Short answer, as of October 2026: a location page is legitimate when it contains information that is true only for that place — a real address or defined service area, local staff, local detail, local proof — and it becomes a doorway page when the only thing that changes from one page to the next is the place name. Google's spam policies name this pattern directly, and sites built on large city-by-service matrices tend to lose those pages together rather than one at a time.
Local businesses and multi-location brands are told, correctly, that they need a page per location. Many then take the instruction further than it goes: a page for every town in the region, then a page for every service in every town. The arithmetic is seductive. Thirty towns and ten services make three hundred pages from one template and an afternoon's work.
It also makes three hundred pages that say the same thing. This article is about recognising the difference and about what to do if you are already on the wrong side of it.
What the policy actually says
Google's spam policies describe doorway abuse as creating sites or pages to rank for specific, similar search queries, leading users to intermediate pages that are not as useful as the final destination. Among the examples it gives are multiple pages targeted at specific regions or cities that funnel users to one page, and pages generated to rank for similar queries without each offering distinct value.
Two points are worth drawing out. The policy is about purpose and usefulness, not about page count; a brand with two hundred real branches can have two hundred location pages. And the test is whether each page is useful in itself, not whether it is technically unique. Rewording a paragraph so that it passes a duplicate-content checker does not change what the page is for.
Why matrices fail the way they do
A site with a few genuinely local pages and a large matrix of generated ones usually looks healthy for a while. The generated pages get indexed. Some collect impressions. A handful even earn clicks.
Then one of two things happens. Either the pages are quietly never shown — indexed, technically present, receiving nothing — or a quality reassessment demotes them as a group. The second is the dangerous one, because the reassessment is rarely confined to the thin pages. A domain where most of the URLs are low-value can see its good pages affected too.
The pattern to look for in your own data is a set of URLs sharing a template that have impressions but no clicks over a long period, followed by a step change in impressions across the whole template in a single week. A drop like that on pages nobody edited is not caused by anything you changed. It is a reassessment of the pattern.
There is a newer cost as well. AI assistants retrieving a page for a local question want a passage that states something specific. A templated page offers the same generic passage as its three hundred siblings, so there is nothing to prefer about any one of them. Thin local pages are not merely unranked; they are uncitable.
The five-question test
Apply these to any local page. A page that cannot answer yes to at least three should not exist as a separate URL.
1. Is there a real-world anchor? A staffed address, or a service area you genuinely travel to and can define. A town you would decline a job in fails immediately.
2. Is there content that would be false on any other page? Named staff, a specific route or response time, local regulations, a landmark you work near, pricing that differs here. If you could paste the body onto the next town's page and it would still be accurate, it is not local content.
3. Is there local proof? Reviews from customers in that area, photos of work done there, a case you can describe. Proof is the hardest thing to fabricate and the most persuasive thing to include.
4. Would a customer in that town choose this page over your main service page? If the main page answers their question equally well, the local page is an intermediate step — which is the definition of a doorway.
5. Has it ever earned a click? Not a ranking, a click. A page with a long history of impressions and no clicks is being shown and rejected.
What separates the two, side by side
| Element | Location page | Doorway page |
|---|---|---|
| Reason it exists | The business operates there | The keyword exists |
| Address or area | Real premises, or a defined service area | Town name inserted into a template |
| Body content | Partly specific to the place | Identical apart from the place name |
| Proof | Local reviews, photos, cases | None, or generic testimonials |
| Number per business | Matches real locations or real service areas | Matches a list of towns |
| Where it sends the visitor | It is the destination | Onward to a generic page or form |
| Behaviour over time | Earns clicks, keeps them | Impressions without clicks, then nothing |
How many local pages should you have
There is no cap, but there is a principle: the number of local pages should match the number of places where you have something specific to say.
For a brand with physical branches, that is one page per branch. For a service-area business with a single base, it is usually a small number of area pages — the base location, plus the handful of surrounding areas where you do meaningful volume and can say something particular. The mechanics for businesses without a storefront are covered in AI visibility for service-area businesses.
City-by-service combinations need the strictest filter. The useful question is not "could someone search for this" but "is our answer for this service in this town different from our answer for the service in general". Occasionally it is — a service priced differently because of distance, or one offered only from a particular branch. Usually it is not, and the service page plus the location page already cover it.
Consolidating a matrix you already have
If the site already carries hundreds of generated pages, the fix is a consolidation, and the order of operations matters.
