Mueller says Google can lose faith in a whole site over programmatic pages

Google Search Advocate John Mueller said that programmatic SEO "often leads to a site that's either spam, borderline spam, or low quality", and that when it does, "our systems have possibly lost faith in your site providing good value to users based on the old pages." His comments were made on Bluesky and reported by Search Engine Roundtable on 7 September 2026, and they describe a loss of trust at the level of the whole site rather than a ranking problem confined to the pages that caused it.

The practical weight of the remark sits in the recovery, not the diagnosis. Mueller said that resolving this "tends to take time and significant effort to show the value", and framed it alongside the way sites recover from spam issues and core update hits. That places mass-generated pages in a category most teams do not plan for: the pages can be deleted in an afternoon, and the site can still be carrying the consequence long after they are gone.

what john mueller said on bluesky

John Mueller, writing on Bluesky in comments reported on 7 September 2026, said that programmatic SEO "often leads to a site that's either spam, borderline spam, or low quality". The mechanic he described is familiar to anyone who has built a page template: a site iterates through domain names, technologies and attributes, generates a page for each combination, and ends up with a large set of URLs where each individual page carries very little value of its own.

His second sentence is the one that changes the risk calculation. "Our systems have possibly lost faith in your site providing good value to users based on the old pages", Mueller said. The subject of that sentence is the site, not the page set. He also said, "It's easy to spin something up with many pages, it's hard to provide real value to users", and that resolving the situation "tends to take time and significant effort to show the value."

None of this is a new policy. Mueller is a Search Advocate restating guidance that already exists in Google's spam policies, and Search Engine Roundtable reported it as commentary rather than as an announcement. What makes it worth reading closely is the specificity about where the damage lands and how long it lasts, which is the part most published advice on page generation skips.

what programmatic seo is, and where the honest line sits

Programmatic SEO is the practice of building one page template and filling it from a data source, so that a single piece of work produces hundreds or thousands of URLs. A flight route page for every origin and destination pair, a product comparison page for every pair of tools in a category, a city page for every service area. The technique itself is neutral. Booking sites, job boards, real estate portals and price comparison services are all built this way, and Google indexes and ranks them.

The line Mueller is pointing at is not the template. It is what the template is fed. A page that pulls genuinely distinct data for each row, so that a reader arriving on it gets an answer they could not get from the sibling page next to it, is doing the work a search result is supposed to do. A page that pulls a spun variable, where the only difference between two URLs is a city name or a tool name swapped into otherwise identical prose, is producing what Mueller described: many pages, very little value per page.

The test is not how the page was written. It is whether the data behind the row exists independently of the page. If a team can point at the source of the difference, a price, an inventory count, a set of reviews, a specification sheet, then the template is a delivery mechanism for real information. If the difference exists only because the template needed something to vary, the page set is the thing Mueller is warning about, whether a person wrote it or a model did.

why site-level trust loss is a different problem from a ranking drop

A page-level ranking drop is bounded and legible. One URL loses positions, its impressions fall, the pages around it are unaffected, and the fix is contained to that URL. Teams have tooling for this, and the feedback loop is short enough that a change made this week shows up in the data within a few weeks.

What Mueller described on Bluesky is a different shape. When he said Google's systems have "possibly lost faith in your site", the consequence attaches to the domain, which means pages that had nothing to do with the generated set can be carrying the cost. A service page written carefully two years ago, a case study, a technical guide that earns links on its own merit: none of those were part of the programmatic push, and all of them sit behind the same trust signal. That is what makes site-level trust loss expensive in a way the raw page count never suggests.

It also breaks the usual diagnostic habit. A team that sees soft declines across a range of unrelated URLs will normally look for a technical cause, a crawl issue, a template regression, an internal linking change. The pattern Mueller describes produces the same symptom from a completely different cause, and no crawl report will surface it, because nothing is technically broken.

what mueller said against what it means in practice

The table below sets each statement reported by Search Engine Roundtable against the operational consequence it implies. The left column is what was said. The right column is the reading, and it is a reading, not an additional claim from Google.

what removing the pages does and does not fix

Removing a thin page set stops it producing new signals. That is real and worth doing, and it is the necessary first step. What it does not do, on Mueller's framing, is restore the position the site held before, because the systems he described formed a view of the site based on those pages and that view does not reset when the URLs return a 410.

His phrasing is worth reading literally. He said resolution takes time and significant effort "to show the value", which puts the burden on demonstration rather than on removal. That is closer to how a site works its way back from a spam issue or a core update hit than to how it recovers from a broken canonical tag. Something has to be shown, repeatedly, over a period, before the assessment changes.

