Template pages can sink the whole domain

Programmatic SEO is the practice of generating large numbers of pages from a template and a dataset, and Google has now said plainly that a weak version of it can damage how its systems judge the entire site, not only the generated pages.

That matters more in 2026 than it did a few years ago. Google's AI Overviews and AI Mode pull their supporting links from the same index and ranking systems as classic search. A domain that Google has stopped trusting is a domain that gets retrieved less often when an AI answer is being assembled.

#What Google said

On September 7, 2026, a site owner on Bluesky described a site that generated pages from combinations of domain names, technologies and attributes. John Mueller of Google replied that "programmatic SEO like this often leads to a site that's either spam, borderline spam, or low quality," and that Google's systems had "possibly lost faith in your site providing good value to users based on the old pages," as reported by Search Engine Roundtable and Search Engine Journal.

Two details in that answer are worth reading slowly.

The first is "old pages." Mueller tied the loss of trust to the site's history, not to whatever it publishes next. New, better pages are judged against that history.

The second is the recovery estimate. Mueller said resolving it "tends to take time & significant effort to show the value," and compared it to recovering from spam and core updates. He gave no threshold and no timeline. He also did not say all programmatic SEO is spam. He was describing that specific setup.

#Why one section can affect the rest

This is not a new mechanism, just a blunt description of one. Google's ranking systems guide says its systems work at the page level, and that "site-wide signals and classifiers are also used and contribute to our understanding of pages."

Google's core updates documentation describes the same thing from the recovery side. If you see a sustained drop "for your site as a whole," it points you to a self-assessment of whether the site overall is helpful. It also says it "could take several months for our systems to learn and confirm that the site as a whole is now producing helpful, reliable, people-first content."

Put those together and the risk is clear. A large block of thin generated pages is not isolated from your good pages. It contributes to a site-level picture, and that picture is slow to change once formed.

#Where the line sits

Google's spam policies define scaled content abuse as "when many pages are generated for the primary purpose of manipulating search rankings and not helping users." The listed examples include generating many pages with AI or similar tools "without adding value for users," scraping or automatically transforming other content, and stitching pages together "without adding value."

None of those examples name templates as the problem. A template is a layout. The problem is what fills it. A useful test for each generated page type:

• Does each page answer a question someone actually searches, or does it exist because the combination was possible?
• Does the page carry data a reader cannot get from the template alone, such as real specs, real prices, real reviews, or first-party measurements?
• If two pages from the set were placed side by side, would a reader learn something different from each?
• Would you link to it from a hand-written page on your own site?

Pages that fail all four are the ones Mueller was describing. Pages that pass are closer to a product catalog or a reference database, which Google has indexed for as long as it has existed.

#Why this reaches AI answers

Google's documentation on AI features and your website says that to appear as a supporting link in AI Overviews or AI Mode, a page "must be indexed and eligible to be shown in Google Search with a snippet," and that there are "no additional technical requirements." The same page says both features may use query fan-out, issuing multiple related searches across subtopics.

So AI visibility in Google inherits whatever Google's search systems think of your site. Fan-out makes this more visible, not less. Each sub-query is a fresh retrieval, and each retrieval is another place where a domain with weak site-level signals can lose to a competitor. A brand that relies on a few strong hand-written pages can see those pages cited less if a large generated section has dragged down the domain around them.

There is a second, quieter cost. Generated pages that target long-tail combinations are exactly the pages fan-out queries tend to hit. If those pages are thin, they are the version of your site the AI system meets first.

#What to do about an existing programmatic section

Start with an inventory, not a deletion. Export every URL pattern your templates produce and group the pages by type. For each type, look at impressions, clicks and whether any page in the group has ever earned a link or a citation.

Then sort each group into one of three actions:

1. Improve. The page type answers real demand but is thin. Add the missing data, consolidate near-duplicates, and write the parts a template cannot.
2. Noindex or consolidate. The page type has some use for visitors on your site but no search value. Keep it for users and keep it out of the index. Google's spam policy itself advises that if you host such content, you should "exclude it from Search."
3. Remove. The page type exists only because the combination was possible. Google's core updates guidance calls deleting content "a last resort," to be used when it "can't be salvaged." Generated combinations with no demand often meet that bar.

After that, set expectations honestly. Google says improvements may take several months to register and that there is "no guarantee" changes produce a visible effect. Track the hand-written pages you care about as well as the cleaned-up section, because site-level recovery shows up first on pages that were never the problem.

#What to do before you build one

If you are planning a programmatic section now, flip the order. Decide what unique data each page will carry before you decide how many pages to generate. Launch a small set, check whether those pages earn impressions and citations, and only then expand. Keep the generated section clearly separated in your URL structure so you can measure it, noindex it, or remove it as a unit if it underperforms.

The cost of a weak programmatic section is not limited to the pages it creates. Google has now said so directly, and with AI answers built on the same systems, the cost reaches the places where buyers increasingly start their research.