SEO
Thin pages, duplicate pages, competing pages
Three findings that do not show on a single page at all and only become visible when you lay the whole website side by side:
5 min read
By Timo Wessels Published
What this is about
Three findings that do not show on a single page at all and only become visible when you lay the whole website side by side:
Thin pages have so little text that they are an answer for nobody. Duplicate pages say almost the same thing under different addresses. Competing pages aim at the same search intent — and take visibility away from each other.
On top of that comes readability: sentence length, paragraph length and the question whether the title of a page even fits what it is about.
Why it matters
Let us start with what does not matter: word count is not a ranking factor. Danny Sullivan from Google confirmed that again in September 2025. An analysis of one million search result pages shows: as soon as a page covers at least half of the semantically expected terms, sheer length practically no longer correlates with position.
So the useful rule is not “at least 1,500 words”, but: match the depth of the weakest page currently in the top 10 — and add one thing nobody else covers.
Still, real thin pages exist. A service page with four sentences, a category page without an introduction, a tag archive with two posts. They are not too short for a threshold, but too empty for an intent. The right way to deal with them is to merge or expand — not to pad.
Cannibalisation is often misunderstood. It comes from overlapping search intent, not from overlapping words. Two pages can contain exactly the same term without competing — a glossary entry and a service page do exactly that. And two pages with completely different wording can still cannibalise each other. The reliable diagnosis goes through Google Search Console: if two URLs share the impressions for the same query without one dominating, they should be merged.
Duplicates in WordPress rarely arise on purpose. They come from paginated archives, filtered views, parameter URLs and from the same page under several addresses. The reverse case is striking too: text blocks that are the same on every page — footer texts, lists of advantages, certificates. They are template, not content, and when they make up most of a page, little is left for the actual content.
And one point on AI visibility: clear, active language with short sentences and without filler words is an item of its own on the GEO checklist (Rheinwerk, Generative Engine Optimization, ch. 12.7). Text divided into understandable sections can be quoted. Text in endless sentences cannot.
How to check it yourself
For thin pages, a look at the WordPress page overview is enough: open every page and ask yourself whether it answers a question someone would actually have. If nothing comes to mind after twenty seconds, the page is thin.
For cannibalisation, use Search Console: open Performance, filter by one query, then switch to the “Pages” tab. If two or more URLs with similar numbers of impressions show up there, they compete.
For duplicates a sample helps: copy a characteristic sentence from a page and search for it on Google with site:yourdomain.com "your sentence". More than one hit means: the text appears more than once on your website.
For readability there is no test better than reading aloud. Where you run out of breath, the sentence is too long.
What to do if it is missing
Merge instead of adding. Two mediocre pages on one topic become one good page plus a 301 redirect. That is almost always the right order — first consolidate, then optimise.
Give every page exactly one job: one main question, plus five to fifteen secondary questions and three to seven named terms, each explained in one clear sentence.
Set tag archives and internal search results to noindex if they are thin — or give them real content.
For the writing: short sentences, one thought per sentence, paragraphs of at most three sentences. Concrete beats vague. “Reduces the plugins from 20 to 12” says more than “less ballast”.
And when a text is older: maintain it instead of replacing it. For central pages a rhythm of 90 to 120 days has proven itself, with a visible update line. Important here: do not change the publication date without a real change in content.
Sources
- Consolidate or expand thin content; merge duplicate content through canonical or 301; noindex for thin parameter pages, internal search results and tag archives; category pages with their own introductory text -- @ctx:atlas-seo-technical-audit@1
- Word count is not a ranking factor (Danny Sullivan, September 2025); from 50 % coverage of the semantically expected terms, length practically no longer correlates with position; process rule “match the depth of the weakest top-10 page and add one thing nobody covers”; cannibalisation is an overlap of search intent rather than of wording; GSC diagnosis through shared impressions without dominance as the method recommended by Mueller; one page, one main query, 5–15 secondary queries, 3–7 named entities; update rhythm of 90–120 days for central pages; do not adjust datePublished without a real change -- @ctx:atlas-seo-practice-2026@3
- Divide content into understandable sections; back facts with sources; clear, active language with short sentences without filler words; make currency visible -- @ctx:atlas-seo-rheinwerk-generative-engine-optimization-12-07-the-overall-checklist-geo-at-a-glance@1 -- Rheinwerk, Generative Engine Optimization, ch. 12.7
- Write concretely rather than vaguely, one thought per sentence, paragraphs of at most three sentences -- @ctx:projects-websites-dhs-docs-website-context-voice@1