The short version: keyword research and analysis is the process of finding what your customers type into Google, working out why they're typing it, and deciding which of those terms are worth building content or pages around. It's not a spreadsheet full of search volumes. It's a decision-making process, and most people stop halfway through it, which is why so many keyword lists sit in a drawer doing nothing.
What people think keyword research is
Most business owners I meet think keyword research means opening a tool, typing in a word, and copying down a list of related terms with numbers next to them. That's the easy 20 percent. I've watched clients spend three hours building a list of 200 keywords, feel very proud of themselves, and then never touch it again because nobody worked out what to do with it.
Real keyword research and analysis has four parts, and skipping any of them is why so many SEO efforts go nowhere: finding the terms, working out the intent behind them, judging whether you can realistically compete for them, and mapping them to actual pages or content pieces someone is going to write. Tools do the first part. Judgement does the other three.
The steps involved, in the order I do them
Here is the process, step by step, the way I run it for clients and for my own site.
- Start with the business, not the tool. Before I open anything, I write down what the business sells, who buys it, and what problem it solves. Sounds obvious. Skipping this step is the number one reason keyword lists end up full of terms nobody who buys anything would ever search for.
- Build a seed list of 15 to 25 terms. These come from the founder's own language, from customer emails, from sales call transcripts if you have them, and from your own product pages. This is the raw material.
- Expand with a keyword tool. I use Ahrefs and Google Keyword Planner side by side because they never agree with each other, and seeing the gap between them tells you something. This usually turns 20 seed terms into 300 to 500 variants.
- Sort by search intent. Every keyword falls into one of four buckets: informational (someone learning), commercial (someone comparing), transactional (someone ready to buy), or navigational (someone looking for a specific brand). Get this wrong and you'll write a blog post for a keyword where the searcher wanted a pricing page.
- Check the SERP, not just the number. Before I decide a keyword is worth targeting, I look at what's currently ranking. If the top 10 results are all government sites, Wikipedia, or huge publishers, no amount of good content is getting a small business onto page one for that term within a year.
- Score for realistic opportunity. I weigh volume against difficulty against how well it matches something the business can deliver on. A keyword with 90 searches a month that converts at 12 percent beats one with 9,000 searches that converts at nothing.
- Cluster into topics, not single keywords. Google has grouped near-identical searches for years now. One well-built page can rank for 40 to 80 related variants if you cluster rather than writing one thin page per keyword.
- Map every keyword to a specific page. This is the step almost everyone skips. A list without an owner and a destination page is just a list.
- Revisit every quarter. Search behaviour moves. What ranked in January doesn't always rank in October, especially now that AI Overviews sit above the normal results for a growing share of queries.
If you want the same rigour applied to content ideas more broadly, not just search terms, I wrote a longer breakdown of that process in how to win at content marketing with kickass content research, which covers the research side that keyword tools never show you.
A real example, and where it went wrong first
A client of mine, a small bookkeeping firm outside Reading, came to me convinced she needed to rank for "accounting software." That term gets around 40,000 searches a month in the UK. It also has Xero, Sage, QuickBooks, and a dozen comparison sites sitting on page one, all with domain authority scores she had no chance of matching within a decade, let alone a year.
We pulled her actual client list instead. Half of her customers were Etsy sellers and small ecommerce owners who hated doing their own VAT returns. So we went after "bookkeeping for Etsy sellers UK," a term with roughly 90 searches a month according to Ahrefs. Ninety. It felt like nothing next to 40,000.
That page brought in 6 new client enquiries in its first four months, and 4 of them converted into paying clients worth around 1,800 pounds a year each. The 40,000-search keyword would have brought her a comment section full of software vendors and zero enquiries, because the intent was wrong from the start. Volume without intent is just a vanity number.
