Programmatic SEO still works. Templated pages do not.
The difference between a page set Google ranks and a page set it filters is whether each page says something only that page can say.
What changed and what did not
Scaled page generation is not penalised for being scaled. It is penalised for being empty. What changed is that generating plausible text got free, so plausible text stopped being worth anything.
The surviving test is simple: if you cannot say something true on this page that you cannot say on any other page in the set, the page should not exist.
Picking the axis that actually varies
Geography is the most common axis and usually the worst, because a city name swapped into the same 1,700 words is exactly what filtering targets. Variant, use case, material, size, compatibility and price band are better, because the content genuinely differs.
Build the data model first. Decide what varies. Only then generate.
- Bad axis: city name in the same copy
- Better axis: use case, material, compatibility
- Best axis: something only your data can populate
How much unique content is enough
There is no threshold, but there is a workable heuristic. If two pages in your set share more than about half their meaningful text, they are competing with each other rather than with anyone else. Measure it, do not guess.
The internal linking part
A programmatic set that links everything to everything gives search engines a mesh, not a map. Link each page to the three most closely related pages with descriptive anchors, and to its parent hub. That is what produces a topical structure.
Read next
Other posts: What is Generative Engine Optimization (GEO)?Answer Engine Optimization: a practical guideAI Overviews and the Indian traffic dropEntity SEO for Indian D2C brandsHow to write content AI engines will citeWhich sources do AI engines cite in India?Local SEO and AI answers share the same dataChoosing an SEO agency in India in 2026
Primary sources and further reading
Google's guidance on scaled content and on generative AI features both bear on this directly: the objection is to content with no value, not to scale itself.
- Google's guide to optimizing for generative AI features
- How Google Search works
- Google Search Central documentation updates
- Google Search Central: AI features and your website
External links open in a new tab. Our own figures and their limits are published on the method page.
Twenty buying prompts, four engines, you against three named competitors. Report in one working day. No call needed.
Programmatic SEO after AI Overviews: common questions
Is programmatic SEO risky in 2026?
Only when the pages are near-duplicates. A set where each page carries genuinely different information is normal, useful publishing.
How many programmatic pages is too many?
There is no count. Fewer pages that each answer one question beat a thousand that answer the same one.
How do I check for duplication?
Compare pages within the set on overlapping word sequences. Anything above roughly fifty percent shared text needs work.
See what the engines say about you
Twenty buying prompts, four engines, your brand against three named competitors. Report in one working day.
- 20 buying prompts, your category
- ChatGPT, Gemini, Perplexity, AI Overviews
- You against three named competitors
- Report in one working day, yours to keep