Playbook · AI search and GEO

Updated

How do you build hundreds of pages without making thin content?

Programmatic SEO works when each page answers a real search with facts that belong to that page, and the template only handles the layout. It turns into thin content when the only thing that changes between pages is a keyword, which Google’s spam policy calls scaled content abuse.

Google defines scaled content abuse as many pages generated mainly to manipulate rankings and not to help users, including pages made with generative AI that add no value.
Source: Google Search Central, updated 28 Aug 2026

From my work

On Turing.tn, a site that showcases IT training, we first built a custom server-side rendering app in PHP, which took every Lighthouse check to 100. Only then did we do programmatic SEO, after a long round of research into the keywords we wanted to target. The order matters: build pages at scale on a slow, JavaScript-only site and you get the same problem, many times over.

This site works the same way. More than a hundred guides and service pages are written as typed data, and one route renders them all.

The rules I use for programmatic SEO

  • Keyword research first. A page exists because people search for that question. “The template can make one” is not a reason.
  • Three facts per page. If a page cannot reach three facts that belong only to it, it stays a draft. Drafts build in development and never go live.
  • Every number carries a source and a date, typed into the data, so a page without one is easy to spot in review.
  • Each page has its own answer in the first two sentences, written by hand.
  • Shared parts, like the price box, come from one place, so a price change reaches every page at once.

Checks that run at build time

  • The build stops if two pages claim the same URL, including the fixed pages that would quietly win over a generated one.
  • A related link to a page that does not exist stops the build too. Better a red build than a dead link a visitor finds first.
  • The sitemap and /llms.txt come from the same data, so a page that is not live is not listed anywhere.

Signs your pages are thin

Google’s own self-check asks whether you produce lots of content on many topics hoping some of it performs, and whether you use extensive automation to do it. If the answer is yes, cut the page count. Deleting pages hurts for about a day.

  • Swap the city or tool name and the page still reads the same.
  • The pages exist for keywords but nobody would bookmark one.
  • Numbers appear with no source, or the same number appears on every page.
  • Pages are stitched from other sites or generated by AI without anyone adding anything.

When not to build pages programmatically

If you only have a dozen real questions to answer, write a dozen pages by hand. It is quicker than building the machine. The template, the data model and the checks pay off when the count runs into the dozens or hundreds and the data is real.

Read next

Sources

Start

Tell me where the product is stuck.

In 30 minutes we look at where things stand, what you need from a CTO right now, and which option fits. Either way, you leave with a clear next step.

How the call goes

  1. 01You tell me about the product, the team and the next 6 months.
  2. 02I ask about the code, the stack and what keeps breaking.
  3. 03We pick the option that fits, or agree that none does.

09:00 to 20:00 CET

Founding rate: the first 3 clients get 20% off, in return for a named testimonial and a case study. On a monthly plan, the 20% applies to the first month.