Skip to content
Chalk Labs / AI SEO
How-to guide

How to Create Programmatic Pages Without Triggering Spam Policies

By Rahil Jain, founder of Chalk LabsUpdated How we research

The short answer

Programmatic pages are safe when each page answers a real question with data that is specific to it and that readers cannot get elsewhere. Google's spam policies target pages made mainly to manipulate rankings, however they are produced. Set a quality gate before launch, check overlap between pages, publish in small batches, and noindex any page that fails.

  • Google judges purpose, not method: many pages made mainly to manipulate rankings can break policy even if people wrote them.
  • Each page needs facts only it can carry: its own data, its own local detail, its own answer.
  • Test overlap before launch, publish in batches, and noindex pages that fail the gate.
  • Google warns that a page for every query variation, including fan-out queries, is not a long-term strategy.

What do Google's spam policies actually say?

Google defines scaled content abuse as many pages generated for the primary purpose of manipulating search rankings 1. It lists using generative AI to produce many pages that add nothing for users, and scraping feeds or search results into pages, as examples 1. Doorway abuse covers pages made to rank for specific, similar queries, including region or city pages that funnel users to one destination 1.

Google's AI search guide goes further. It says creating separate content for every query variation, including fan-out queries, mainly to manipulate rankings or AI responses violates the scaled content abuse policy 2. A fan-out query is one of the extra searches an AI tool runs behind the scenes. The guide adds that more pages do not make a site higher quality 2.

Do it step by step

0 of 8 done0%

  1. Write the question in one sentence for a sample page. If you cannot, the page has no purpose. Cut it.

  2. Name the column in your dataset that makes page A different from page B. Check that it is real, sourced and not scraped without added value.

    Tip: Aim for at least two fields that change the answer, not only the title.

  3. Write the first three pages manually. Fix the template only after you see where they feel thin.

  4. Use the checklist below. Decide in advance who reviews a sample from each batch and what happens when a page fails.

  5. Compare every pair in the batch with the script below. Rewrite or merge pairs above your limit.

  6. Launch 10 to 20 pages. Link them from one hub page and add them to your sitemap.

  7. Check Search Console for indexed pages, impressions and any manual actions. Read some pages as a customer would.

  8. Add noindex to pages that fail the gate and merge pages that duplicate each other. Redirect merged URLs with a 301 to the best page.

Your ticks are saved in this browser only.

Copy-ready templates

Quality gate checklist

markdown
A page ships only if every box is ticked:
- [ ] It answers one clear question a person would ask
- [ ] At least two facts on it are specific to this page
- [ ] Every fact has a named source and a date
- [ ] The first two sentences give the answer
- [ ] A human has read it and could not mistake it for its neighbour
- [ ] Overlap with every sibling page is below our limit
- [ ] A hub page links to it with descriptive anchor text
- [ ] We would be happy for a customer to land on it

Apply to a random sample of each batch, and to every page of the first batch.

Six-word overlap checker (Node)

bash
cat > overlap.mjs <<'EOF'
import { readFileSync } from "node:fs";
const grams = (f, n = 6) => {
  const w = readFileSync(f, "utf8").toLowerCase().replace(/[^a-z0-9\s]/g, " ").split(/\s+/).filter(Boolean);
  const s = new Set();
  for (let i = 0; i + n <= w.length; i++) s.add(w.slice(i, i + n).join(" "));
  return s;
};
const [a, b] = [grams(process.argv[2]), grams(process.argv[3])];
const both = [...a].filter((x) => b.has(x)).length;
console.log(Math.round((100 * both) / (a.size + b.size - both)) + "% overlap");
EOF
node overlap.mjs page-a.txt page-b.txt

Export each page's visible text to a .txt file first. This is a rough screen, not a judgement of quality.

Noindex tag for a failed page

html
<meta name="robots" content="noindex">

Place in the head of the page. Do not also block the URL in robots.txt, or Google cannot see the tag.

Is it the method or the purpose that gets you in trouble?

The purpose. Google's guidance on generative AI says that using AI to generate many pages without adding value may violate the scaled content abuse policy, and that models predict words, not retrieve facts, so output needs manual fact-checking 3. Templates and databases are not the problem. Thin, interchangeable pages are.

Google's people-first questions point the same way. They warn against producing lots of content on many topics hoping some will rank, and against summarising others' work without adding value 4. If a page would exist only because a keyword exists, do not build it.

What makes a programmatic page genuinely unique?

Unique data, not unique wording. Swapping a city name or synonyms into the same paragraph does not count, and Google lists producing pages through synonyms or translation of scraped content as abuse 1. A page is unique if its core answer changes when the input changes.

Ask: what can this page say that no neighbouring page can? A cost page for a city can use that city's prices, rules, offices and examples. A comparison page can use real test results for that pair. If the answer is 'nothing', merge the pages.

Page ideas that pass and fail the uniqueness test
Page ideaSource of unique valueVerdict
Price guide per cityLocal quotes you collectedPass, if the data is real
Service in [town] x 400Only the town name changesFail: doorway risk
Tool A vs Tool B, all pairsTested differences per pairPass only where you tested
Glossary term per pageOriginal definition and examplePass
Translated copies of scraped textNoneFail

How do you set a quality gate before launch?

