Few techniques promise so much and charge so much. Programmatic SEO lets you publish hundreds of pages from a spreadsheet, and that very ease is what puts the site in the crosshairs of spam policies. The promise is long-tail coverage in weeks, and the risk is the whole domain losing credit with the search engine.
The discussion returned to the center of SEO strategy in September 2026, when John Mueller replied to a site that had adopted the technique and explained why recovery is slow. The statement didn't introduce a new policy, but it named a mechanism that had already been showing up after every round of Google updates.
The practical point is that no one has to choose between scale and quality. There's a range where automated page generation works well, and it has observable criteria.
What is programmatic SEO and how does it work?
Programmatic SEO is the creation of pages at scale from a template and a structured database, where each combination of variables generates its own URL. A course catalog crossed with cities and formats produces thousands of pages with the same skeleton and variable content.
Caption: in programmatic SEO, every page is born from the intersection of the template and a data point of its own.
The logic is one of coverage. Instead of writing one page at a time for each specific query, the team models the search pattern and lets the system fill in the gaps.
Three elements sustain any implementation:
- The database, which holds the entities and attributes of each page.
- The template, which defines the title, heading structure, content blocks, and internal links.
- The publishing rule, which decides which combinations deserve a page and which are left out.
The third part is the one that's almost always missing. When every possible combination becomes a URL, the site publishes pages with no search demand, no original information, and no audience, and that's where the problem starts.
None of this is new, and the technique sustains entire businesses. Marketplaces, travel sites, and price comparison tools live off generated pages, because each one carries data that exists only in that database.
Why does Google associate programmatic SEO with spam?
Google associates programmatic SEO with spam because the technique became popular as a volume shortcut, not as a way to organize real information. Two statements from John Mueller, three years apart, show that the company's reading hasn't changed: the problem isn't automation, but the empty page it tends to produce.
In July 2023, Mueller responded to an observation about sites scraping third-party data to build hyper-specific landing pages and summed it up on the old Twitter by saying that “programmatic SEO is often just a fancy banner for spam.”
On September 7, 2026, on Bluesky, he went further and described the cumulative effect on the domain. There are four sentences worth reading in order, because each one answers a different question for anyone who has already published these pages.
The first names the pattern: “programmatic SEO like that tends to lead to a site that is spam, near-spam, or low quality.” The second explains the asymmetry of effort: “it's easy to put up something with lots of pages, it's hard to deliver real value to users.”
The third is the one that changes the diagnosis for anyone who has lost rankings: “our systems may have lost faith that your site provides good value to users, based on the older pages.” The assessment lands on the site, not on the isolated URL.
The fourth closes out the timeline expectation: “fixing this usually requires significant time and effort to demonstrate value.” There's no reconsideration button for quality assessed by an algorithm, so the clock runs on the site's reprocessing time.
It's worth noting what these statements are not. Neither of them creates policy, and neither condemns automated page generation on its own. Google's official documentation is even more precise on this point.
What are scaled content abuse and doorway pages?
Scaled content abuse and doorway pages are the two spam policies that reach pages generated in bulk. Both sit inside Google Search's spam policies, and both describe the intent behind the publication, not the tool used to produce it.
Scaled content abuse, in Google's definition, happens “when many pages are generated for the primary purpose of manipulating search rankings and not helping users.” The documentation lists five examples, and most of them describe common programmatic operations:
- Using generative AI tools or similar to generate many pages without adding value.
- Scraping feeds, search results, or other content to generate many pages, including through automated transformations like synonymizing and translation.
- Stitching or combining content from different pages without adding value.
- Creating multiple sites with the intent to hide the scaled nature of the content.
- Creating many pages where the content makes little or no sense to the reader but contains search keywords.
Doorway pages follow a different logic. Google defines them as sites or pages “created to rank for specific, similar search queries” that “lead users to intermediate pages that are less useful than the final destination.”
This is the policy that catches the classic case of a course crossed with a city, when the hundred variations all end at the same form and none of them answers anything specific about that city. The reader lands on a step that exists only to capture the search.
On AI-generated content, the official guidance is explicit and less restrictive than the market tends to repeat. Google doesn't ban the use of these tools.
The Search Central page on AI-generated content states that “using generative AI tools or similar to generate many pages without adding value for users may violate the scaled content abuse spam policy.” That framing repeats throughout the page: the target is scale without value, not automation itself.
Anyone wanting to dig deeper into that line finds the breakdown in how to create content with AI without destroying your SEO, with the review criteria that keep AI-assisted production inside policy.
How do Google's updates hit programmatic pages?
Google's updates hit programmatic pages through two distinct paths, and confusing them leads to the wrong diagnosis. A spam update reprocesses the detection of practices that violate the policies, while a core update reassesses overall content quality, with no violation involved at all.
The official Search Status Dashboard tracks the recent cadence, and it's useful for dating any ranking drop.
The August 2026 spam update started on August 18 and took just over two and a half days to finish. The May 2026 core update, on the other hand, started on May 21 and dragged on for nearly twelve days.
That difference in duration explains a good share of rushed diagnoses. Assessing the impact on day three of a long rollout measures partial propagation, not the result.
Here's how the two fronts differ in practice:
|
Criteria |
Spam update |
Core update |
|
What it reassesses |
Spam policy violation |
Content quality and usefulness |
|
Typical drop signal |
Sharp loss across URL blocks |
Gradual, distributed loss |
|
Recent rollout duration |
About two to three days |
Days to weeks |
|
Recovery path |
Remove the practice and wait for reprocessing |
Raise the quality of the site as a whole |
Table: Comparison between the two update fronts that most affect pages generated at scale.
