Good seo keyword ideas come from three places: what people already find you for, what your competitors rank for and you do not, and what your customers say in their own words. Everything else is a variation on those. The generating part is easy — any decent tool will hand you 150 ideas from one seed. The filtering is the actual skill, and it is where most keyword lists die.
Start with the sources, then the filter, because a list you cannot triage is worse than a shorter list you can act on.
Source one: Google Search Console
This is the highest-value source you already own, and most sites underuse it. Open the Performance report, set the range to 12 months, and sort by impressions descending with clicks near zero. Those are queries where Google already shows you and nobody clicks — usually because you rank on page two or three, or because the page that ranks is not quite about that query.
Every one of those is a validated idea. Google has confirmed there is demand and that it considers you relevant. Two moves follow: strengthen the existing page for terms where you sit at positions 11 to 20, and build a dedicated page for terms where the ranking page is only tangentially related. This is the closest thing to free traffic in SEO, and unlike third-party volume figures, this data is Google’s own.
Source two: competitor gaps
Pick three to five sites that actually appear in your target SERPs — not the rivals your sales team names, the ones Google keeps showing. Run a content gap analysis and you get the terms where several of them rank and you do not, with each rival’s position per keyword.
The best entries are the ones where four out of five competitors rank. That overlap is a demand signal your market has already validated, and it usually maps to a real buying question rather than a vanity term.
Source three: seed expansion
Take five to ten terms describing what you sell and expand each one. A keyword explorer returns matching terms, related terms, and question variants with volume, difficulty, CPC, global volume, and SERP features attached. SEO Rocket returns up to 150 ideas per search on country-specific indexes, with filtering, sorting, and CSV export free after one query, and lets you save the keepers into a project pool that feeds the writer and the rank tracker.
One practical note on country indexes: querying the US index for a site serving Singapore or the UK produces near-empty, misleading data. Set the market to match your actual audience before you judge any volume number.
Source four: the SERP itself
Google gives away ideas for free on every results page. People Also Ask boxes expand infinitely as you click, and each expansion is a real query. The related searches strip at the bottom of the page is a curated list of adjacent intents. Autocomplete suggestions, prefixed with question words or “vs” or “alternative,” surface phrasings you would not think of.
These carry no volume figures, which is fine. Many are low-volume long-tail terms that convert well and face almost no competition. Treat them as the raw material for supporting content around your main pages.
Sources five through nine
- Your own site search. If you have an internal search box, the query log is people telling you exactly what they expected to find and did not.
- Sales and support conversations. Recurring questions in tickets and calls are keywords in customer language, which is usually less polished and more specific than marketing language.
- Reddit, forums, and niche communities. Search your category and read how people phrase problems. The vocabulary gap between customers and vendors is where uncontested long-tail lives.
- Review sites. Complaints about competitors are comparison keywords waiting to happen — “X alternative for small teams” writes itself from a two-star review.
- Google Ads search terms report. If you run paid search, the terms that actually triggered your ads, with conversion data attached, are the most commercially proven list you will ever get.
Filtering a raw list down to a plan
A thousand ideas is not a plan. Run every list through this sequence:
- Cut anything with no commercial connection. High volume with no path to revenue is a distraction, however good it looks in a report.
- Group by intent. Informational, comparison, transactional, navigational. Each needs a different page type, and mixing them on one page serves none of them.
- Cluster near-duplicates. “Best running shoes,” “top running shoes,” and “running shoes reviews” are one page. Building three creates cannibalization and splits your own signals.
- Check the SERP for the top candidates. Look at the weakest page-one result. If it is thin and has few referring domains, keep the keyword. If page one is uniformly strong, drop it or take a long-tail variant instead.
- Score by opportunity, not volume. Commercial relevance times winnability beats raw search numbers every time.
- Sequence by capacity. If you publish four pages a month, your quarterly plan has twelve entries. Anything past that is a backlog, not a plan.
What the volume numbers are really telling you
Every third-party volume figure is a modeled estimate, usually a twelve-month average from periodic crawls. That matters in three concrete ways. Seasonal terms are flattened — a keyword with a ten-week demand window looks like a steady trickle all year. Very low-volume terms are often reported as zero when real demand exists. And two vendors will hand you different numbers for the same keyword because they model different clickstreams.
Use volume for relative sizing, never as a forecast. A keyword reported at 200 a month with clear buying intent will usually out-earn one reported at 5,000 that is pure browsing. And once a page is live, Search Console tells you the truth about what it actually gets.
Turning the list into published pages
The gap between a keyword list and traffic is production, and this is where most plans stall. For each target, build the brief from the SERP: the subtopics that appear across the top ten, the questions in People Also Ask, the format the ranking pages share, and one thing none of them offer — your data, a real price, a documented edge case.
Then write to the brief, not to a word count. Length should land wherever full coverage puts it, benchmarked against the weakest page-one competitor rather than the longest. If you are producing at volume, keep a hard quality gate before publish — minimum body length, title under 60 characters, meta description in range, a real section structure — so nothing half-finished reaches the site.
That is the loop SEO Rocket is built around: research into a saved keyword pool, drafting against a validated template with automatic repair on failures, then rank tracking with top-100 snapshots and movement deltas so you can see what the pages actually did. Expect a quarter before the picture is clear, and read trends rather than daily positions — two or three places of movement between checks is normal noise, not a result.
