Straight answer: keyword research doesn’t tell you what to write, it tells you what people already type into a search box, which words they use to describe their problem, and roughly how many of them there are. It answers demand, phrasing, and competition. It cannot answer whether your business should exist, whether your offer is good, or what your customers need from you next. Confuse those two things and you’ll build a very well-optimised page nobody wanted.
Three questions it answers
After twenty-odd years doing this for clients across the UK, US and Israel, I’ve stopped treating keyword research as a magic strategy tool. It’s a research method with a narrow job. Done well, it answers exactly three things.
1. What words do people use, not what words I think they use
I once had a client, a bookkeeping firm in Manchester, who insisted on the phrase “financial reporting services” across their homepage. Their keyword tool showed that phrase getting 40 searches a month in the UK. “Management accounts template” got 880. “How to read a profit and loss statement” got 1,300. Their own customers weren’t searching in industry jargon, they were searching like confused small business owners, because that’s what they were. The gap between what a company calls itself and what a searcher types is the single biggest thing keyword research exposes, and most businesses never check it.
2. How much demand exists before I commit time to it
This is the practical, boring, useful bit. If a keyword shows 20 searches a month in the UK and 30 in the US combined, and it would take you three weeks to build the authority to rank for it, that’s a bad trade. I use tools like Ahrefs and Semrush side by side because their volume estimates disagree by 30 to 50 percent depending on the keyword, and averaging two sources gives me a steadier number than trusting either one blind. Volume tells you the size of the room before you decide whether it’s worth walking into.
3. Where the competition is weak enough for a smaller site to win
Keyword difficulty scores are rough, but they’re useful for triage. A term with 5,000 monthly searches and a difficulty score of 70, dominated by the BBC, Which? and three national comparison sites, is not a fight worth having for a two-person consultancy. A term with 300 searches and a difficulty score of 12, where the top result is a five-year-old forum thread, is worth an afternoon. Keyword research is really a shortlisting exercise, not a strategy in itself.
What it cannot answer, no matter how good the tool is
Here’s the bit most guides skip past. Keyword volume is backward-looking. It tells you what people have already searched, using words that already exist for a problem people already know they have. It cannot tell you about a problem your audience hasn’t named yet, a new product category, or a shift in how people talk about something. When AI overviews and chat-based search started pulling traffic away from classic blue links, plenty of keyword tools kept showing “stable” volume on terms that were quietly losing clicks to zero-click answers. The number was right and the insight it implied was wrong. That’s an uncomfortable thing to sit with if you’ve built your whole content plan around a spreadsheet of search volumes, but it’s true, and I’ve watched it happen to my own traffic on terms I’d ranked for since 2019.
This is exactly why I never let keyword data run alone. I pair it with what the searcher wants when they type that phrase, because two people can type the identical five words and want completely different things. “Best CRM for small business” can mean “I want a listicle to compare,” or “I want to buy right now,” or “I’m a student writing an essay.” The keyword alone never tells you which. You find that out by reading the top ten results and noticing what format wins.
A short story: the client who ranked for the wrong reason
A SaaS client I worked with in 2023 ranked position two for “project management software for agencies,” a term with roughly 2,400 monthly searches in the US. Brilliant, right? Except their trial-to-paid conversion from that page was under 1 percent, while a scrappier page targeting “how to bill clients for project overruns,” at 90 searches a month, converted at 11 percent. The high-volume keyword answered a question they wanted answered. The low-volume keyword answered a question a buyer with a credit card in hand had. We shifted their content calendar to chase specific, painful, low-volume questions instead of broad category terms, and revenue from organic content nearly doubled in five months without a single extra visitor on the big page. The lesson: keyword research answers “how many people search this,” not “which of these searchers will pay me.”
The step-by-step I run each month
This is the process I use with clients, stripped down to the parts that matter:
- Pull 15 to 20 seed terms from what clients ask us in sales calls and support emails, not from guesswork.
- Run each seed through a keyword tool to surface the 10 to 15 related phrases it suggests, including the “questions” and “also asked” tabs, which surface the exact wording searchers use.
- Note volume, difficulty, and who’s ranking in the top five right now, checking whether they’re big brands, forums, or thin content you could outdo in a week.
- Group the list by intent, not by topic. Buying-intent terms go into one bucket, learning-intent terms into another. This is where proper content research earns its keep, because it tells you what to build for each bucket, a comparison page versus a how-to guide versus a calculator.
- Pick the smallest viable win first, usually something under 200 monthly searches with weak competition, and prove the process works before scaling it up.
That last step matters more than people admit. I’d rather rank number one for ten small terms in six weeks than start a year-long campaign for one huge term with no proof it’ll ever land.
Where keyword research meets the rest of your site
Once you’ve got your list, the next decision is structural: does each keyword deserve its own page, or should several cluster onto one? I’ve seen sites split “email marketing tips,” “email marketing strategy” and “email marketing best practices” into three thin, near-identical pages that cannibalised each other in the rankings for a year before anyone noticed. If you’re facing that choice, it’s worth reading through how to structure a page around several related keywords rather than building a new page for every phrase your tool spits out.
You also need somewhere to put the answers once you’ve found the questions. Whether that’s a blog post, a video script, a downloadable template, or a short explainer graphic depends on the format already winning for that term, and on what your team can produce well and consistently. I keep a running shortlist of tools that make producing that content less painful, because the best keyword list in the world is worthless if it sits in a spreadsheet nobody turns into a page.
The uncomfortable bit about volume numbers
Search volume numbers are estimates, not facts, and every tool rounds and models differently. I’ve pulled the same keyword from Ahrefs, Semrush and Google’s own Keyword Planner in the same week and got three different figures, sometimes 20 percent apart, occasionally double. None of them are lying, they’re all sampling clickstream and search log data differently and smoothing it over twelve months. Treat volume as a rough size, small, medium, large, rather than a precise forecast, and you’ll make better decisions than someone chasing the third decimal place of a number that was never that exact to begin with.
What good keyword research feels like in practice
When it’s working, keyword research feels less like data analysis and more like eavesdropping on your market. You start noticing patterns: the phrases customers repeat on sales calls turning up in the “people also ask” box, the same confused question appearing in your support inbox and your keyword tool in the same week. That overlap is the signal worth chasing, far more than any single volume number. It’s why I still do this manually for my own site rather than fully automating it, because the pattern recognition matters more than the spreadsheet.
Pair that with the free competitor-traffic tools worth having open while you research.
Frequently asked questions
Does keyword research tell you what content to write?
No, it tells you what words people search and roughly how often. What to write comes from combining that with intent, what format already ranks, and what your business can credibly answer better than the current top results.
How accurate are keyword search volume numbers?
Treat them as estimates within a range, not exact counts. Different tools commonly disagree by 30 to 50 percent on the same term because they model clickstream data differently, so use volume to judge rough size rather than precise traffic forecasts.
Should I chase high-volume or low-volume keywords first?
Start with low-volume, low-competition terms tied to a clear buying question, since they convert faster and prove your process works before you invest months chasing a big, competitive term that may never pay off.
Can keyword research replace understanding my customers?
No. It shows you the words strangers type into Google, not why your specific customers buy. The best keyword lists come from combining tool data with real questions pulled from sales calls, support tickets, and customer conversations.
