This keyword research checklist is the sequence I run on a new site, in order, start to finish. It takes an afternoon for a small site and a couple of days for something with existing traffic. The output is not a spreadsheet of 2,000 rows — it is a ranked list of pages to build, with a reason attached to each one.
Skip steps at your own risk, but especially do not skip the SERP check. Most wasted content budgets die there.
Steps 1–2: Build the seed list from your business and your data
1. Write down what you sell before opening a tool
List your products, services, and the problems each solves, in the words a customer would use rather than your internal names. Add the questions your sales or support team answers weekly. Twenty to forty lines is plenty. This beats anything a tool generates because it is pre-filtered by what you can actually deliver — tools expand seeds, they cannot invent your business model.
2. Pull what you already rank for
Open Google Search Console, set the range to the last three months, and export your queries. Sort by impressions with a position between 8 and 30. Those are pages Google already considers relevant but has not committed to, and they are the cheapest wins on the entire site. You will also find queries you never targeted, which tell you how the market describes your category.
Step 3: Expand with multiple seeds at once
Single-seed research gives you a narrow cluster and a false sense of completeness. Run five or six seeds together — a product term, a problem term, a competitor category name, a buyer job title, and a “how to” phrasing — and the overlap between them shows where demand actually concentrates.
SEO Rocket’s explorer returns up to 150 ideas per search with volume, difficulty, CPC, global volume, and SERP features, and supports multiple seeds in one query. Whatever tool you use, confirm one thing first: it must be querying the country index you sell into. A Singapore or UK business pulling US volume gets numbers that are wrong in both directions.
Step 4: Mine competitors for gaps
Pick three to five direct rivals — the ones actually on page one for your money terms, not the biggest brands in the industry. Run a content gap analysis to see which keywords they rank for and you do not.
Read that output with a filter in mind. Most gaps are irrelevant: their blog covers adjacent topics you have no business writing about. What you want is the subset where two or more competitors rank and you are absent, because agreement between rivals is a decent signal that the query converts. SEO Rocket’s content gap runs across up to five competitors with per-rival position columns, which makes the overlap obvious at a glance.
Step 5: Classify every keyword by intent, then sort by CPC
Sort the combined list into four buckets, because each requires a different page and produces different revenue:
- Transactional — “buy”, “pricing”, “cost”, “quote”, “near me”. Highest value, lowest volume.
- Commercial investigation — “best”, “vs”, “alternative”, “review”, “for [industry]”. This is where buyers decide.
- Informational — “how to”, “what is”, “why”. Builds authority, converts slowly.
- Navigational — brand names. Track your own, ignore everyone else’s unless you are writing a comparison.
Then sort by CPC within each bucket. Advertisers do not sustain high bids on queries that never produce customers, so CPC is the cheapest available proxy for commercial value. A term at 200 searches and $18 CPC usually beats one at 6,000 searches and $0.40.
Step 6: Read the actual SERP for your top candidates
This is the step people skip, and it is the one that saves quarters. For every keyword you seriously intend to target, open the results and look.
- What content type ranks? If page one is nine product category pages, a blog post will not rank no matter how good it is. Match the format.
- Who holds the weakest slot? Look at positions eight through ten, not the average. If those are thin pages, forum threads, or outdated posts, the query is available.
- How much room is left? Count the ads, map pack, AI Overview, and featured snippet above the first organic result. On some queries, position one is below the fold.
- Is intent what you assumed? Plenty of terms look commercial and return purely informational content. Google’s result set is the definitive statement of what a query means.
Fifteen minutes here regularly disqualifies half a shortlist. That is the checklist working.
Steps 7–8: Score difficulty honestly and cluster into pages
7. Difficulty is relative to you, not absolute
Difficulty scores are modeled estimates, usually driven by link counts to the ranking pages. Useful for sorting, useless as a verdict. Check referring domains to the specific ranking URL rather than the whole domain. Many page-one results have almost no page-level links and rank on domain authority alone, which means a genuinely better page plus a handful of relevant links can displace them. Links are necessary but not sufficient; thin content loses even with links behind it.
8. Cluster keywords into pages
One page can rank for dozens of related queries. Group terms sharing the same intent — “seo audit cost”, “how much does an seo audit cost”, and “seo audit pricing” are one page, not three. Publishing three creates cannibalization, where your own pages split signals and none ranks well. Pick the highest-volume term in each cluster as the focus keyword and treat the rest as section headings.
Step 9: Sequence the queue and store it where it gets used
Order the build queue: transactional pages first, comparison and alternative content second, informational last. Teams that do this backwards publish sixty blog posts, triple traffic, and see no revenue change.
Then save the shortlist into a project keyword pool rather than a folder of CSVs. The point is that the same list feeds drafting and rank tracking without anyone retyping it — research that dead-ends in a download folder gets tracked at maybe a third of its actual size, because the copying is tedious and people quietly stop doing it.
Re-running the checklist
Repeat steps two, four, and six quarterly. Search Console tells you what changed for you, the content gap tells you what competitors added, and the SERP check tells you whether your assumptions still hold. Volume estimates are modeled twelve-month averages, so they lag real shifts by months — your own impression data catches trends first, every time.
Expect the list to churn. Roughly a fifth of what you shortlist in January will look wrong by June, because competitors publish, SERP layouts change, and AI Overviews absorb informational queries that used to send clicks. That is not a flaw in the process; it is the reason the process is quarterly rather than annual.
That is the whole keyword research checklist. If you want research, writing, and tracking sharing one keyword pool instead of three tools and a spreadsheet, SEO Rocket does that for a flat US$50 a month. The process above works regardless of what you run it in.