Write the gate as a short list of pass-or-fail checks, and apply it to a sample of every batch. A page passes only if it has its own data, its own answer in the first lines, a named source for each fact, and a reason a human would visit it.

Add a numeric overlap limit. In our own AI SEO library, a checker measures how many six-word phrases two pages of the same type share. Region and industry pages get a warning when one tenth of their phrases overlap and are blocked at one fifth. Pick limits that suit your template, and test them on a few pages before you build the rest.

How should you roll pages out?

Start with 10 to 20 pages, not thousands. Publish them, wait for indexing and traffic data, and read the Search Console results before building more. Google's AI guidance notes that indexing and serving are not guaranteed even when a page meets every requirement 2.

Keep pages findable through normal links. A hub page that lists the set, with descriptive anchor text, helps readers and crawlers; see internal linking for AI search. Do not hide pages behind search boxes or load them only when scripts run.

When should you noindex a programmatic page?

Noindex a page when it fails the gate but must stay for users, such as a filtered view or a record with too little data. Google supports a robots meta tag with noindex, or an X-Robots-Tag header for files that are not HTML 5.

Do not block the page in robots.txt and expect it to disappear. Google says a page blocked by robots.txt can still appear in results if other pages link to it, because the crawler never sees the noindex tag 5. The page must stay crawlable for noindex to work.

What about pages for several languages or regions?

Treat each as a real market, not a copy. Google warns that translating only boilerplate while keeping the bulk of a page in one language creates a poor experience 6. For near-identical content across regional URLs in one language, it advises choosing a preferred version and using canonical and hreflang tags 6.

See how to do AI SEO in multiple languages for the language side. Our view is simple: if you cannot add a fact that is true only for that region, do not make the page.

When should you not use programmatic pages at all?

When you lack a dataset worth publishing. A handful of hand-written pages usually beats hundreds of weak ones, and the editorial route is covered in programmatic vs editorial content. For background on the term, read what programmatic SEO is and what scaled content abuse is.

Frequently asked questions

Is programmatic SEO against Google's rules?

No. Google's policy targets pages made mainly to manipulate rankings, whatever the method 1. Pages built from a real dataset that help users are fine. Thin, interchangeable pages, or pages for every query variation, are the risk 2.

Can I use AI to write programmatic pages?

You can use it as a tool, but not as a way to mass-produce pages without adding value 3. Add original data, fact-check every claim, and have a person review a sample of every batch.

How many programmatic pages are too many?

There is no safe number. Google says more pages do not make a site higher quality 2. The right number is how many pages you can fill with distinct, sourced facts.

What if some of my pages are already indexed and thin?

Improve them with real data, merge them into a stronger page with a 301 redirect, or add noindex while you fix them. Keep the page crawlable so Google can see the noindex tag 5.

Do city pages count as doorway pages?

They can. Google lists pages targeted at specific regions or cities that funnel users to one page as an example of doorway abuse 1. City pages are fine when they carry facts true only of that city.

About the author

Rahil Jain

Rahil Jain is an AI SEO (answer engine and generative engine optimisation) consultant and co-founder of Chalk Labs.

LinkedIn · Method and credentials

Sources

  1. Spam policies for Google web search · Google Search CentralOpened October 20, 2026. Supports: Definitions and examples of scaled content abuse and doorway abuse, including generative AI at scale, scraping and synonym or translation rewrites, and region or city pages funnelling users. Last updated 2026-08-28.
  2. Optimizing your website for generative AI features on Google Search · Google Search CentralOpened October 20, 2026. Supports: Separate pages for every query variation or fan-out query, mainly to manipulate rankings or AI responses, violate the scaled content abuse policy; more pages do not make a site higher quality; indexing is not guaranteed. Last updated 2026-07-10.
  3. Google Search's guidance on using generative AI content on your website · Google Search CentralOpened October 20, 2026. Supports: Using AI to generate many pages without adding value may violate the scaled content abuse policy; generative models predict word sequences and output must be fact-checked. Last updated 2026-10-01.
  4. Creating helpful, reliable, people-first content · Google Search CentralOpened October 20, 2026. Supports: Warning signs of search-engine-first content: lots of content on many topics hoping some will rank, summarising others' work without adding value. Last updated 2026-10-05.
  5. Block Search indexing with noindex · Google Search CentralOpened October 20, 2026. Supports: Noindex via robots meta tag or X-Robots-Tag header; robots.txt does not prevent indexing and the page must be crawlable for noindex to work. Last updated 2025-12-10.
  6. Managing multi-regional and multilingual sites · Google Search CentralOpened October 20, 2026. Supports: Translating only boilerplate creates a poor experience; for similar content across regional URLs in one language, choose a preferred version and use canonical and hreflang. Last updated 2025-12-10.
How this page was made. Built on Rahil Jain's AI SEO method and drafted with AI research assistance. Every statistic links to the page we read it on (opened October 20, 2026 or nearby). Published October 20, 2026; last updated October 20, 2026. Chalk Labs sells AI SEO services, so we have a commercial interest in this topic. Spotted an error? Tell us and we will fix it.