doorway pages, and why a thousand landing pages backfire.

Doorway pages are pages built to rank for many similar searches while sending every visitor to the same place or showing near-identical content. Google’s spam policies list them as spam. A site with a thousand landing pages that differ by one swapped word will usually see most of them left out of the index, and the pages that are indexed compete with each other. More URLs do not mean more reach.

This article covers what these pages look like, how to test a page before you build it, what to build instead, and how to clean up if you already have them.

what counts as a doorway page

Google describes doorway pages in its spam policies as pages or sites created to rank for specific, similar queries. They funnel users to one destination, or they offer content that is nearly the same on every page. The visitor gains nothing from landing on one of them rather than another.

These are the common forms:

• City-swap location pages. One page per city name, with the same paragraphs and only the place changed.
• Keyword-swap variants. One template that produces a page for a tool for an audience, repeated for every audience you can list.
• Size and format matrices. Every combination of a few options, generated from one template, each with the same steps and the same answers.
• Near-duplicate comparison pages. One page per competitor, where only the competitor’s name changes.

The common thread is that a template did the writing and the query was the only input.

why they fail

There are three problems, and they stack.

First, Google treats them as spam. A page made to catch a search, rather than to help the person who made it, breaks the policy even when the tool or service behind it is real. A separate policy on scaled content abuse covers large numbers of low-value pages made mainly to rank, however they were produced, including with AI.

Second, thin near-duplicates tend to be left unindexed. Google finds the page, sees that it adds nothing the others do not, and moves on. A large share of unindexed pages can also drag on how the site as a whole is judged.

Third, pages that target the same intent compete with each other. This is cannibalisation. Instead of one strong page, you get several weak ones swapping positions, and none of them wins.

the test to apply to any planned page

Ask three questions before a page gets a URL:

1. Does it serve a distinct need, with content a person would want? Not a distinct keyword. A distinct need.
2. Would you build it if search did not exist? If the only reason for the page is the query, it is a doorway.
3. Could you write it by hand? If you could not write each page separately and still have something different to say on each, a template is hiding that there is nothing there.

One “no” means the page should not exist, or should be merged into one that does.

what to build instead

Build one page per real intent, and write it by hand. A page per location is right only when you genuinely operate there and have something specific to say about it: how the work is done, who it is for, what is different. Hubs and category pages belong only when there is enough real content behind them to justify the page.

how we apply this on our own product

This is how we apply it on our own product, PDFOak. It has 22 tool pages. Each is one real job with a working tool behind it, and each has a hand-written article. Routing, metadata, canonical tags, structured data, breadcrumbs, the sitemap and related links all come from a registry. The writing never does. A tool without a hand-written article and a working tool cannot get a URL, because the build refuses.

We deliberately did not build an auto-generated format matrix, keyword-swap or size-swap variants from one template, location pages, or a target such as “1000 tools”. We expect the site to grow to roughly 40 to 60 tools, one per real job.

Seven compression pages target different file sizes, which is exactly the kind of set that could turn into one template repeated. We rewrote their text until the similarity between any two pages fell from 0.74 to under 0.01, measured by word-sequence overlap. They share no steps, limitations or FAQs, and a test enforces it.

Guide category pages are generated only when a topic has three or more guides, so a thin archive page never becomes an indexable URL.

how to clean up if you already have them

1. Inventory. List every page built from the template, with its indexing status, impressions and clicks from Search Console.
2. Measure similarity. Compare the text of the pages. Pages that overlap heavily are the candidates for removal.
3. Use noindex as a holding step. It takes weak pages out of the index while you decide, without breaking any links.
4. Merge or remove. Fold overlapping pages into one stronger page and redirect the old URLs to it. Remove pages that serve no need.
5. Keep what earns traffic. A page that already gets clicks has shown it serves someone. Improve it, do not delete it.
6. Rewrite the rest by hand. Give each remaining page its own purpose and content, or retire it.

Then track indexed pages and impressions over the following months, as described in what an SEO agency should report.

Our website SEO service starts with this kind of audit: which pages deserve to exist, and which are costing you.