The short answer
Programmatic SEO is generating many pages from one template with automatic insertion of data from a database: 'service in city' pages, comparisons, calculators, spec directories. The approach is not banned and remains one of the strongest channels in 2026 — but only if real data stands behind every page, not a word swapped into the same text.
What changed: in 2026 Google tightened its grip twice. In March a spam update shipped alongside a core update, and in June another spam update, global, across all languages. By industry breakdowns, sites with thousands of near-identical template pages lost 60 to 90% of their traffic. What was penalized was not the idea of scaling but its lazy version.
The line of what is allowed runs along one question: does each page give something unique? If yes, programmatic works. If pages differ only by the inserted city, that is what Google calls scaled content abuse, and it is exactly what the updates mow down.
What happened in the 2026 spam updates
So as not to build strategy on rumor, let me lay out the facts as they are documented.
Google formulated the scaled content abuse policy back in March 2024 and defined it as "generating many pages where the content is only slightly different, with the primary purpose of manipulating search rankings rather than helping users." Its enforcement only strengthened afterward.
In 2026 two spam updates shipped. The March one ran almost simultaneously with a core update — a rare case of two changes back to back, which made volatility especially strong. The June spam update, per Search Engine Roundtable, started on June 24 and finished on June 26, was global and affected all languages and regions. Google characterized it as "a normal spam update."
Who was hit hardest, by industry breakdowns:
- sites publishing dozens and hundreds of AI-generated pages per day with no editorial review;
- classic programmatic projects of the "best service in city" kind across hundreds of locations where only the city name changed;
- thin-content pages where there was neither first-hand data nor expertise beneath the template.
Losses on such projects reached 60–90% of organic traffic — essentially a reset of the channel. Recovery from an algorithmic filter for scaled content, by industry estimates, takes 3–6 months after the content is actually reworked.
An important note on sources: the exact percentages and timelines are third-party data (SERP tracking tools and SEO breakdowns), not official Google statistics. Google confirms the fact and the policy but does not name loss figures. I quote ranges as the measurements give them, not as a guarantee.
Where the line runs between programmatic and spam
The difference is not in the number of pages but in whether unique value stands behind them. Here is the simple test I use myself.
The substitution test. Mentally remove from the page the variable that changes from page to page — city, product, parameter. If what is left is indistinguishable from hundreds of other pages, it is a spam risk. If unique information remains (real prices for this city, specs of this exact product, local nuances), it is normal programmatic.
| Trait | Healthy programmatic | Scaled content abuse |
|---|---|---|
| Data on the page | Unique to each | One text with a word swap |
| Content source | Own database | Text generator over an empty core |
| Value to the user | Answers a specific query | Exists for a URL in the index |
| What is left without the variable | The page still makes sense | Pages are indistinguishable |
A classic example of a violation from industry breakdowns: a legal site generated tens of thousands of city pages with an identical body and a name swap — and got filtered. Recovery took months. The same template, but with real data per city (local courts, timelines, specifics), would have passed fine.
How to do programmatic right in 2026
The order I consider workable and safe.
- Start with the database, not the template. Programmatic is strong where you have a unique array of data: prices, specs, calculations, statistics. No database, no programmatic; a text generator will not replace it.
- One page = one real intent. Each page must answer a query people actually ask. A page created only for a keyword, not for a question, is a filter candidate.
- Check indexability and quality on a sample. Do not roll out ten thousand pages at once. Launch a hundred, see how they get indexed in Google, and only then scale. A mass dump of thin pages is a signal in itself.
- Keep the technical base. Speed, clean structure, markup — technical optimization on a programmatic project matters more than on a regular one, because a mistake gets replicated across thousands of URLs.
- AI for formatting, not for filling. A neural network structures existing unique data well. But if it writes text to fill a template over an empty core, you produce exactly what the updates catch. Competent AI automation speeds up data processing, it does not substitute for it.
- Plan a review. Once a quarter, check which pages get impressions at all. Thousands of zero pages are ballast dragging down the whole site's assessment.
Why it is worth it
After two updates it is easy to conclude "programmatic is dead." That is wrong. Its lazy version died — the one where URL count matters more than content.
Programmatic with real data remains one of the most effective channels precisely because competitors got scared and left. On a small market like Estonia this is especially visible: many niches are still weakly covered, and a structural project with a unique database reaches the top more cheaply than on large markets. The same logic applies here that I described in the Estonian casino affiliate niche teardown: the one who does it for real wins while others churn.
And keep a realistic horizon. Programmatic, like any SEO, does not deliver results in a week — I wrote about timelines in detail in the piece on how long SEO results take. And if you are unsure whether to invest in scale or first pick up quick leads with ads, see the breakdown of Google Ads or SEO.
If you want to check whether your programmatic will survive the next spam update, I can review the structure, page quality and risks as part of SEO services and tell you plainly where you have healthy scaling and where a filter blank.
