Key Takeaways
- Keyword research is finding the words and questions your customers type into Google (and ask AI), measuring how many people use them and how hard they are to rank for, and deciding which page should win each one.
- Intent beats volume. A term searched 500 times a month by people ready to hire beats one searched 50,000 times by students writing essays.
- The output is a map — one primary term and its cluster of variations per page — not a list. Two pages chasing the same term compete with each other.
Keyword research is the process of discovering the exact words and questions your potential customers use when they search, measuring how much demand each one has and how hard it will be to rank for, and then assigning each worthwhile term to the page that should answer it. It is the foundation of on-page SEO and content strategy alike: without it you are guessing what people search for, and businesses guess wrong far more often than they expect — usually by using their own vocabulary instead of their customers’.
The Four Numbers That Matter
| Metric | What it tells you | How to use it |
|---|---|---|
| Search volume | Average monthly searches for the exact term in a country | Prioritize, but do not worship — volume includes people who will never buy |
| Intent | Why someone searches: to learn (informational), to compare (commercial), to buy or act (transactional), to reach a specific site (navigational) | Match the page type: guides for learning, service and comparison pages for commercial and transactional |
| Difficulty | How strong the pages already ranking are (usually a 0–100 score based on their backlinks) | Newer or smaller sites should start where difficulty is lower and climb |
| Cost per click (CPC) | What advertisers pay for a click on the term | A proxy for commercial value: high CPC means the click is worth money to someone |
A term with modest volume, commercial intent, moderate difficulty and a high CPC is usually worth more to a business than a giant informational term. This is the single most common mistake in DIY keyword research: chasing the big number.
The 6-Step Keyword Research Process
From seed terms to a page map
- List seed terms. Your services, products and the problems they solve, in the words customers use on calls and in emails — not your internal jargon.
- Expand. Put each seed into a keyword tool and Google itself (autocomplete, "People also ask," related searches) and collect every variation and question. Expect hundreds per seed.
- Get the numbers. Volume, intent, difficulty and CPC for your target country. Export everything — you will filter later.
- Filter for fit. Remove terms you cannot honestly serve (wrong product, wrong geography, job-seekers, students). Keep a short list of commercial terms and a longer list of questions.
- Cluster. Group terms that mean the same thing — "custom ERP development," "custom ERP software development company," "bespoke ERP" — into one cluster with one primary term. One cluster, one page.
- Map to pages. Commercial clusters go to service pages; question clusters go to blog posts that link back to the service page. Note which clusters have no page yet — that is your content plan.
A Worked Example From Our Own Research
When we planned the US pages on this site, we exported roughly 125,000 keyword rows across ERP, SEO and web development. The raw list was dominated by terms like "erp" (74,000 searches a month) and "web development" (1.5 million) — huge, informational and impossible to win. Filtering for intent and fit changed the picture completely:
- "seo service" — 12,100 searches, commercial intent, difficulty 34, CPC around US$14. A buyer’s term with a realistic difficulty: it became the primary target for our US SEO services page.
- "custom web development" — 3,600 searches, difficulty 23. Small but precisely our buyer: mapped to the website development page.
- "what does erp stand for" — 8,100 searches, purely informational. Not a service-page term; it strengthened an existing explainer instead of getting its own page.
- "how to make a website mobile friendly" — 33,100 searches, difficulty 50, informational. Worth a dedicated guide because the people asking it own websites that are not mobile-friendly, which is a problem we solve.
Notice the pattern: the winners were rarely the biggest numbers. They were the terms where the searcher’s problem and our service were the same thing.
Long-Tail Keywords and Questions
Long-tail keywords are longer, more specific phrases with lower volume each — "erp for small manufacturing business" instead of "erp." Individually small, collectively they usually outnumber the head terms, they are far easier to rank for, and the intent is clearer. Question phrasings ("how much does an erp system cost") are the most valuable long tail of all in 2026, because they are exactly what people type into AI assistants and what AI Overviews answer. Every question your customers ask deserves a page or a section that answers it directly.
Keyword Research for AI Search
Traditional tools measure Google queries; people ask AI engines in longer, more conversational sentences. The fix is not a new tool but a different reading of the data: treat every question cluster as a prompt, answer it in the first paragraph of the page, structure the rest so a machine can quote it, and add FAQ schema. Then test the questions yourself in ChatGPT, Perplexity and Google’s AI Mode and record who gets cited — that is your AI-visibility baseline.
Common Keyword Research Mistakes
- Using your own vocabulary. You sell "enterprise resource planning solutions"; customers search "inventory software for small business."
- Chasing volume. The biggest terms are informational and dominated by giants.
- One keyword per page, literally. Pages rank for dozens of variations; target the cluster, not a single string.
- Two pages, one term. They cannibalize each other; merge them.
- Doing it once. Search behavior shifts; re-run the research yearly and whenever you add a service.
Keyword research is step two of every engagement we run — after the audit and before a word is written. It feeds the content strategy and the on-page work described in on-page vs. off-page SEO.
Frequently Asked Questions
What are SEO keywords?
The words and phrases people type into search engines that you want your pages to appear for. In practice each page targets a cluster of related terms — a primary keyword plus its natural variations and questions — rather than a single string.
What is a good keyword to target?
One your customers actually use, with intent that matches what the page offers, difficulty your site can realistically beat, and enough volume to matter. For a business, a modest-volume commercial term usually beats a high-volume informational one.
How many keywords should a page target?
One primary term and its cluster of variations — often 10–30 closely related phrases and questions. Trying to target unrelated terms on one page dilutes it; use separate pages for separate topics.
Are free keyword research tools good enough?
For a start, yes: Google autocomplete, "People also ask," Google Trends and Search Console reveal a lot. Paid tools add reliable volume, difficulty and CPC data at scale, which matters once you are planning more than a handful of pages.
How often should keyword research be redone?
Review it yearly, whenever you add a service or enter a new market, and after major shifts in search behavior — the growth of AI answer engines is one such shift, favoring question-style research.
Want this done for your site, not just explained?
Book a free strategy call. We look at your real search data together and tell you honestly where the leverage is — and what we would leave alone.
Book a Free Strategy Call