The distinction matters because the operational response is the opposite in each case. A site that deleted pages by mistake after a core update loses coverage without fixing the underlying problem, and the mechanics of each type are detailed in Google spam update: how is it different from the core update.
There's also a third scenario: the programmatic page hosted on a third-party domain to piggyback on its authority. That practice has its own policy and is covered in parasite SEO.
How do you do SEO at scale without losing organic performance?
To do SEO at scale without losing organic performance, every generated page needs information that doesn't exist on any other URL on the site. The test is direct: when removing the city and course name makes two pages indistinguishable, they shouldn't be two pages.
This uniqueness criterion beats any word count. A 300-word page with its own price, schedule, and address delivers more than a 1,500-word page built out of synonyms.
Five decisions separate useful scale from inflated scale:
- Demand before the template. Only the combination with documented search volume or intent goes live, and the rest stays in the database without becoming a URL.
- Original data per page. Price, deadline, admission score, address, availability, and a local testimonial give the page a reason to exist.
- A minimum editorial block. A passage written by a person, specific to that combination, takes the page out of the auto-fill pattern.
- A coherent link architecture. The generated page needs to be reachable through navigation, not just the sitemap, which calls for an internal link audit.
- Periodic pruning. A page with no impressions, no clicks, and no links for two quarters is a candidate for consolidation or removal.
Pruning is the step almost nobody executes, and it's the cheapest of all. Consolidating fifty weak pages into five good ones reduces the surface area under evaluation and concentrates signal.
There's also an alternative path. Product-led SEO meets part of long-tail demand with useful tools and calculators, which attract search traffic without multiplying near-identical URLs.
Anyone who'd rather treat this as a project, with a cannibalization diagnosis and a consolidation plan, finds the scope on the agency's SEO service page.
What tips apply to programmatic SEO at educational institutions?
At educational institutions, tips for programmatic SEO start with choosing which axis to scale. Crossing the course catalog with cities is the most tempting combination and the riskiest one, because most combinations have no local information of their own to sustain a standalone page.
The axis that works is the one that changes the information, not just the label. An in-person course with an address, schedule, tuition, and seat count per campus has real content per page, while the same distance-learning course replicated across two hundred cities has exactly the same content.
Three applications tend to sustain organic performance in this sector:
- A page per course and per physical campus, with operational data only the institution has.
- A page per admissions cycle, with the date, official notice, application method, and admission score for that period.
- A page per field of study, aggregating courses, job market outlook, and career path.
The counterexample deserves a name too. A page per course-and-neighborhood combination, per popular variation of a course name, or per synonym of the same search adds no information, and those are the first three candidates for pruning in any audit.
It's also worth tracking the effect at the right level of granularity. Monitoring only total sessions hides the behavior of URL blocks, and the segment-level reading path is in how do I find out my site's position on Google.
What changes in SEO for AI when a site scales pages?
In SEO for AI, a generic programmatic page loses even more ground than it did in traditional search. AI response systems retrieve self-contained passages and cite whoever brings verifiable information, so a page built from swapped variables rarely offers what the model is looking for.
The reason is mechanical. AI Overviews and AI Mode use the same index and the same ranking systems as search, and on top of that they break the question down into sub-questions.
In that setup, what increases the odds of citation is the opposite of auto-fill: a named source, a dated statistic, and a passage that answers on its own. A page with no original data has none of that to offer.
The practical consequence for SEO strategy is a reversal of priority. Twenty pages with exclusive data yield more visibility in generated answers than two hundred pages with the same text, and the effect of the AI layer on traffic is mapped in Google AI Mode.
One methodological caveat applies here. Google states there's no additional requirement to appear in AI features, so none of this is published policy — it's market reading of the passage-retrieval mechanism.
How do you recover Google's trust after a drop?
Recovering Google's trust after a drop tied to pages generated at scale requires reducing the bad surface area before producing new content. Since the signal is assessed at the site level, adding good content on top of a large base of weak pages dilutes the effort instead of concentrating it.
The sequence that tends to work has four steps, and the order matters more than speed:
- Inventory. List every URL generated by the template and cross it with impressions, clicks, and links received.
- Decide by block. Classify each block as keep, rewrite, consolidate, or remove, without treating URLs one by one.
- Execute the cut. Consolidate with a redirect to the equivalent destination, and use a 410 only for pages with no substitute.
- Prove the value. Actually raise the quality of the pages you keep, with original data and human review, before scaling any new axis.
On timelines, expectations need to be realistic. Mueller's own statement points to significant time and effort, and quality reassessment happens continuously, with no reconsideration request like the one that exists for manual actions.
While that plays out, the indicator that matters isn't the site's average position. It's the progress of the block you kept, measured against its own history.
Frequently asked questions about programmatic SEO
Where should you start auditing your programmatic SEO?
Start with the inventory, because almost every wrong programmatic SEO diagnosis comes from not having one. Before discussing template, keyword, or publishing volume, the team needs to know how many URLs came from automated generation, how many got impressions in the last ninety days, and how many never got a single click.
With that map in hand, the decision stops being ideological. It becomes clear which scaling axis is sustaining organic performance and which is only increasing the domain's surface area under evaluation, and the conversation shifts from “scale or not” to “scale what.”
The September 2026 statement added urgency to the topic, but it didn't change the criteria. The same test still applies: does the page exist because someone needs it, or because the database allowed it to be created?
If your operation already produces content at volume and the question is how to keep quality consistent page by page, the natural next step is understanding the review criteria in how to create content with AI without destroying your SEO.
Portugal