Google can lose faith in your whole site: the real risk in templated pages

Most websites built in the last five years contain templated pages generated from a spreadsheet or a database: one page per city, per service, per integration, per product attribute. It is a legitimate technique, and it is also the technique behind most of the sites that lose their rankings overnight. The difference is not the template. It is whether each page carries value that exists nowhere else on your site.

What Google actually said, and why it is different from the usual warning

On 7 September 2026, Google Search Advocate John Mueller posted on Bluesky about sites that had gone all-in on programmatic SEO. The quoted part that matters: programmatic SEO “often leads to a site that’s either spam, borderline spam, or low quality”, and when it does, “our systems have possibly lost faith in your site providing good value to users based on the old pages”. He added that resolving it “tends to take time & significant effort to show the value”.

Two things in that are worth separating, because the industry reaction mostly collapsed them into “programmatic SEO is dead”, which is not what was said.

The first is unremarkable: thin pages do not rank. Everybody already knew that.

The second is the one that should change how you plan a website. The assessment is site-level and it is retrospective. Google is not only deciding whether a given page deserves a position; it is forming a view about whether this domain tends to produce something worth showing, and it forms that view partly from pages you may have published two years ago and forgotten. That is why sites in this position report the specific, confusing symptom of good new content that simply does not get traction. The new page is fine. The context it arrives in is not.

This also explains the awkward asymmetry in timing. You can generate 4,000 pages in a week and delete them in an afternoon, but you cannot reverse a trust assessment in an afternoon, because nothing about deleting pages demonstrates value. Demonstrating value takes the thing Mueller named: time and effort.

Templated pages are not the same thing as scaled content abuse

Google’s spam policies define scaled content abuse as generating many pages primarily to manipulate search rankings, with little or no value added for users, and state explicitly that the policy applies regardless of how the pages were produced — automation, AI, human writers or any combination. The method is not the violation. The absence of value is.

That distinction is usable, because almost every honest business site publishes templated pages. Product pages are templated. Location pages are templated. So are job listings, event pages, integration pages and store finders. Here is where the line actually falls, dimension by dimension.

Read the right-hand column as a single failure with six symptoms: the pages were created for the index rather than for a reader. That is the thing being penalised, and it is detectable without any judgement about how the text was written.

The five-question value test, before you publish

This is the test we run on a sample of pages before a templated section ships, and on an existing section when a site has lost traffic and nobody knows which part caused it. Score each question 0, 1 or 2 for a single page, and repeat on ten randomly chosen pages rather than your best one.

1. Unique substance. Does this page contain at least one fact, number, image, quote or named entity that appears on no other page of this site? Score 2 if there are several, 1 if there is exactly one, 0 if the only difference is the noun in the heading.
2. Evidence of demand. Is there evidence that somebody searches for this specific thing — keyword data, existing Search Console impressions on a related page, sales enquiries that use these words? Score 0 if the page exists because the permutation existed.
3. Answer completeness. Could a reader finish this page and stop searching? Score 2 if it resolves the question, 1 if it resolves part of it, 0 if it exists to funnel the reader to a form.
4. Human accountability. Has a person with actual domain knowledge read this page, corrected it and put their name to it? Score 0 for unread output regardless of how good the generator is.
5. Maintainability. If a fact on this page changes in six months, will anyone know and fix it? Score against your real capacity, not your intentions.

The reason to score a sample rather than argue about the concept is that everyone defends templated pages with their best example. The index contains the average.

A worked comparison: 400 pages or 40?

Take a services business that wants to rank in 400 towns. Two plans, same annual budget.

The upkeep row is the one that settles most arguments, because it is arithmetic rather than opinion. Nobody has 133 unbudgeted hours a year, so in practice Plan A’s pages are never revisited, which is exactly how a page set that was merely thin at launch becomes inaccurate two years later. Volume is not the cheap option; it is a deferred cost with interest.

This is also the structural point in multi-location SEO: pages that cover overlapping intent without differentiated substance do not add reach, they split it. And the general version of the question is covered in how many pages a website should have — the count is an output of what you can substantiate, not an input you choose.

If you already published them: the order of operations

Most sites reading this are not planning a templated section, they are carrying one. The sequence matters, because the common instinct — delete everything on Friday — removes your ability to tell what was working.

1. Inventory before you touch anything. Export every URL in the section with 12 months of clicks, impressions, entry-page conversions and internal and external link counts. You cannot triage what you have not measured, and Search Console’s indexing data is not back-filled for gaps, so pull it now rather than later.
2. Classify each URL into one of five outcomes using the table below. Expect a distribution, not a single answer.
3. Act on the clear cases first — the 0–4 pages with no traffic, no links and no user purpose. This is usually the bulk of the set and the least risky work.
4. Improve the pages with demand rather than deleting them. A page with impressions and a bad answer is a brief, not a liability.
5. Publish something that demonstrates value in the same topic area: primary research, real data, a genuinely useful tool. Removal alone gives Google nothing new to assess.
6. Measure over quarters. Recovery is documented by Google as taking months and being tied to later updates. Weekly rank checks in this period cause bad decisions.

