Doorway pages and thin content: how to scale pages without getting penalized
Short answer: A doorway page is a page made to rank for one narrow search and then pass the visitor on to somewhere else. Thin content is a page that adds little that is new or useful. Google, the search engine most business sites depend on for traffic, does not ban pages made in bulk or with AI. Its spam policies judge purpose and value: pages made mainly to win rankings, with nothing real on each one, are the problem. To scale safely, give every page facts that only that page has, test a small batch first, and keep a hand-built path that people can browse.
You want to publish dozens or hundreds of pages, one for each city, service or product, and you want to know where the line is before Google draws it for you.
I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.
The line is not the number of pages. It is whether each page gives a real person something they could not get from the page next to it. A business with 40 service areas can safely publish 40 pages if each one has real local facts. A business that changes only the town name on one template and publishes 400 copies is asking for trouble. This guide uses Google's own wording, then gives you a test, a pilot plan and a checklist you can copy.
Part of this page covers making pages from a database. One tool that does this is Ploy, an AI marketing platform that builds and hosts websites and can turn spreadsheet rows into pages. The advice below stands on its own if you never use it.
This page may contain affiliate links from which Blur Studio earns commissions.
This is information, not legal or professional advice. Google's rules change, so check its current spam policies before you scale anything. For the related question of whether AI-written pages get penalized, see does Google penalize AI-written pages.
What counts as a doorway page or thin content
Google's spam policies page (checked 21 November 2026) names several related problems. They overlap, so it helps to see them side by side.
| Term | Google's wording, shortened | Plain meaning |
|---|---|---|
| Doorway abuse | Sites or pages created to rank for specific, similar search queries that lead users to intermediate pages that are not as useful as the final destination | Many near-identical pages that exist to catch searches and push people to one place |
| Scaled content abuse | Many pages generated for the primary purpose of manipulating search rankings and not helping users, no matter how they are created | Bulk pages with little value. A human, a script or AI can make them. The method does not matter |
| Thin content with little or no added value (manual action) | Thin affiliate pages, content from other sources such as scraped content, and doorways | The label Google uses when a person on its team reviews your site and finds these |
Google's list of doorway examples includes pages or domains "targeted at specific regions or cities that funnel users to one page" and "substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy". The second one means a page that is little more than a list of links to other pages, with no real content of its own.
For scaled content abuse, Google's examples include using generative AI tools "to generate many pages without adding value for users", scraping feeds or search results to build many pages, and stitching together content from other pages without adding value. Notice the repeated phrase: without adding value. Using AI is not on the list as a problem by itself. Google's February 2023 post on AI content says "not all use of automation, including AI generation, is spam", and also that AI does not give content any special gains.
Manual action or algorithm: how a penalty shows up
There are two ways a site loses visibility, and they feel different.
- Manual action. A person at Google reviews your site and applies a penalty. You will see it in the Manual actions report in Google Search Console, the free Google tool that shows how your site performs in search. After you fix the problem, you use "Request Review" in that report.
- Algorithmic demotion. Google's systems rank your pages lower. There is no notice. Only the Manual actions report shows a manual action, so a traffic drop after an update does not prove you were penalized.
Recovery times vary. Gary Illyes, a Google search team member, spoke on this at Search Central Live (Google's event for site owners) in Barcelona on 2 October 2026, as recapped by Search Engine Journal, a search industry news site, on 4 October. The recap gave a typical manual action removal of 1 to 2 weeks (slowest 4 to 6 weeks or longer) and a typical core update recovery of 3 to 6 months (slowest 6 months to a year). It did not say how "typical" was measured, so treat these as rough ranges.
John Mueller, another Google search team member, commented on programmatic SEO, which means making many pages from data and templates, in a Bluesky (social network) post on 7 September 2026, reported by TechWyse, a marketing agency news page. He said it often leads to a site that is "either spam, borderline spam, or low quality", that trust loss is site-wide and that recovery tends to take time and significant effort. That is a trade-press report of a social post, so read it as a warning, not a rule. The part to remember is site-wide: weak pages can hurt your good pages too.
Good location and service pages versus doorways
Most small businesses meet this problem through location pages: "plumber in Springfield", "plumber in Shelbyville" and so on. Google does not say these pages are banned. It says pages that funnel people to one page, or that are copies with a swapped name, are doorways.
Here is a simple test. I call it the swap test.
- Open page A (the Springfield page).
- Replace every mention of Springfield with Shelbyville.
- Ask: does it still read as true and complete? If yes, page A has nothing that belongs to Springfield. It fails.
Then ask a second question, the existence test: would this page exist if search engines did not? A page for a real service area that you really serve, with a real phone number, real jobs and real prices, would exist anyway. A page for a town you have never worked in would not.
What makes a page worth having
Each page needs facts that only that page has. Examples, depending on your business:
- Real jobs you did in that place, with dates and photos you took (with the customer's permission).
- Rules that differ by place, such as permit steps, tax rates or delivery times.
- Prices or price ranges you actually quote in that place.