Step 1: Pull clicks per URL for the last ninety days. Clicks, not impressions. Sort the templated pages into those that earned at least one and those that earned none.
Step 2: Build the keep list. Pages with clicks stay, as do pages that pass the five-question test regardless of clicks. Write the list down as data — a file the build reads — rather than as a decision someone has to remember.
Step 3: Map every dropped page to a real destination. Each removed URL redirects permanently to the most relevant page that still exists: the location page for that town if there is one, otherwise the service page. Never redirect everything to the homepage, and verify that every target actually exists. A redirect into a missing page is worse than the doorway it replaced.
Step 4: Rewrite the internal links. Links from blog posts, footers and service pages that pointed at dropped URLs should point at the new destination directly. Leaving them to pass through redirects wastes crawl and hides the new structure. How internal links shape retrieval is covered in internal linking for AI retrieval.
Step 5: Rebuild the sitemap from the keep list. Only pages you want indexed belong in it, with honest last-modified dates.
Step 6: Improve what you kept. This is the step that produces the gain. Fifty pages with real local content will do more than five hundred without.
Step 7: Wait, and measure the right thing. Crawlers take weeks to revisit removed URLs. Judge the result on clicks and enquiries to the pages you kept over the following two to three months, not on the total number of indexed pages, which is supposed to fall.
The same discipline applies whenever URLs change at scale; the fuller checklist is in protecting AI visibility through a site migration.
The blog version of the same mistake
Doorway patterns are not confined to service pages. A blog that publishes "Guide to [service] in [town]" for every town in a list is the same matrix one layer down, and it is treated the same way. Automated publishing makes this easy to do by accident: a content schedule that needs a topic every week will reach for the next town on the list unless it is told not to.
The guard is a rule in the content process, not good intentions. No post whose only distinguishing feature is a place name; no post whose title differs from an existing one only by location; a check for near-duplicate titles before anything is published. Fewer, deeper pieces outperform a calendar of variations.
Making local pages citable
A location page that passes the test can also earn AI citations for local questions, provided it states its facts in sentences that survive being quoted alone.
"We cover the whole area" gives an assistant nothing. "The Riverside branch serves the city centre and the four northern suburbs, with same-day appointments on weekdays" is a complete, attributable statement. Write the address, hours, area served, services available at this location and typical cost in that self-contained form, and mark the page up with local business structured data that matches the visible text exactly.
Then check that what the page says agrees with the Google Business Profile and the main directories. DigiRank Expert's citations audit compares a location's name, address, phone and website across 21 directories, and the geo-grid scan shows where in the service area each location actually ranks — useful evidence when deciding which area pages are worth keeping. Both are included from the $99/mo Starter plan; recurring scheduled scans and automated listing fixes arrive with Agency. The full module list is on the features page.
Frequently asked questions
What is the difference between a location page and a doorway page? A location page contains information that is true only for that place — a real address or defined service area, local staff, local detail and proof. A doorway page changes only the place name in a template and exists to capture a search rather than to serve a visitor. The test is usefulness, not uniqueness of wording.
Are city pages bad for SEO? Not when the business genuinely operates in the city and the page says something specific to it. They become a risk when produced in bulk for towns where there is nothing particular to say, because search engines treat large sets of near-identical regional pages as doorway abuse.
How many location pages should a service-area business have? As many as there are areas where you have something specific to say — usually the base location and a small number of surrounding areas where you do real volume. A page for every town within driving distance is almost always too many.
Should I create a page for every service in every city? Rarely. Create one only where your answer for that service in that town genuinely differs from your general service page, such as different pricing or availability. Otherwise the service page and the location page already cover the query.
How do I fix a site that already has hundreds of thin location pages? Keep the pages that have earned clicks or pass a usefulness test, permanently redirect each of the rest to the most relevant surviving page, rewrite internal links to the new destinations, rebuild the sitemap from the keep list, and then improve the pages you kept.
Will removing thin pages hurt my traffic? Pages that have earned no clicks have no traffic to lose. Total indexed pages will fall, which is the intended result. Judge the consolidation on clicks and enquiries to the remaining pages over the following two to three months.
Do AI assistants cite location pages? They can, when the page states local facts in self-contained sentences and those facts agree with the business's profile and directory listings. Templated pages that repeat the same generic text offer nothing specific to quote and are rarely cited.