For a team deciding what to do with an existing page set, that reframes the choice. The question is not only whether to delete. It is whether the pages can be made to carry the value they were supposed to carry, by cutting the set down to the rows where real data exists and building those out properly, or whether the underlying data was never there and the set should go. A thorough technical and content audit is the usual way to find out which of those two situations a site is actually in, because the answer is a property of the data source, not of the page count.

what to measure per template, not per site

The measurement mistake that produces these page sets is counting output. A dashboard that reports pages published, or total indexed URLs, will show a programmatic push as a success right up until it is not. The metrics that would have caught the problem early are all per template and per page set, and none of them are about volume.

• Query coverage per page set. How many distinct queries does the set actually surface for, divided by the number of pages in it. A set of 4,000 pages surfacing for 60 queries is not covering a long tail. It is one page repeated 4,000 times.
• Engagement per page, measured on the set. Not the site average. A generated set will usually sit far below the hand-built pages on the same domain, and the size of that gap is the clearest early read on whether the template is delivering anything.
• Share of the set that has ever received an impression. Pages that have never surfaced for anything are the clearest statement that the row behind them was not distinct enough to be worth a result.
• Whether the data behind each row is independently verifiable. This is a content check rather than an analytics one, and it is the only one that can be run before publishing instead of after.

Run per template, these numbers turn a judgement call into an observation. A template where 90 percent of pages have never had an impression is answering the question on its own, and a team can stop that set at 200 pages instead of finding out at 20,000.

how this connects to google's scaled content abuse policy

Google's spam policies already cover scaled content abuse, which addresses generating many pages primarily to manipulate rankings rather than to help users. Mueller's comments sit alongside that policy rather than adding to it, and he did not announce a change to it or cite it as the mechanism at work. The connection worth drawing is about the definition rather than the enforcement: both the policy and Mueller's phrasing put the emphasis on whether pages provide value to users, and neither one draws a line based on how the pages were produced.

what the source did not establish

This section matters as much as the rest, because the gap between what Mueller said and what will be written about it is wide.

• Mueller gave no threshold. There is no number of pages named at which a site becomes a problem, and nothing in his comments supports one.
• He gave no recovery timeline. He said resolution takes time and significant effort, and he did not put that in weeks or in months.
• He did not confirm that any specific site had been affected. The comments describe a pattern, not a case.
• He drew no distinction between AI-written and template-written pages. The framing is about value to users, and it does not turn on the production method.
• This is not a policy change or an algorithm announcement. It is a Search Advocate restating existing guidance in a public reply.

what this means for thai marketers

Programmatic and AI-assisted page generation has become an ordinary growth tactic for Thai sites, particularly for bilingual builds where an English template and a Thai template multiply the same row set across two locales. That doubling is where the risk compounds quietly, because a set of 3,000 rows becomes 6,000 URLs without anyone deciding to publish 6,000 pages.

The check is the same one Mueller's comments imply anywhere. For each generated set on the site, is there a data source behind the rows that a reader would recognise as distinct information, and is the Thai version carrying that data or carrying a machine translation of an English page that was already thin. A Thai page set built from genuinely local data, real Bangkok pricing, real service coverage, real inventory, is doing something an English page cannot. A Thai page set built by swapping a district name into the same paragraph is the exact case being described.

For teams planning a bilingual expansion, the sequence that follows from this is to prove the template on a small set in one locale, measure query coverage and engagement on that set specifically, and only then scale it and translate it. Getting the Thai and English search strategy right at the template stage is far cheaper than unwinding a site-wide trust problem afterwards, particularly given that Mueller offered no timeline for how long unwinding takes.

faq

is this a new google penalty for programmatic seo?

No, this is not a new penalty or a policy change. John Mueller is a Search Advocate restating existing guidance in comments on Bluesky reported on 7 September 2026, and Google's spam policies already covered scaled content abuse before he said it.

how many pages is too many?

The source did not state a number, and Mueller gave no threshold at which a page count becomes a problem. His comments are about whether pages provide real value to users, which is a property of the data behind each page rather than of how many pages exist.

if i delete the thin pages, will my rankings come back?

Removing the pages stops them producing new signals but does not by itself restore the site, on Mueller's framing. He said resolving this "tends to take time and significant effort to show the value", which puts the recovery on demonstrating value over a period rather than on the deletion itself.

does this apply to ai-written pages specifically?

Mueller drew no distinction between AI-written and template-written pages in these comments. The framing is about whether the output provides value to users, so a hand-built template producing thin pages and a model producing thin pages fall on the same side of the line.

should sites in thailand change anything right now?

Nothing changed in Google's systems on 7 September 2026, so there is no urgent action. The useful response is to check any existing generated page set for whether the data behind each row is genuinely distinct, and to be particularly careful where a bilingual build has doubled a thin set across two locales.

where to take this next

The reporting from Search Engine Roundtable on Mueller's comments is available in full here, and it is short enough to read in the original rather than through summaries. If a site is already carrying a large generated page set and the per-template numbers look uncomfortable, the next step is working out which rows have real data behind them and which never did. Relevant Audience works with teams on exactly that kind of assessment, and a conversation is a reasonable place to start.