AI keyword research means using a chatbot like ChatGPT or Claude to figure out the exact words and phrases your customers type (or say) when they’re looking for what you sell — then using that language to shape your website, blog posts, and product pages. You don’t need spreadsheets, keyword-difficulty scores, or an SEO certification to do it well. You need the right prompts, a little context about your business, and a way to double-check what the AI hands back.
If you’ve ever stared at a “keywords” tool and felt your eyes glaze over, this is the version of keyword research built for someone running a business, not studying for an SEO exam.
Why keyword research even matters when everyone’s asking AI instead of Googling
Here’s the thing that changes everything about how you should think about this: your customers aren’t typing three-word phrases into a search box anymore. A lot of them are having a conversation with a chatbot instead. A January 2026 Search Engine Land study found 37% of consumers now start their searches with AI rather than Google, and a July 2025 AP-NORC survey found 60% of U.S. adults have used AI to search for information, with adoption reaching 74% among people under 30.
And it’s not just casual browsing. About half of U.S. adults now use AI chatbots, up from a third in 2024, and searching for information is the single most common chatbot use case, ahead of work tasks. On the buying side, a global IBM study of over 18,000 consumers found nearly half now turn to AI during their buying journeys, using it to research products, interpret reviews, and hunt for deals.
This matters for keyword research specifically because the way people phrase things has changed. People can use far longer, more specific prompts with AI tools than in traditional search, and keyword research now needs to cover these conversational, intent-rich prompts too. Nobody’s typing “hormone balance meal plan.” They’re asking something closer to a full sentence — and that’s exactly what your content needs to match.
Start with context, not a keyword list
The biggest mistake I see people make when they first try this: they open ChatGPT and type “give me keywords for my bakery.” That’s too vague to be useful. Treat the AI like a new hire who needs a briefing before it can do good work.
Give it:
- What you sell and who you sell it to (be specific — “custom birthday cakes for parents in a 20-mile radius,” not “baked goods”)
- The problem your customer has right before they find you
- Any language your customers already use — pull it from DMs, reviews, or emails if you can
One practical workflow: open an AI chatbot and give it context about your business — for example, “Act as an SEO strategist for [your business]. My target audience is [specific description]. Give me 15 strong seed keywords related to [your offers]. Focus on evergreen and rising trends.” That single prompt turns a blank page into a working list in under a minute.
Ask AI to think like your customer, not like a search engine
Once you have a seed list, push further. Ask the AI to separate keywords by what the person actually wants when they type it — informational (just learning), commercial (comparing options), or transactional (ready to buy). This distinction matters more than it sounds like it should. Two keywords can look identical in a keyword planner but attract completely different audiences — “best CRM for small business” signals commercial intent, while “what is a CRM” signals someone still in early research. If you write one blog post trying to serve both, it usually serves neither.
A prompt that works well here: “Group these keywords into informational, commercial, and transactional categories, and tell me what kind of page or content would satisfy someone searching each one.” You’ll usually get back a much smarter map than the raw list you started with, because AI has shifted keyword research from manual sorting by volume to semantic clustering and automated intent classification, grouping keywords by meaning rather than exact-match similarity so the results align with what actually drives revenue.
Don’t skip the human sources — they’re where the gold is
AI is great at generating options, but it’s guessing unless you feed it real language. Before you finalize anything, go pull the actual words your customers use. Start with your Search Console data, support tickets, sales calls, and onboarding questions to find the language your audience already uses — this often beats volume-first keyword tools — and check Reddit, YouTube comments, and forums to spot real intent and long-tail phrasing that bigger brands miss. Paste a batch of that raw language into your AI chat and ask it to cluster and clean it up. That’s a far better foundation than anything generated from scratch.
This is also where a lot of “keyword volume” advice starts to break down. An estimated 15% of daily searches are entirely new, so it’s worth targeting zero-search-volume long-tail queries pulled from places like Reddit, support logs, and AI-tool related questions — the kind of specific phrasing a keyword tool would never surface because no one’s searched it enough times yet to register.
Check AI’s suggestions against something real
Here’s my honest opinion after using this workflow for a while: AI keyword lists are a strong first draft, not gospel. It will occasionally suggest a phrase that sounds plausible but nobody actually searches. Cross-check the list against something free before you build content around it. For most small business owners, start with the free stack — Google Search Console, Google Keyword Planner, AnswerThePublic, and Google Trends together give you more than enough to build a solid six months of content.
You genuinely don’t need a perfect strategy before you start writing. You don’t need a perfect keyword strategy before you start — you need five keywords and a willingness to show up. Publish something built around one, see what happens, then repeat.
Remember: you’re now writing for two audiences
This is the part that trips people up. You’re not just trying to rank in Google anymore — you’re trying to get mentioned when someone asks ChatGPT or Perplexity a question about your industry. AI citations research shows 91% of AI citations come from brand-managed sources, with AI models prioritizing structured, verifiable facts, making owned data the primary driver of AI visibility. That means clear, specific, well-organized content on your own site still matters — arguably more than ever, since AI tools have to find something worth quoting.
The reassuring part is that this doesn’t require a totally different skill set. Research from Semrush confirms that long-tail, detailed queries still drive strong click-through rates even when an AI Overview appears on the page — the content that gets skipped is thin and vague, while the content that thrives is useful, specific, and human. Write for the actual person with the actual problem, using the actual words they use, and both Google and the AI tools quoting it tend to reward that.
A simple starting workflow you can use this week
- Ask your AI chatbot for 15–20 seed keywords based on specific context about your business and customer
- Ask it to sort those into informational, commercial, and transactional buckets
- Pull 10 real phrases from your own reviews, DMs, or sales calls and add them to the mix
- Cross-check the strongest 5–10 against Google Trends or AnswerThePublic to see if anyone’s actually searching for them
- Pick one, write one piece of content around it, and see what happens before you build the next ten
Keyword research used to mean staring at a spreadsheet of numbers you didn’t fully understand. Now it’s more like a conversation — you describe your customer, the AI helps you hear how they actually talk, and you write to that person instead of to an algorithm. That’s a much better use of your Tuesday afternoon.
Hi! I use AI to help research and write posts on this site. I do my best to keep things accurate, but please double-check anything important — and nothing here replaces advice from a licensed or certified professional.