On 7 September 2026 John Mueller answered a site owner on Bluesky, and by the evening the quote was everywhere with the qualifiers stripped out. The headline version was “Google says programmatic SEO is spam”. That is not what he wrote, and the part everyone skipped is the part that actually helps you.
What was actually asked, and what was actually said
The site owner runs a technology lookup site. He had generated a large set of pages by combining domains, technologies and attributes, then noindexed the thin ones. His problem was that months later his core pages, including his about page, were still stuck on “Page is not indexed: Crawled - currently not indexed”.
Mueller replied in a chain of five posts. The relevant ones:
“Regarding the site you describe, it’s hard to say, but I worry a bit because it sounds like a programmatic SEO play, where you iterate through a large list of domain names, technologies, and attributes, finding all possible combinations.”
“It’s easy for a programmer to do, even easier with code generators now. You end up with tons of pages that have ‘some’ value, but the overall picture is not that exciting.”
“Programmatic SEO like this often leads to a site that’s either spam, borderline spam, or low quality. It’s easy to spin something up with many pages, it’s hard to provide real value to users. Our systems have possibly lost faith in your site providing good value to users based on the old pages.”
“Resolving this tends to take time and significant effort to show the value.”
Three qualifiers carry the whole meaning and every trade headline dropped all three. “Like this”, meaning that specific combinatorial pattern. “Often”, not always. “Possibly”, because Mueller was guessing from a description, not looking at the site.
He is consistent on this. In July 2023 he wrote that “programmatic SEO is often a fancy banner for spam”. Same word again. Often.
The sentence that matters most
“Our systems have possibly lost faith in your site providing good value to users based on the old pages.”
That is a site-level statement, and it is the reason this owner is stuck. He already deleted the thin pages. He already applied noindex. Months later his about page still will not index.
If the assessment were per-URL, removing the bad URLs would fix it. It is not. The signal attached to the whole domain, and the good pages inherited it. Noindexing junk does not reset the judgement any more than taking down a bad review deletes the memory of it.
That is also why Mueller says it takes time and significant effort. There is no switch.
Status names as they appear in the Search Console Page indexing report.
The Search Console read I would run first
The two statuses look almost identical in the report and they mean opposite things. Sort your Page indexing report by these two before you change anything.
Discovered - currently not indexed means Google knows the URL exists and has not spent a fetch on it. That is a budget and priority call. If your generated set sits here, you have a volume problem, and cutting the URL count usually moves it inside a few weeks. My full walkthrough of the difference between discovered and crawled covers the sorting.
Crawled - currently not indexed means Google fetched the page, read it, and decided against it. That is the quality call Mueller was describing.
Now the diagnostic. Look up a page that has nothing to do with the generated set. Your about page, your pricing page, your contact page. If those carry Crawled - currently not indexed as well, you are not looking at a thin page problem. You are looking at the site-level loss of faith, and no amount of pruning the generated folder will move it on its own.
How to prove which folder is dragging you down
Split your sitemap. One sitemap file per template folder, submitted separately: one for the generated pages, one for the editorial pages, one for the core site pages.
Search Console reports indexing per sitemap. Within a couple of weeks you have three separate indexed counts instead of one blended number, and the argument about which folder is the problem stops being an argument.
Then watch Crawl Stats. The report splits crawl purpose into Discovery and Refresh. A site Google has lost confidence in shows discovery crawling collapsing towards nothing while refresh crawling continues on the handful of pages it already trusts. That ratio moving back up is the earliest recovery signal you get, and it moves long before rankings do.
What will not help is the URL Inspection tool’s Request Indexing button. It is capped at roughly ten a day and I have written before about what those request indexing limits actually are. It does not override a quality assessment. Ten forced fetches against a site-level judgement is not a plan.
So is programmatic SEO bad?
The pattern Mueller described is bad. Enumerate every combination of a list, publish the cross product, and hope volume carries it. That is what “programmatic SEO like this” points at, and it fails because the pages have some value each and no value together.
Generating pages from a database is not the same thing. I build service area clusters programmatically all the time, and the ones that index carry something per page that could not be find-and-replaced. On one UK locksmith cluster 138 of 193 pages are indexed and climbing, on a mobile notary build it is 243 of 250. Same technique, opposite outcome.
The difference is not the code that made the pages. It is whether a person searching for that specific thing gets an answer worth the click. Google has never had a rule against generated pages. It has a rule against pages that exist only because they were cheap to make.
Sources
- John Mueller, reply on Bluesky, 7 September 2026, and the four posts that follow it in the same thread.
- John Mueller, post on X, 25 July 2023.
- Page Indexing report, Google Search Console Help.
- Crawl Stats report, Google Search Console Help.