- Named staff, crews or offices that serve it.
- Reviews from customers there.
- For a data page, numbers from your own data that are different on each row.
If you cannot fill at least three of these for a place, do not publish a page for it. Cover it on a single "areas we serve" page that lists the places plainly. That page is a normal hub, not a doorway, as long as it has real information and links to pages that have real content.
Safe scale rules
These rules come from Google's wording above and from basic sense. They do not guarantee a result, because no one outside Google can promise that.
- Start with the reader's question, not the keyword list. Write one sentence for each page: "This page helps someone who wants to know X." If two pages have the same sentence, merge them.
- Put unique data on every page. Every page must pass the swap test: the swapped version should no longer be true. Facts come from your business, not from rewording the same paragraph.
- Keep a browseable structure. Home page, then a hub (for example "Services"), then the pages. People and crawlers should reach every page through normal links, not only through a search box or a sitemap.
- Have a person review every page before it goes live. This matters most when AI helped write it. Google's AI content guidance warns that outputs may contain inaccuracies, which it calls hallucinations. A person who knows the business must check names, prices and claims.
- Give each page a real goal. A phone number that works, a form that works, a real next step. A page nobody can act on is a sign it exists for search only.
- Publish in batches and measure. Never launch 500 pages in one day with no feedback. The pilot plan below shows how.
- Remove what does not earn its place. Pages that fail should be improved, merged into a stronger page, or removed. More on that in the audit.
Worked example with made-up numbers
Here is a made-up business to show the method. All figures below are invented for the example.
Harbor Drain Co is a fictional plumbing company that works in 12 towns. Its owner wants 12 town pages and, because a competitor has 200, is tempted to add 188 more for towns it has never visited.
| Option | Pages | What each page has | Risk |
|---|---|---|---|
| A. Copy and swap | 200 | Same text, town name changed | High. Fails the swap test on every page |
| B. Real towns only | 12 | Local jobs, price range, crew, permit note, reviews | Low. Each page has facts only it has |
| C. Hub and spokes | 1 hub plus 12 | Hub lists all towns, links to 12 full pages | Low, and easier for visitors to browse |
The owner picks C. For the town of Oakridge, the page gets: three real job write-ups with photos, the town's permit step for water heaters, the price range the company quoted there in the last year, and two customer reviews. The Pinewood page gets different facts. Each page now passes the swap test, because the pages cannot be swapped.
For the other 188 towns, the owner adds one line on the hub: "We also serve nearby areas. Call to check." No pages are made. That choice costs nothing and removes the main risk.
Pilot plan: test before you scale
Google says crawling a new page can take from a few days to a few weeks, and that asking for indexing does not guarantee it. So a pilot needs patience.
- Week 1. Build 10 pages that pass the checklist below. Publish them and add them to your sitemap.
- Weeks 2 to 4. Open the Page indexing report in Google Search Console. Look at the status of each page. "Discovered, currently not indexed" means Google knows the page but has not crawled it yet. "Crawled, currently not indexed" means Google fetched it and chose not to index it. That second status is common, but if most of your pilot pages sit there, treat it as a quality signal.
- Decision. Pass: most pages are indexed, and real people use them (calls, form fills, time on page). Fail: most are not indexed, or nobody uses them. Fix quality first. Do not add volume.
- Next batch. Add 20 to 30 more, then repeat the check. Stop adding when new pages stop being indexed.
Pass and fail here are your own rules, not Google's. Write them down before you start so you do not argue yourself into scaling after a bad pilot.
Audit checklist you can copy
Use this on any set of pages, new or old. Score each page. Any "no" in the first six lines means fix or remove the page.
- Does the page pass the swap test (it would not make sense on another place or product)?
- Would the page exist if search engines did not?
- Does it contain at least three facts unique to this page?
- Has a person who knows the business read and approved it?
- Can a visitor take a real action (call, buy, book, download)?
- Is it reachable from a hub page through normal links?
- Is the main text different from the other pages by more than the name or a few words?
- Does the title and heading describe this page only?
- Are there photos, tables or data that you made or own?
- Is the page accurate today, with a review month if the facts change?
What to do with a page that fails
- Add real facts. Best when the page serves a real need.
- Merge. Combine two thin pages into one strong page and redirect the weaker address to it with a permanent redirect (a 301).
- Noindex. Add a noindex tag, which tells search engines not to list the page. Google's spam page says to exclude such content from Search, and Google says the crawler must be able to fetch the page to see the tag. Blocking it in robots.txt, a plain file that tells crawlers what they may visit, does not work for this, because Google never sees the noindex.
- Delete. Remove it and redirect to the closest useful page.
After a manual action, fix the pages across the whole site, not only the examples Google shows, then use "Request Review".
Making pages from a database, and where Ploy fits
Programmatic pages are legitimate when each row holds real, different data. A directory of 300 real suppliers, a price list by region from your own quotes, or a library of 200 real product specs can each justify one page per row. They are risky when the row only holds a keyword.