The bit nobody likes admitting about search volume
Here's something worth saying plainly: most search volume numbers you're looking at are estimates, not facts. Google Keyword Planner rounds and buckets its figures unless you're actively spending on ads, and even then it's giving you an average over 12 months that can hide huge swings. Ahrefs, SEMrush, and Moz all pull from different data sources and clickstream panels, which is why the same keyword can show 200 searches in one tool and 900 in another. I've seen it happen on the same afternoon with the same keyword.
Nobody outside Google knows the real number. What you're buying when you pay for these tools isn't accuracy, it's a consistent way to compare terms against each other. Treat the numbers as directional, not gospel, and you'll make better decisions than someone who trusts the figure to the exact digit.
The second uncomfortable truth is this: a lot of agency keyword research is theatre. A client pays for a "comprehensive keyword audit," gets a 40-page PDF with 800 rows, and 95 percent of those rows never get built into anything because nobody assigned them to a page or a writer. The report exists to justify the invoice, not to drive rankings. If you're paying for keyword research, ask up front how the deliverable maps to actual content that will get written, not just how many keywords will be in the spreadsheet.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
How AI search has changed the process in 2026
The other shift worth naming: keyword research used to be entirely about ranking in 10 blue links. Now a growing share of searches get answered inside an AI Overview or a chat answer before anyone clicks through to a website at all. I covered how that's reshaping click-through behaviour in this week's roundup of AI news this week for small business, and the trend has only sped up since.
What this means for keyword research: intent still matters more than ever, but you also need to check whether a term triggers an AI Overview before deciding it's worth chasing. Some transactional and local terms still send traffic to websites reliably. Some broad informational terms get swallowed whole by the AI answer, and even a page one ranking sends almost nobody through. I check this manually now for every priority keyword before we commit content resources to it, something that wasn't part of the process even two years ago. If you want a fuller playbook on staying visible without your traffic getting eaten, I wrote about it in AI for SEO in 2026.
AI tools have also changed how the research itself gets done. I now use AI models to cluster keyword lists, spot intent patterns across hundreds of terms in minutes, and draft the first pass of content briefs, work that used to take a full day by hand. That doesn't replace judgement about which keywords matter for a specific business, but it removes the grunt work. If you're still doing this entirely by hand, it's worth the short learning curve. I wrote a practical path for that in how to become fluent in AI in 90 days.
Turning the list into pages that rank
The final piece, and the one that separates a keyword list from a keyword strategy, is turning terms into content that matches the intent you identified. A commercial-intent term needs a comparison page, not a 2,000-word think piece. A transactional term needs a page that lets someone buy or book, with the friction removed, not another blog post. A cluster of informational terms around one topic can often live on a single detailed guide rather than ten thin articles competing with each other.
I'd rather see a business target 15 well-chosen keyword clusters, each built into a strong page, than 150 keywords scattered across shallow content that never earns a top-10 position for anything. Depth on fewer terms beats breadth on many, every time I've tested it.
Frequently asked questions
What is the difference between keyword research and keyword analysis?
Keyword research is the finding stage, building the list of terms people search for around your topic. Keyword analysis is the judging stage, working out intent, difficulty, and whether a term is realistically worth targeting. Most people do the first and skip the second, which is why so many keyword lists never turn into rankings.
How many keywords should a small business target?
Fewer than you think. I'd rather a business commit to 15 to 30 well-researched keyword clusters built into strong, specific pages than scatter effort across 200 loosely related terms. Depth on fewer terms consistently outperforms breadth on many in the results I've tracked.
Are free keyword tools good enough, or do I need to pay for Ahrefs or SEMrush?
Google Keyword Planner and Google Search Console are free and good enough to start, especially for a business with an existing site and some traffic already. Paid tools become worth the 99 to 199 dollars a month once you're managing multiple sites or need to see competitor gaps you can't get anywhere else.
How often should keyword research be redone?
Every three to four months for an active site, and immediately after any major shift in your product line or market. Search behaviour changes faster now that AI answers sit above traditional results for a growing share of queries, so a keyword list built a year ago is worth checking, not trusting blindly.
