Good programmatic SEO from real data with unique checked pages, versus mass AI publishing with identical pages

Good vs Bad Programmatic SEO:A Checklist for Scalable Content That Ranks

By · November 11, 2026 · 10 min read


Generative AI has taken away the biggest barrier to publishing at scale. A brand can now produce thousands of pages in an afternoon and call the result programmatic SEO. For a while, many of those brands saw strong search traffic and real organic revenue.

Management teams like the idea for obvious reasons: time to market is short and production costs are low. Those same two traits are what make the approach fail. Aggressive programmatic projects are now collapsing or drawing manual penalties, because nobody stopped to ask whether each page deserved to exist.

This guide explains where the line sits between genuine programmatic SEO and mass AI publishing, what Google has said about it, and which checks help a scaled content program survive.

What Google has said about scaled content

Google has been direct on this point,it strengthened its scaled content abuse policy so that it applies whether the pages come from automation, from people or from a mix of both. The company later reported 45% less low-quality, unoriginal content showing up in results.

The policy language matters. The word "programmatic" does not appear in Google's spam policies. What appears is scaled content abuse, defined as many pages generated mainly to manipulate rankings instead of helping users, regardless of how they were made. Using generative AI to produce many pages without adding value is one of the named examples.

So a template is not what gets penalized, and neither is automation. The question that decides the outcome is whether each page adds value.

More recent signals point the same way. On September 7, 2026, John Mueller replied on Bluesky to a site that had generated pages from every combination of domain names, technologies and attributes. He said programmatic SEO of that kind often produces a site that is spam, borderline spam or low quality. He also said pages already published may keep shaping how Google judges the whole site, which means a bad batch can drag down pages that were fine.

Programmatic SEO is a publishing method, not a content type

Programmatic SEO is a way of publishing. It is not a kind of content. It has three parts:

A repeatable search pattern that many people use in slightly different forms.

A structured dataset that makes each page unique and useful.

A template that answers the query for every variation.

A currency converter shows how this works. The pattern is "convert X to Y." The dataset is a live rate for every currency pair. The template turns each row into a page. Nobody writes thousands of pages by hand, the data makes each page different, and the site publishes only as many pages as the data can support.

Mass AI publishing looks identical from the outside, which is why people confuse the two. It keeps the template and the publishing system, but the thing that makes each page different comes from a prompt. Swap the city name and you have a new page with nothing new on it.

A quick test with two cost of living pages

Picture two pages for "cost of living in Lisbon," built from the same template and aimed at the same search pattern.

The first page shows rent ranges, transport passes and grocery prices pulled from a maintained dataset. It updates when the numbers change.

The second page looks just as complete. It has a rent figure, a monthly budget and a confident summary. But the numbers are the model's best guess, with no source and no date. The same sentences appear on every other city page with only the name changed.

Only the first page counts as programmatic SEO. The way to tell them apart is to ask where each number came from and when it was last updated. The first page can answer both questions. The second cannot.

How to tell good pSEO from bad pSEO

Most teams struggle to say whether their pages are programmatic SEO or mass publishing, because the output looks the same on the surface. A useful comparison looks underneath: where the data comes from, whether each page maps to a distinct intent, how fresh the numbers are, and whether the template adds anything beyond swapped keywords.

None of these checks ask whether AI wrote the page. A page built on real data and drafted with AI help can perform well. A hand-written page with no real data behind it has no better claim to rank.

Treat the comparison as a diagnostic. If one check fails, you have a workflow problem to fix. If several fail, you are mass publishing, whatever name the project carries.

From static templates to AI pipelines

Programmatic SEO has been around for years, and the pages teams built in 2016 look much like the ones they build today. What changed is how easy it became to ship a bad one.

The middle era is worth studying. A rule-based natural language generation engine such as Rosae NLG writes sentences from structured data, and it can only say what the data and the template author allow. Building one was slow, and that slowness did useful work, because it forced teams to have real data before they had pages.

Language models removed that limit. A model will write a fluent, confident paragraph about a row that has nothing in it, and a reader cannot tell the difference. Fluent text with no data behind it is exactly what scaled content abuse policies target.

The checks had to evolve with the tools. A workable modern process splits the work this way: people do the research and decide each page's structure, the model drafts sections from that structure, and a subject matter expert reviews the facts and the brand fit before anything goes live.

Language models did not make programmatic SEO better or worse. Tools used to stop teams from scaling too fast. Now teams have to stop themselves, and a written checklist is the practical way to do it.

The pSEO checklist to run before any page ships

Before the first page

Find a unique data source. If you cannot name it, you do not have a programmatic SEO program.

Define the minimum data completeness a row needs before it can generate a page, and enforce that rule in code.

Confirm that each template maps to a distinct search intent and not just a keyword variant.

Write the metadata rules for titles, descriptions and schema with the same care as the body content.

Decide on a disclosure line that explains how the pages are built and how often the data refreshes.

Ask what you can add that nobody else has, such as your own measurements or two datasets combined into a new one.

At launch

Ship a test batch of 25 to 50 pages instead of the full set. Publishing 10,000 pages before you know your index rate is the most common failure in this field.

Set up indexation tests from day one. Use server logs and not Search Console alone.

