Google can lose faith in your whole site over mass-generated pages

TL;DR: Google’s John Mueller said last week that publishing a big batch of near-identical, low-value pages can make Google stop trusting the whole website, not only the pages themselves. Think of a travel site that generates one page per city from a spreadsheet, where every page says the same thing with the city name swapped in. Mueller’s point is that once Google’s systems decide a site like that isn’t worth much, deleting the pages doesn’t flip a switch back. Getting the trust back takes time and real effort. He announced no new rule and gave no numbers. What he did was describe the cost, and the cost lands on the site, not on the batch.

First, what is programmatic SEO?

Programmatic SEO means building pages from a template plus a list, instead of writing them one at a time. You have a page design — “Plumbers in ___” — and a list of two thousand towns. A script fills in the blank and publishes two thousand pages. Each one gets its own web address, and each one can, in principle, show up in Google for someone searching that town.

Nothing about that is against the rules. Job boards, property listings, currency converters and store locators are all built this way, and they’re genuinely useful, because each page carries information that only applies to that row: different jobs, different houses, different exchange rates.

The failure mode is when the rows don’t actually differ. The town changes, the sentence doesn’t. You end up with two thousand pages that are one page wearing two thousand name tags.

What Mueller said

Writing on Bluesky, Google Search Advocate John Mueller said programmatic SEO often produces a site that is spam, borderline spam or low quality — and when it does, Google’s systems may have “lost faith in your site providing good value to users” on the strength of those pages. He described recovery as something that takes time and significant effort to demonstrate value, and put it in the same category as coming back from a spam problem or a core update hit. Search Engine Roundtable reported the remarks on 7 September, linking his original post; Relevant Audience covered them separately.

Note what is absent. No page-count threshold. No recovery timeline. No named site, no algorithm update, no distinction drawn between pages written by a language model and pages written by a template from 2018. This is guidance restated, not policy changed.

The bit worth paying attention to: site level, not page level

Google’s published rulebook on spam — the search spam policies — defines scaled content abuse as generating many pages mainly to manipulate rankings rather than to help anyone, and the remedy it describes is to keep those pages out of Search. Read on its own, that sounds contained. Bad pages, bad pages removed.

Mueller is describing something wider. The pages are the evidence; the judgement is about the site. That distinction changes the maths for anyone who has ever shipped a large templated section as an experiment, on the assumption that the downside was capped at “those URLs won’t rank.” If the downside is actually the site’s standing overall, then the money pages that had nothing to do with the experiment are exposed too.

It also changes what “cleaning up” means. Pulling ten thousand thin pages down is an afternoon’s work. Convincing a system that has already formed a view of the site is not, and Mueller was explicit that it isn’t fast.

Why this bites harder now

Five years ago, generating forty thousand pages was an engineering project. You needed a data source, a build pipeline and somebody to maintain it, and that friction filtered out a lot of bad ideas before they shipped. Now it’s an API key and a weekend. The barrier that used to stop a mediocre programmatic build from ever reaching production is gone, and the volume of these sections on the open web has gone up accordingly.

So the guidance hasn’t moved, but the number of sites in a position to trip over it has.

What to do about it

Judge each template separately. Sitewide traffic charts hide this completely. Pull the numbers for one page set at a time — how many of those URLs are indexed at all, how many get impressions in a month, how many get clicks. A template where the great majority of pages have never had an impression is telling you something.

Check that the rows carry real information. Open three pages from the set and read them side by side. If the only differences are the ones a find-and-replace would produce, the set is thin no matter how many words are on each page.

Treat removal as step one. Deleting or noindexing (a noindex tag tells search engines to leave a page out of results) stops the bleeding. It doesn’t demonstrate anything. What demonstrates value is what you publish afterwards.

Validate in one locale before scaling. Building the same weak template in six languages multiplies the evidence against you rather than the reach.

Know what you already have. Plenty of sites are carrying a programmatic section somebody shipped years ago and forgot. Recovery advice is irrelevant if nobody has counted the pages.

That last item is the tedious one, and it’s tedious in proportion to how many sites you look after. Auditing every URL on a site — not a sample, every one — and doing it again next quarter to see whether an old templated section is still sitting there, is exactly the work agencies defer until a ranking drop forces it. Preferium runs 47 automated checks across every page and scores the site 0–1000, so a thin, duplicated or orphaned page set shows up as a number rather than as a hunch, and it re-runs on its own instead of waiting for someone to schedule it. See what the scan covers.

Key takeaways

• John Mueller said low-value programmatic pages can make Google’s systems lose faith in the site as a whole, not only in those pages.
• Recovery is slow. Mueller framed it like coming back from a spam problem or a core update, and gave no timeline.
• Google’s written spam policy describes scaled content abuse in page-level terms; Mueller’s framing of the consequence is site-level. Plan for the wider one.
• Nothing new was announced: no thresholds, no update, no named sites.
• Programmatic pages are fine when each row carries information the others don’t. They are the problem when the template is the only thing on the page.
• Measure one template at a time, and know what old page sets your sites are still hosting before you need to.