Keyword research is how you figure out what your customers actually type (or ask) when they need something you offer. It turns guesses about "what people want" into a list of real queries you can build pages around.
Done well, it shapes your whole content plan. It tells you which topics deserve a page, which pages should be merged, and which searches are worth chasing given your budget and timeline.
What is keyword research?
Keyword research is the process of finding, analyzing and prioritizing the search terms people use to look for products, services and information related to your business. The output is usually a spreadsheet of keywords grouped by topic, with notes on intent, demand and competition, plus a decision about which page should target each group.
Google's own SEO Starter Guide frames it simply: think about the words a person might search for to find your content. It also points out that experts and beginners often describe the same thing in different words. Keyword research is the disciplined version of that advice.
A "keyword" doesn't have to be one word. Most useful keywords are phrases like "payroll software for restaurants" or "how to fix a leaking faucet," and many are full questions.
Why keyword research matters for search and AI answers
Without keyword research, teams tend to write about what they find interesting, using their own internal jargon. Customers search in their own words. Research closes that gap so your pages match the language and needs of the people you want to reach.
It also protects your time. A search volume estimate tells you whether anyone is looking, and a keyword difficulty score hints at how much work it might take to compete. Together they help you avoid spending months on a page nobody needs or one you can't realistically rank.
The role is shifting a bit with AI search. Google says its AI Overviews and AI Mode may use a technique called "query fan-out," running several related searches across subtopics to build a single answer, according to its AI features documentation. That rewards sites that cover a topic thoroughly, not just one exact phrase. Keyword research is still the best map of those subtopics, and Google states that no special optimization is needed beyond normal SEO fundamentals to be eligible.
How keyword research works
Most processes follow the same broad steps. The Ahrefs keyword research guide lays out a similar sequence: generate ideas, organize them, classify intent, evaluate with metrics, then prioritize.
1. Start with seed keywords
Seed keywords are the broad terms that describe your business, like "accounting software" or "dog grooming." Pull them from sales calls, support tickets, reviews and the way customers describe their problems. Your own site's data is a goldmine too: the Queries tab in the Search Console Performance report shows the searches that already bring you impressions and clicks.
2. Expand the list
Feed your seeds into a keyword tool (Google Keyword Planner, Ahrefs, Semrush and similar) to find related phrases, questions and modifiers. Look at the pages competitors rank with, autocomplete suggestions, People Also Ask boxes, forums and community threads. You're looking for every reasonable way a person might describe the need.
3. Group by meaning and intent
Next, cluster keywords that mean the same thing so one page can serve the whole group. A quick test: if Google shows mostly the same results for two queries, they probably belong on one page. Then label each group by keyword intent, such as informational, commercial, transactional or navigational, so you know what kind of page to build.
4. Evaluate and prioritize
Score each group on demand, difficulty and business value. A high-volume term that has nothing to do with what you sell is a distraction. A modest one that maps directly to a paid product can be worth far more. Seasonality matters too. Google Trends shows relative interest over time on a 0-100 scale, which is handy for spotting peaks and declines, though it doesn't give raw search counts.
Keyword research in practice
Picture a hypothetical local bakery that wants more online cake orders. Its seed list might be "custom cakes," "birthday cakes" and "wedding cakes," plus the city name.
Expanding those seeds surfaces phrases like "gluten-free birthday cake near me," "how far in advance to order a wedding cake" and "cake delivery same day." Grouping shows three clear page types: transactional product pages for each cake type, a local landing page, and a helpful guide answering ordering-timeline questions.
Prioritizing might put the local landing page and the birthday-cake page first, because they map directly to orders. The ordering guide comes next, since it answers pre-purchase questions. The national term "wedding cake ideas" might be skipped. It has demand, but searchers mostly want inspiration photos, not a local order.
Best practices for keyword research
- Start with the customer, not the tool. Tools give you volumes. Customers give you language. Collect real phrasing from calls, chats and reviews before you open any software.
- Map one primary topic to one page. Decide which URL owns each keyword group. This prevents keyword cannibalization, where several of your pages compete for the same query.
- Check the live results yourself. Search the term and look at what ranks. If the top results are all videos, product listings or forum threads, a blog post may not fit what searchers want.
- Don't ignore low-volume terms. Long-tail keywords are specific, often closer to a purchase, and usually less competitive. Many of them add up.
- Weigh business value explicitly. Give each keyword group a simple score for how closely it relates to what you sell. It keeps the plan honest.
- Write naturally. Google's starter guide notes that its language matching is sophisticated, so you don't need every variation on the page. It also says keyword stuffing is against Google's spam policies.
- Revisit regularly. Language, products and search features change, and Search Console will surface new queries every month.
Common keyword research mistakes
Chasing volume alone. Big numbers are tempting, but broad head terms are often vague in intent and fiercely contested. A page that ranks for a smaller, precise query can bring more qualified visitors.
Treating tool numbers as exact. Search volumes and difficulty scores are estimates. Different tools often disagree, sometimes by a lot. Use them to compare options, not to forecast precise traffic.
Creating a page for every variation. "Best CRM for startups" and "top CRM for startups" don't need separate articles. Splitting them dilutes your effort and can confuse search engines about which page to show.
Skipping intent. Targeting a "what is" query with a sales page, or a "buy" query with a long essay, usually fails no matter how well optimized the page is.
How Butter approaches keyword research
At Butter, we treat keyword research as the foundation for everything else, including GEO. We start with the questions real customers ask, then check how those questions show up both in Google results and in AI assistants through our monthly AI prompt testing. That tells us which topics need a page, which pages need tightening, and where a client is missing from answers entirely.
On our GEO+SEO plan, that research directly drives the eight articles we publish each month and the on-page optimization we do on existing pages. Search Console data in the Butter Analytics dashboard helps us spot new queries as they appear. If you want to see how this fits into a broader program, our SEO services page has the details.
Frequently asked questions
How often should I do keyword research?
Do a thorough round when you plan a site or a new content program, then refresh it a few times a year. Check Search Console monthly for new queries you already appear for. Major product launches or market shifts are also good triggers for a fresh look.
Can I do keyword research for free?
Yes. Google Search Console, Google Trends, Google Keyword Planner (inside a Google Ads account), autocomplete and People Also Ask all cost nothing. Paid tools add bigger databases, difficulty scores and competitor data, which save time but aren't required to get started.
Does keyword research still matter with AI search?
Yes. AI answers are built from searches and sources, and Google says its AI features use the same SEO fundamentals as regular search. Keyword research tells you which questions and subtopics your audience cares about, which is exactly what you need to cover to be useful and citable.
How many keywords should one page target?
One primary topic per page, plus the close variations and related questions that share the same intent. A strong page often ranks for many related queries at once. If two keywords need different answers, they usually need different pages.