Ploy's database, called PloyDB, works like a spreadsheet inside the site. According to Ploy's docs, one template turns every row into a page, adding a record publishes a new page, and edits reach visitors usually within a minute without a republish. Limits listed in the docs include 97 columns per table, 128 KB per text value, CSV imports up to 32 MB and 10 MB per file. The docs do not discuss search rankings. Ploy's blog names Clay, a software company, as a customer that uses it for programmatic pages. Read more in how PloyDB and the agentic CMS work.
The tool makes pages easy to produce. It does not make the data unique. If your spreadsheet has only a keyword column, a faster tool just makes more thin pages faster. If your spreadsheet has real columns (price, local rule, crew, photo, review), the template can turn them into pages that pass the swap test.
Pick or skip
- Pick a database-to-pages tool if you already own a table of real, different facts and you update it often. A tool that publishes from rows saves time on layout and keeps pages consistent.
- Skip it if your plan is to generate text to fill a keyword list. No tool fixes that plan.
Whatever tool you use, keep your own copy of the data, since CSV is for moving data around and not a full backup. For the exit and export side, see who owns your site and content.
How I'd approach this
I would not ask "how many pages can I make?" I would ask "how many pages have I got real material for?" Usually the answer is smaller than the owner hoped, and the pages are better for it.
If you want help building a data-driven page set with real data behind it, or cleaning up one that has already gone wrong, Blur Studio builds the custom version: AI agents that use your business knowledge, connections to your CRM and calendar, and ongoing care. See what that includes.
First-week plan
- Day 1. List every page you have or plan. For each, write the one-sentence reader question. Merge duplicates.
- Day 2. Run the swap test and the existence test on each. Mark pass or fail.
- Day 3. Open Search Console. Check the Manual actions report (it should say no issues) and the Page indexing report for "Crawled, currently not indexed" pages.
- Day 4. For failing pages, choose add facts, merge, noindex or delete. If you will build pages from a table, try Ploy's free plan on a small test set of 10 rows, with real columns.
- Day 5. Have a person who knows the business read the 10 pilot pages and sign off.
- Day 6 and 7. Publish the pilot, submit the sitemap, and set a calendar date 3 weeks out to apply your written pass or fail rule.
Decision rule: if a page can only be told apart from its neighbor by a place name, do not publish it. If it holds facts only it has, and a person has checked it, publish it and measure.
Related guide: making hundreds of pages from a spreadsheet.
Questions people ask
What is a doorway page in simple words?
A page made to catch one narrow search and send the visitor on to another page, so the page itself gives little. Google's spam policies call this doorway abuse and give examples such as many region or city pages that funnel users to one page.
Does Google penalize AI-written pages?
Google says not all use of AI generation is spam, and also that AI gives content no special gains. The problem is many pages made mainly to manipulate rankings that add no value, however they are made. See our guide on AI-written pages. Have a person review AI-written pages for accuracy.
Are city or location pages always bad for SEO?
No. Google's concern is pages targeted at cities that funnel users to one page, or near-identical pages. A page for a place you really serve, with real local facts, jobs, prices or reviews, passes the swap test and gives a visitor something real.
How many pages is too many?
Google gives no number. Its test is purpose and value, so a site with 10,000 useful pages can be fine and a site with 50 copies can be a problem. Publish in batches and check Search Console between each batch.
How do I know if I have a manual action?
Open the Manual actions report in Google Search Console. If it lists an issue, you have one, and it explains what to fix. If it says no issues, a traffic drop may come from a ranking change, which is not a penalty notice.
How long does recovery take?
It varies. As recapped by Search Engine Journal from a talk by Google's Gary Illyes on 2 October 2026, manual action removal is typically 1 to 2 weeks and core update recovery 3 to 6 months, but the recap does not say how typical was measured. Treat these as rough ranges.
Should I noindex thin pages or delete them?
If the page has a use for visitors, such as a policy or a filter page, noindex keeps it available and out of search. If no one needs it, delete it and redirect to the closest useful page. Do not block it in robots.txt as a substitute, because Google then cannot see a noindex tag.
Can Ploy build programmatic pages safely?
Ploy's database turns rows into pages, but the tool cannot make your data unique. It is a fair fit if your rows hold real, different facts. If you want to try it, you can start with a small test set. Its docs say nothing about search rankings, so judge the pages by the swap test and by indexing results.
More in this topic: Websites: Google rules, privacy and ownership (6 guides) →
Sources
- Spam policies for Google web search (Google Search Central)
- Thin content with little or no added value manual action (Google Search Console Help)
- Google Search's guidance about AI-generated content (Google Search Central Blog)
- Using generative AI content on your website (Google Search Central)
- Creating helpful, reliable, people-first content (Google Search Central)
- Spam policies and core update announcement, March 2024 (Google Search Central Blog)
- Google, crawling and indexing recovery ranges (Search Engine Journal)
- Google's John Mueller on programmatic SEO and site trust (TechWyse)
- PloyDB database docs (Ploy)
- Ploy pricing (Ploy)