What are doorway pages, according to Google?
Google's spam policies describe doorway abuse as sites or pages created to rank for specific, similar search queries, which lead people to intermediate pages that aren't as useful as the final destination. One of the listed examples reads almost like a careless location rollout: pages targeted at specific regions or cities that funnel users to one page. If you're planning a page per town, that's the example to read twice, and it's the first thing I check in the local SEO help that doesn't need an office I offer.
Doorways aren't cloaking. Google defines cloaking as presenting different content to users and search engines with the intent to manipulate rankings and mislead users. A doorway shows everyone the same thin page. Cloaking shows Google one page and people another.
What are some examples of doorway pages?
Google lists four examples in all, and each one has a version that shows up on location pages. I use them as the questions the test has to answer.
| Google's example | What it looks like on a location site | The question I ask |
|---|---|---|
| Pages for specific regions or cities that funnel users to one page | Every town page's only job is to push the reader to one contact form | Can a reader finish what they came for on the town page? |
| Pages generated to funnel visitors into the usable part of a site | Thin town pages, with the useful content sitting on one service page | Is the useful content on the town page, or behind it? |
| Substantially similar pages, closer to search results than a browsable hierarchy | Town pages only the sitemap links to, with no hub a person can click through | Can someone click from the home page to every town? |
| Multiple sites with slight variations to the URL and home page | A separate domain for each town or region | Is it one site with one section for the areas served? |
Further down the same policy page sits scaled content abuse: many pages generated mainly to manipulate rankings rather than help people, however they're made. Pages built from one template can trip both at once, which is why I test the set as a set.
How does Google spot doorway pages, and is there a penalty?
Google says it catches spam policy violations through automated systems and, where needed, human review that can end in a manual action. That manual action is what most people mean by a doorway page penalty, and Google says sites that break these policies may rank lower or not appear in results at all.
What Google doesn't publish is how the automated side works, a page count or a similarity percentage that marks the line. I won't pretend there's one. The test below checks whether each page is worth landing on, which is what the policy describes.
Rather have me do this? One person, whole job.
Book a call on UpworkWhat's the difference between a doorway page and a landing page?
A landing page in SEO is a page built for one search that a visitor can finish their task on. A doorway page is built for the search and sends the visitor somewhere else to finish. Both can target a town. The difference is what happens after someone arrives.
| Landing page | Doorway page |
|---|---|
| Answers the searcher's question on the page itself | Holds a heading and a button, with the answer somewhere else |
| Says things that are only true for that service or town | Reads the same with any town name swapped in |
| Sits in a section a person can browse to | Is reached mainly from search or the sitemap |
| Exists because people there need it | Exists because a search phrase has volume |
When is a location page legitimate?
When a person in that town would find something on it they couldn't find on the page for the next town. That gives me the test I start with: if you could swap the town name and the page would still read true, it's a doorway, however well it's written.
What passes is usually plain. Which services you really offer in that area and which you don't. How the work runs there, like collection times or travel limits. Local conditions that change the job, such as older housing or rural access. Proof that belongs to the area: work you've done there and questions people there actually ask.
What fails is anything that reads the same everywhere: a paragraph of town history, a list of neighbourhoods with no reason to be there, population figures, or the same three service paragraphs with a new name in the heading. Those facts can be true about the town without being true about your business in the town.
Service area pages sit on the same line. A business that travels to its customers can have a page per area it covers, as long as each page earns its place. The swap test sits alongside the rest of my local SEO checklist, and it's the one that decides whether a town page gets built at all.
The business has to really serve the town. A page for a town it doesn't cover is a doorway on day one, whatever's written on it.
How I test a set of location pages before launch
I run this on staging, or on a Screaming Frog crawl of the drafts, before any page is indexable. Apart from the first step it covers every page in the set, not a sample.
- Step 01
Read three pages side by side
A big town, a small one and one from the middle, read in one sitting. If the name's the only thing that changes, the set fails here.
