and volume ranges | Free, but Google requires billing details in a Google Ads account to use "Get ideas for new keywords" (Google Ads Help) |
| Semrush Keyword Strategy Builder | Turns up to 5 seed keywords into topics, pillar pages and subpages | Paid SEO plans; the cheapest listed at $139 a month (Semrush pricing) |
| Ahrefs Keywords Explorer | AI suggestions and keyword clusters | Listed on the Standard plan and up, $249 a month (Ahrefs pricing) |
Prices are in US dollars as the vendors list them. For hands-on steps in the two most used tools, see Keyword Planner and Ahrefs step by step.
How to find keywords with AI: give it your real words
A chatbot asked cold for "keywords for a plumber" returns the same generic list every plumber gets. Feed it words only you have.
Your Search Console queries. In Search Console, open Performance, then Queries, and export. Google says the most you can export from the screen "is 1,000 rows of data" (Google Search Central Blog), plenty for a small business. Upload the file and ask:
- "These are searches that showed my site. Which ones mean the person wants to hire someone? List 30 more phrases a customer with the same need might type."
Customer messages. Paste 10–20 emails, form submissions or call notes, with names removed. Ask:
- "List the exact phrases these customers use for their problem. Then list how they'd likely type each one into Google."
Customers say "my basement smells musty"; owners write "mold remediation". The first phrasing is often the one nobody else targets.
Your service list and area. Ask for variations by service, problem, price and place:
- "For each service below, list searches about the problem, the cost, the comparison with other options, and the service plus each city I serve. No search volumes."
Asking for "no search volumes" matters: a chatbot without live data may invent them.
For more on mining your own data, see keyword research with Search Console.
How to group keywords with AI
Group by the answer, not the shared word. Two searches belong together when one page could fully answer both. "Water heater repair" and "water heater leaking" want the same visit from a technician. "Tankless vs tank water heater" wants a comparison: a different page.
With a chatbot. Paste your list and ask:
- "Group these phrases so that each group could be answered by one page. Name each group by the page it needs (service page, price page, comparison article, city page). Put each phrase in only one group. Flag phrases you're unsure about."
Then spot-check the flagged ones in Google: search both phrases. If the same kinds of pages show up for both, keep them together.
With a paid tool. Both big tools sort keywords into pages for you:
- Semrush's Keyword Strategy Builder "analyzes up to 10,000 keywords and groups them into topics, pillar pages, and subpages", and ranks pages by relevance, search volume, keyword difficulty, domain diversity and SERP features (Semrush Knowledge Base).
- Ahrefs groups keywords by Parent Topic. To find it, Ahrefs says, "we take the #1 ranking page for the keyword and find the query responsible for sending the most search traffic to that page" (Ahrefs Help Center).
Ahrefs' method follows what Google already ranks, which is closer to how Google sees it than word matching. Keywords sharing a Parent Topic can usually be served by one page. That matches Google's own advice that you don't need a page per phrasing:
"Google's language matching systems are sophisticated and can understand how your page relates to many queries, even if you don't explicitly use the exact terms in them." (Google Search Central: SEO Starter Guide)
Put the result into a keyword map, one row per page. How to build and use one is in after keyword research.
Where AI grouping goes wrong
Read the groups before you act on them. Four mistakes come up often.
| Mistake | What it looks like | Fix |
|---|---|---|
| Too many groups | A separate page for "AC repair", "air conditioner repair" and "AC fix" | Merge them: one page answers all three |
| Mixed intent | "Roof replacement cost" grouped with "roof replacement near me" | Search both; split if Google shows price articles for one and local businesses for the other |
| Services you don't sell | Groups for jobs you never take | Delete them; traffic you can't serve costs you time |
| Missing rare searches | Very specific phrases absent from your export | Google hides "anonymized queries", ones not "issued by more than a few dozen users over a two-to-three month period" (Google Search Central Blog); add them from customer messages |
Too many groups is the costly one. Pages written for every small variation, especially with AI, can cross Google's line: it warns that using AI "to generate many pages without adding value for users may violate Google's spam policy on scaled content abuse" (Google Search Central: Guidance on using generative AI content). See rank for multiple keywords with one page.
Example
Illustrative cases with invented numbers.
An HVAC company (US dollars)
The owner exports 600 queries from Search Console and adds 15 service-call notes. ChatGPT returns 180 phrases in 22 groups. He merges near-duplicates down to 9 pages: 5 service pages, 2 price pages and 2 comparison articles. Cost: $0 and one afternoon.
A Calgary accountant
She pays for one month of Semrush (US$139) and enters three seeds: "small business accountant", "bookkeeping" and "corporate tax return". The tool proposes 14 pages. She keeps 6 that match services she sells, exports the list and cancels.
A cleaning company with a big list
An agency hands the owner 2,000 keywords. Instead of paying for a tool, he uploads the list to a paid chatbot plan and asks for groups by page. Searching the 12 flagged pairs in Google splits 3 groups. He ends with 11 pages, not 40.
FAQ
Can ChatGPT do keyword research?
Yes, for finding phrasings and grouping them, best when fed your own search data. It can't count searches; check volumes in Keyword Planner.
What is AI keyword clustering?
Sorting a keyword list into groups that one page can answer. Semrush and Ahrefs do it automatically; Ahrefs groups by what already ranks.
Do I need a paid tool to group keywords?
Not for a few hundred phrases. A chatbot plus a Google spot-check of unsure pairs works. Paid clustering saves time on thousands.
How many keywords should go on one page?
As many as one page can fully answer. One main search plus its close variations is normal; Google matches the variations on its own.
Is it safe to write the pages with AI too?
Yes, if each page adds value. Google warns mass-produced AI pages without value can break its scaled content abuse policy. See does Google penalize AI content?.
Different for you?
- You want to know which AI numbers to trust: AI keyword research: where to check the numbers.
- You're starting from zero with free tools: keyword research for a local business.
- You have groups and need pages: after keyword research.
- You'd rather have the research, grouping and pages done for you: SEO & GEO Agent.