Most keyword tools group keywords that look alike. RankDots groups keywords that Google already treats alike — by comparing who actually ranks for each one. When the top results for two keywords overlap heavily, one page can win them both. When they do not, one page cannot, however similar the words look. That single difference changes what a “page” in your plan means.
The two approaches
Grouping by words
The common approach reads the keywords themselves: shared terms, word stems, or closeness of meaning measured by a language model. It is fast, cheap, and needs nothing but the list. Its assumption is that keywords which look or mean alike deserve the same page.
Grouping by evidence
RankDots instead asks Google. For each keyword, it reads the results page; when two keywords return largely the same ranking pages, Google is treating them as one search need — and one page of yours can serve both. The grouping is built from that overlap. This is how Step 3 of the method builds your pages. Step 3 — How we build pages from keywords
Why words mislead
The words fail in both directions, and both failures are expensive. The examples below are illustrative, not from any project.
Similar words, different searches
Take “email marketing” and “email marketing software”. One added word — and Google shows guides and explainers for the first, product lists and comparisons for the second. They share almost every letter and almost no intent. A words-based grouper puts them on one page; that page then explains to people who came to buy, and sells to people who came to learn, and satisfies neither. Google sees a page serving two needs and ranks it hesitantly for both.
Different words, same search
Now take “how to get rid of ants” and “ant infestation in house”. Barely a shared word — and largely the same results. Pairs like this are everywhere: two phrasings that read differently while eight of their top ten results are the same pages. A words-based grouper splits them into two pages; you write both; and your own two articles then compete against each other for one set of results, splitting whatever strength either would have had.
The results page settles it in both cases, because the results page is Google publishing its verdict. Who ranks for a keyword is not a clue about what Google thinks the query means — it is the answer. The Google top 20 column in your project shows you that verdict for any keyword, so the grouping is never something you have to take on faith.
Side by side
| Similar-words grouping | SERP-overlap grouping | |
| What it compares | The words: shared terms, stems, closeness of meaning | The evidence: which pages rank for each keyword |
| What it needs | Only the keyword list — fast and cheap | A results page per keyword — slower, data-heavy |
| Where it errs | Words that look alike but mean different searches; words that differ but mean the same one | Volatile results pages; keywords with no results data |
| What an error costs | One page serving two needs badly, or two pages fighting for one need | A grouping that can shift when results shift |
Note what the right-hand column’s weaknesses are: they are conditions in the world, not blind spots in the method. When results pages are stable — which, for the commercial and informational queries a content plan is built on, they mostly are — grouping by evidence inherits Google’s own judgement. And reading the results pages pays twice: the same evidence powers Weak spots, so the plan knows not only which keywords belong together but which of them you can win.
When the other approach is still right
An honest comparison names the cases it loses, and words-based grouping has real ones:
- No evidence to read. Brand-new phrases and deep long-tail queries can have thin, unstable or empty results pages. Where the SERP gives no signal, similarity of meaning is the only signal there is.
- Organising, not ranking. If you are structuring a content calendar, a site’s navigation or an internal taxonomy, thematic grouping by meaning is exactly what you want — nothing about it needs Google’s opinion.
- Fast triage of huge raw lists. Sorting tens of thousands of raw keywords into rough buckets before any deeper work is a job for cheap, instant grouping.
- Genuinely volatile results. News and trending topics reshuffle their results daily. An overlap measured today may not describe next week; on such queries, no grouping method is stable, and meaning is at least steady.
If your work lives in those cases, a semantic tool is not wrong — it is fit for that job. A content plan meant to rank on Google is a different job, and for that one, the evidence of who ranks beats the appearance of the words.
Comments
0 comments
Article is closed for comments.