Keyword research used to mean one thing: find the words people type into Google. That’s only half the job now. People ask ChatGPT, Perplexity, and Google’s AI Overviews full questions, in plain sentences, and expect a direct answer. If your keyword list ignores that, you’re optimizing for a version of search that’s shrinking. Here’s how to do both, without needing a big budget or a fancy tool stack.
Start With the Old-School Basics
The fundamentals still work. Before you touch AI queries, get your traditional list solid. Keyword research means finding the terms your customers use and how often they search them. A local bakery might target “gluten free birthday cake near me” instead of just “bakery.” Specific wins.
- Use a free tool like Google Keyword Planner or AnswerThePublic to pull search volumes.
- Sort keywords by intent: informational (“how to store sourdough”) versus buying (“order sourdough online”).
- Check the difficulty score. As a beginner, chase terms under 30, since ranking for “best pizza” against national chains is a losing game.
Plenty of small businesses hire keyword research services for this step alone, but you can get a working list in an afternoon yourself.
Then Add Conversational AI Queries
This is the part most 2027 guides still skip. AI tools don’t get fed keywords. They get questions. So, whether you are opting for keyword research services or doing it yourself, you need to understand this.
How People Actually Ask AI
Someone typing into Google searches “protein powder side effects.” The same person asks ChatGPT, “Is it safe to take protein powder every day if I have a sensitive stomach?” Longer. More specific. Full of context you can answer. To find these, do three things:
- Open ChatGPT or Perplexity and ask questions the way your customer would. Note the follow-ups the AI suggests.
- Read the “People also ask” boxes on Google, then rephrase each as a natural sentence.
- Check your own customer emails and support tickets for the exact phrasing people use.
Why Format Matters
AI answer engines pull from content that reads like a direct reply. So structure a page around one clear question with a tight answer in the first two sentences. Then expand. Pages built this way get cited in AI Overviews far more often than keyword-stuffed ones, and a citation there can send steady referral traffic even when you don’t rank first on the blue links.
Map Keywords to Pages, Not Just a List
A spreadsheet of 200 keywords does nothing until you assign them. One target keyword per page. Group the close variations under it. Say you sell home gym equipment. Your “adjustable dumbbells” page can also cover “space saving dumbbells” and “are adjustable dumbbells worth it,” because they’re the same buyer asking slightly different things. Three keywords, one page.
- Give each page a primary keyword and two or three supporting ones.
- Write one page that answers the AI-style question fully, and link it to your product page.
- Skip keywords you can’t realistically satisfy. If someone wants a free tool and you sell paid gear, that traffic won’t convert.
Good SEO services do this mapping first, before writing a single word, because it stops you from producing five pages that compete against each other.
Refresh Every Quarter
Search behavior shifts fast now. A term that had no volume in January can spike by summer once a trend hits. Set a calendar reminder every three months to rerun your list, add new AI questions you’ve spotted, and drop the keywords that never brought anyone in.
Conclusion
If this all feels like a lot, it is at first. Start with ten keywords, not two hundred. Pick the ones closest to a sale, build a page for each, and add the conversational versions as you go. Many owners eventually outsource the ongoing work to professional SEO services once it proves out, but the early wins usually come from just paying attention to how your own customers ask for what you sell. Small list. Real intent. That’s the whole game.