One warning about step 3. If the section drove meaningful conversions despite being thin, remove it in batches of no more than a quarter of the set and watch enquiries, not rankings. Thin-for-search and useless-to-customers are different findings, and a page that converts is telling you something the keyword data will not.

What “showing value” looks like in practice

Mueller’s phrasing — effort to show the value — is vague because the mechanism is. What is not vague is what the rest of the market treats as valuable. The 2026 Google Ranking Factors Expert Survey, published on 9 September 2026 by Cyrus Shepard and Dawn Shepard with 131 practitioners scoring more than 100 candidate factors, put content relevance and search intent match at the top, with content quality and links close behind. Meta descriptions came last. Nothing in the top of that list is produced by volume.

Concretely, on a domain that needs to re-establish itself, these are the things that have actually moved sites we have worked on:

• Original data. Something only you can publish: your own pricing ranges, aggregated anonymised results, a survey of your customers, timing data from your own operations.
• Depth on a narrow topic rather than breadth across a shallow one. One page that genuinely answers a hard question outranks forty that gesture at easy ones.
• Named authorship and a credible about page. Not for a score, but because it makes the site attributable, which is also what makes it citable by AI answer engines. See E-E-A-T for business websites for what that actually requires.
• Links earned by the good pages. The 2026 survey is a useful corrective here: backlinks placed second, not nowhere.
• Fewer, better pages, visibly maintained. A dated “reviewed” line that is true beats a sitemap with 4,000 entries.

If you are commissioning a website right now

This has a procurement implication, and it is the reason we are writing about a Bluesky post. Page count is still used as a proxy for value in web design proposals: 50 pages, 200 pages, “SEO pages included”. On the evidence above, an agency that offers you 300 location pages as a deliverable is offering you a site-level liability with a per-page price attached.

Three questions to ask before you sign anything that includes bulk pages:

1. What is the data source for the per-page content? If the answer is a language model and a keyword list, the deliverable is volume, not value.
2. Who reviews each page, and what does review mean? A named person and a checklist, or nobody.
3. What happens in month 13? If no one is resourced to maintain the pages, they will be out of date before they are profitable.

The honest version of this deliverable is smaller and more expensive per page, which is a harder thing to sell and a better thing to buy. It is also how we scope website design and development work: the page inventory is derived from what can be substantiated and maintained, not from what fits the package tier.

Frequently asked questions

Is programmatic SEO against Google's guidelines?

No. Programmatic SEO is a publishing method and Google does not prohibit methods. What Google prohibits is scaled content abuse, which its spam policies define as generating many pages primarily to manipulate search rankings with little or no value added for users. A generated page that carries accurate, differentiated, genuinely useful information is fine. A thousand pages that differ only in a swapped city name are the thing being described.

Can low-value pages hurt the rest of my site?

Yes, and this is the part most site owners get wrong. John Mueller said on 7 September 2026 that Google's systems can lose faith in a site's ability to provide value based on its old pages, which is a site-level judgement rather than a problem confined to the pages that caused it. In practice, well-researched pages on the same domain can under-perform because of what surrounds them.

How many location or service pages is too many?

There is no numeric limit, because the constraint is value per page rather than count. A useful test: can you write at least one sentence per page that is true of that page and false of every other page on your site, from primary knowledge rather than a rewrite? If you can do that 200 times, publish 200. If you can do it 30 times, publish 30 and cover the rest with one strong parent page.

If I delete the thin pages, do my rankings come back?

Not immediately, and not automatically. Deleting the pages removes the cause but does not itself demonstrate value, which is what the site-level assessment turns on. Google's own core update guidance describes recovery as taking months and being tied to subsequent updates. Plan for two to three quarters of rebuilding, not a fortnight.

Should I noindex thin pages or delete them?

Delete and return 410 when the page has no user purpose and no links or traffic. Redirect with a 301 when a better page already covers the same intent. Use noindex when the page is genuinely useful to visitors but has no business being in search results, such as filtered listings. Use noindex as a measurement step when you are unsure: apply it, wait four to six weeks, and decide with data.

Does AI-generated content trigger this?

Not by itself. Google's scaled content abuse policy is explicit that it applies regardless of how the content was produced, whether by automation, by AI, by humans or by a combination. The failure mode of AI content is that it makes it cheap to produce pages at a volume nobody has the capacity to research, review or maintain, which is how sites end up publishing thin pages at scale without intending to.

The takeaway

Templated pages are one of the most efficient things a website can do and one of the most common ways a website destroys itself. The mechanism that makes them dangerous is site-level: Google is judging what your domain generally does, so 900 empty pages set the context in which your 40 good ones are read. Before the next batch ships, run the five-question test on a sample of ten and total the scores. If the average is below eight, you do not have a publishing project, you have a research project that has not been done yet, and shipping it anyway is the expensive option.