SEO Keyword Research

You find the searches that the best-fit customer actually makes, judge which ones the site can win, and map each group of queries to a single page. You do it without paid SEO suites, so you are explicit about where every number comes from and how rough it is. The output is a keyword map a writer can act on, not a long list of terms.

When to use

• "What keywords should we target?"
• "What should we write about to get found on Google?"
• "Why do we get no organic traffic?"
• "Find quick SEO wins."
• Planning a new section, product page or content cluster.

What you need

• Business context: products, best-fit customers, priority markets and languages, what a lead or sale is worth.
• Google Search Console for the site: connected app if available, otherwise the owner's signed-in browser. This is the most valuable free source: real queries, impressions, clicks and positions.
• Optional, free with an account: Google Keyword Planner (inside Google Ads; often shows volume ranges rather than exact numbers when the account has little spend), Bing Webmaster Tools keyword research, Google Analytics (connected app).
• Customer language: sales calls, support tickets, reviews, on-site search logs (ask_teammate for sales and support).
• If you have no account access, work from public sources and say so; ask via clarify for exports.

Method

1. Seed list from customers, not tools. Write 10 to 30 seed topics from the product, the problems it solves, the alternatives, and the words customers use in tickets and reviews. Include problem words ("month end takes too long"), solution words ("multi-currency bookkeeping"), and comparison words ("X vs Y", "X alternative").
2. Mine Search Console first. Export queries for the last 3 to 12 months with clicks, impressions, CTR and position. Pull out:
   • Striking distance: queries at average position 8 to 20 with meaningful impressions. Improving the existing page is often the quickest win.
   • Low CTR at good positions: position 1 to 5 with CTR well below the site's own average for that position. Usually a title or snippet problem.
   • Unexpected queries: topics people already find you for that you have no page for.
3. Expand with free sources. Use your browser modestly (no automated mass querying):
   • Google autocomplete: type each seed plus letters and question words (how, what, why, best, vs, for).
   • People Also Ask and related searches on the results page.
   • Google Trends to compare relative interest and seasonality between phrasings; it shows relative interest, not volume.
   • Keyword Planner and Bing keyword research for volume ranges.
   • Forums and communities (Reddit, Quora, niche forums) for questions in the audience's words. Set the search location and language to the target market; results differ by country.
4. Classify intent. For each term: informational (learn), commercial investigation (compare), transactional (buy or sign up), navigational (a brand). Check intent by looking at the results page: if the top results are product pages, a blog post will struggle, and the reverse.
5. Cluster by shared results. Two queries belong on the same page when their top 10 results overlap heavily. A working rule: 3 or more shared URLs in the top 10 means one page; 0 or 1 means separate pages. Check the important clusters by hand in the browser. Do not create separate pages for plurals and close variants.
6. Judge ability to win from the results page. For each cluster note who ranks: big brands and government sites (hard), established niche sites (medium), forums, thin pages, outdated posts or user-generated content (opportunity). Note results features (map pack, video, shopping, AI answers) that push organic links down.
7. Score and prioritise with the sheet in references/keyword-map-template.md: business value 0 to 3, demand (from the best source you have), ability to win 0 to 3, and whether a page exists. Business value is weighted heaviest: a low-volume query from buyers ready to choose often beats a high-volume curiosity query.
8. Map clusters to URLs. For each priority cluster: the target page (existing URL or new), primary query, secondary queries, intent, format the results page rewards, and the internal pages that should link to it. One cluster, one page; flag existing pages competing for the same cluster.
9. Hand over. Write the map as a sheet, put the top 10 opportunities on a show_card, and pass priorities to the content plan (see also editorial-calendar-planning). Store the priority clusters in memory.

Output

• Keyword map (CSV or sheet) with the columns in references/keyword-map-template.md.
• Top 10 opportunities with the reason and the recommended action (new page, improve page, rewrite title).
• Quick wins list from Search Console (striking distance, low CTR).
• A sources note: which tool gave which numbers, the date, and the location setting.

Checks before you finish

• Every volume figure states its source (Search Console impressions, Keyword Planner range, Trends relative) and date.
• No invented volumes or difficulty scores.
• Intent was checked on the actual results page for each priority cluster.
• Each cluster maps to exactly one URL; competing pages are flagged.
• Priorities weigh business value, not volume alone.
• Location and language match the target market.

Pitfalls

• Volume chasing. High-volume terms with the wrong intent bring visitors who never buy. Start from business value.
• Trusting one number. Free tools give ranges and relative trends. Use them to compare, not as forecasts.
• One page per keyword variant. Close variants that share results belong together; splitting them makes pages compete.
• Ignoring what already ranks. Search Console often shows faster wins than new content.
• Skipping the results page. Intent and difficulty are visible there. Look before you plan.
• Paid tool assumptions. Do not ask the owner to buy a tool for this. If a figure needs a paid tool to get, say what you would need and work with ranges meanwhile.
• Automated scraping of search engines. It breaks their terms and gets blocked. Use the browser at a human pace and the free tools above.