Review a sample by hand. Ten pages read from start to finish will tell you more than any automated score.

Ongoing

Track crawl hits per URL by template every month.

Write a kill rule before launch. For example, if a set share of a page group earns no impressions within a set window, including impressions in AI features, prune or merge those pages.

Review again whenever the template changes, because a template edit changes every page at once.

Prune on a schedule and not after an update. By the time a core update forces the issue, a site-level trust problem may already be in play.

Programs that survive and grow through several Google updates tend to share one habit: they had a written rule for deleting pages before they ever needed it.

Building a pSEO program for AI search

Search is shifting underneath these programs. In the first four months of 2026, 68% of U.S. Google searches ended without a click, according to SparkToro's analysis of Similarweb data. That is up from about 60% in 2024 and about 45% in 2016. SparkToro points to AI Overviews, which it says now appear on more than 20% of searches. Ahrefs estimates that AI Overviews cut clicks on the top result by 58%.

For a programmatic page, this changes the goal. Traffic cannot be the only test of whether a page earns its place, because the click may never arrive. What remains is influence. Rand Fishkin has noted that your content still shapes AI answers and Google's other zero-click features even while visits fall.

Three habits follow from that.

Build fewer pages with data nobody else has. If an AI answer needs a source for Lisbon's rent ranges, it can link to the page that holds them. It has no reason to link to a page that repeats what a dozen others already say.

Decide the query pattern before building the template. If the searcher wants a single fact, Google will probably supply it directly. If they need to compare, calculate, book or find a local business, your page can still be the destination. SparkToro says branded, local and high-intent transactional searches still benefit from SEO.

Measure by template and not only by site. Search Console's generative AI performance report shows how often links to your pages appear in AI Overviews and AI Mode. Export the pages, group them by template, and compare each group with its clicks in the regular performance report. A template with plenty of AI impressions and few clicks may be earning influence without visits. A template with neither is a candidate for your kill rule.

The core test has not changed. If a page's only value was that it ranked, zero-click search removes its reason to exist. If the value was the data, an AI answer still has to point to it. Before you scale anything, pick one page and ask where each number on it came from.

Frequently asked questions

What is programmatic SEO?

Programmatic SEO is a publishing method that creates many pages from a repeatable search pattern, a structured dataset and a template. The data makes each page different, and the template answers the query for every variation.

Is programmatic SEO against Google's rules?

No. Google's spam policies do not mention programmatic SEO by name. They target scaled content abuse, which means generating many pages mainly to manipulate rankings instead of helping users.

Does Google penalize AI-generated content?

Google's policy applies whether automation, humans or both produce the pages. The issue is value. Using generative AI to create many pages without adding anything useful is a named example of scaled content abuse.

What is scaled content abuse?

It is Google's term for producing many pages mainly to influence search rankings with little value for users. Google strengthened this policy on March 5, 2024, and later reported 45% less low-quality, unoriginal content in results.

What did John Mueller say about programmatic SEO?

On September 7, 2026, he replied on Bluesky to a site that built pages from every combination of domain names, technologies and attributes. He said this kind of programmatic SEO often leads to a site that is spam, borderline spam or low quality, and that already published pages may keep affecting how Google views the whole site.

How can I tell whether my pages are good pSEO or mass publishing?

Ask where each number on a page came from and when it was last updated. If the page can answer both questions from a maintained dataset, it is on the right side of the line. If the numbers are model guesses with no source or date, it is mass publishing.

How many pages should I publish at launch?

Start with a test batch of 25 to 50 pages. Publishing thousands of pages before you know your index rate is the most common failure in programmatic SEO.

What is a kill rule for pSEO pages?

A kill rule is a written plan, made before launch, that says when to prune or merge pages. For example, if a set share of a page group earns no impressions within a set window, including in AI features, those pages are removed or combined.

Why should I prune pages on a schedule?

Waiting for a core update to force the issue can be too late, because a site-level trust problem may already exist. Scheduled pruning keeps weak pages from dragging down the rest of the site.

How do AI Overviews affect programmatic SEO?

They reduce clicks. SparkToro reports that 68% of U.S. Google searches ended without a click in the first four months of 2026, and Ahrefs estimates AI Overviews cut clicks on the top result by 58%. Pages with unique data can still earn influence by being cited in AI answers.

How do I measure pSEO performance in AI search?

Use Search Console's generative AI performance report to see how often your pages appear in AI Overviews and AI Mode. Group pages by template and compare those impressions with clicks from the regular performance report.

Reference

Search Engine Journal - https://www.searchenginejournal.com/good-vs-bad-pseo-learnings-from-a-decade-of-building-scalable-content-engines/594204/

Metaflow life - https://metaflow.life/blog/what-is-programmatic-seo

SEO Kreativ - https://www.seo-kreativ.de/en/blog/programmatic-seo/

Search Engine Journal - https://www.searchenginejournal.com/the-content-framework-that-worked-in-2019-is-now-working-against-you/579051/

Tags: programmatic SEO, pSEO, scaled content abuse, AI content at scale, Google spam policy, pSEO checklist, mass AI publishing, zero-click search, AI Overviews, content pruning, unique data, John Mueller, SEO templates, scalable content engines