- Step 02
Measure the shared text across the whole set
I strip the header, footer and menus, split each page's main text into overlapping runs of eight words, and compare every page with every other one. Each page gets two numbers: how much of its text no other page has, and its highest overlap with any single page. The numbers tell me which pages to read first. They aren't a pass mark.
- Step 03
Count the facts only that town can claim
Per page, I list every sentence that would be false on another town's page: services, coverage, timings, local conditions, proof. A page with none of those is a doorway at any length.
- Step 04
Check a reader can finish on the page
If every button on a town page points at the same contact page, and the answer to the reader's question isn't on the town page itself, it's working as a funnel. That's the pattern Google's city example describes.
- Step 05
Check the pages sit in a browsable section
Each town page should be reachable from a hub a person can click through, and link to its neighbours where that's useful. Pages only the sitemap knows about look more like search results than a part of the site.
- Step 06
Search the output for template leaks
I crawl the set for placeholder text, empty variables, the wrong town in a heading and two pages for one place.
- Step 07
Hold back the pages with nothing to say
Every failing page goes on a list with the reason next to it, and it stays unpublished or noindexed until it's fixed. Launching the passing pages first is fine. Launching the lot and fixing later isn't.
Don't let a page generator publish straight to the live site. Generated pages go out as a batch, and a batch with a problem in it goes out with the problem on every page.
What did 487 generated pages show about scale?
The clearest example I've got is Scrap a Vehicle's 487 generated location pages for a UK scrap car business. When I ran this test on the live set on 27 September 2026, it flagged the set instead of passing it. These are the numbers, all from the case study.
| What I measured | Result | What it shows |
|---|---|---|
| Length of the Elgin page | 1,224 words, with its own title and heading | The page is long, so length isn't the problem |
| Elgin's shared 8-word runs | More than half shared with Luton, Stoke-on-Trent and Edinburgh | A long page can still be a doorway |
| Template leaks | A placeholder page and a second Edinburgh page still live | One slip ships at scale |
| Car make pages against Birmingham | Positions 6.7 to 7.7 against 25.2 | Narrower searches beat a contested city |
| Best page by clicks | Stoke-on-Trent, 185 clicks at position 8.6 | Towns can still work when they earn it |
Three lessons come out of it. Word count on its own tells me nothing, because a 1,200 word page that shares most of its words with every other page is a long doorway. Towns aren't the only thing a business can build pages around, and a car make page faces fewer pages built for it than a big city. And the template leak step exists because on a generated set one slip lands on hundreds of pages.
None of this makes 487 a safe number or a risky one, because Google's policy doesn't count pages.
What do I do with pages that fail the test?
Each failing page gets one of these, decided page by page. It's the same call as deciding which thin pages to improve, merge or remove, just made before launch instead of after.
- i.
Add what's missing
If the business really serves the town and there's something true to say, the page gets the local facts, services and proof it lacked. Then it goes back through the test.
- ii.
Merge small towns into a regional page
Five towns with nothing different to say often make one honest page about the area they sit in. That page names the towns it covers and says what's the same across them and what isn't.
- iii.
Keep it out of the index until it's ready
A noindex keeps the page live for anyone who needs it while it stays out of search. It's a holding state, not a fix, so each one gets a date to be looked at again.
- iv.
Remove it with the right status
A page for a town the business doesn't serve comes down. If Google has indexed it, a 404 or 410 says it's gone, and a 301 only points at a page that replaces it, like the regional page.
- v.
Fold two pages for one place into one
Two pages for the same town compete with each other. I keep the stronger one, 301 the other into it, and check which one Google was showing. My cannibalization checker reads that straight from your own Search Console data.
If you've got a set of location pages waiting to go live, or already live and earning nothing, I'll run this test on the whole set remotely and give you a verdict for every page: publish, fix, merge or hold back.
Book a call on Upwork
