Is programmatic SEO dead? What Google's 2026 updates removed

Is programmatic SEO dead? No. But Google shipped six ranking updates between March and October 2026, and they removed a lot of templated pages. The pages that went had nothing on them that the next page didn’t also have. Pages built on their own data mostly held. This post covers four things: what Google released, what its policy says, which page types still rank, and how to tell a policy hit from ordinary ranking churn.

Is programmatic SEO dead?

Google doesn’t penalise templates. Its spam policies target pages “generated for the primary purpose of manipulating search rankings.” The cases that lost traffic in 2026 fit that description: combinatorial builds, pages padded with AI text, and thin affiliate catalogues. Some data-backed sites gained in the same period, including marketplaces like trip.com and indeed.com in the May core update.

John Mueller’s September 2026 comment gets quoted as proof that pSEO is over. He was replying about one specific site that generated pages from combinations of “domain names, technologies, and attributes,” and what he said was: “Programmatic SEO like this often leads to a site that’s either spam, borderline spam, or low quality.” The key words are “like this.”

The 2026 update timeline

Most “pSEO died in 2026” posts get the dates wrong or invent a September core update. Here is what the Search Status Dashboard and the trade press confirm:

Sources: SEO Kreativ, Search Engine Journal on March core, Search Engine Land on May core, SEJ on June spam, the August incident page and Search Engine Roundtable on September. The dashboard shows no core update after May.

Google announced no new spam policies with any of these updates. Scaled content abuse has been a policy since March 2024, so the 2026 updates enforced an existing rule rather than introducing a new one.

What did Google’s March and May 2026 core updates change for templated pages?

Google hasn’t said. It published no blog post for either core update. Everything below comes from third-party data.

• March, Germany. SISTRIX counted 134 confirmed losers against 32 winners, about 4 to 1. Online shops were the largest losing group.
• March, UK. In the SISTRIX analysis, Aleyda Solis saw “a shift away from many intermediary, aggregator, directory, and quick-answer utility sites.” Pricerunner, a comparison site with real price data, gained 45.94%.
• May. Solis found that the winners were the pages that best matched a query’s intent, market and result type. Her warning applies directly to templates: don’t be “a weaker echo of a source that already owns it.”

My reading of this data is that a template repeating what the brand, the official site or the local domain already says is the likely loser. That is an inference from the numbers above. Google hasn’t said it.

What does Google’s scaled content abuse policy actually prohibit?

The policy’s examples include:

• using generative AI or “other similar tools to generate many pages without adding value for users”
• scraping feeds or search results to generate pages
• stitching together content from other pages “without adding value”
• spreading content across several sites to hide that it was produced at scale
• pages that make little sense to a reader but contain keywords

Two neighbouring policies matter for programmatic sites. Doorway abuse covers “pages targeted at specific regions or cities that funnel users to one page” and “substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy.” Thin affiliation is narrower than people assume: “Good affiliate sites add value by offering meaningful content or features.”

Does it matter if the pages are AI-generated?

No. Google’s March 2024 announcement says the policy applies “no matter whether content is produced through automation, human efforts, or some combination.” A team of writers filling a city template by hand is covered too.

Google’s AI content guidance adds one requirement on top: it is “critical to manually factcheck and review all AI-generated content.” If you generate 50,000 pages, plan for how you will check them.

What the spam updates actually removed

Glenn Gabe published four blinded cases from the August spam update. They are the best first-hand evidence available:

1. An ultra-YMYL site with programmatic country pages, some carrying 100% AI text. It lost rankings for more than 200,000 queries, and the rankings disappeared entirely instead of slipping.
2. An Amazon affiliate site built from Amazon data. Some categories had no images, and the boilerplate text was flagged as 97%+ likely AI. It lost over 14,000 queries.
3. A product-information site with more than 1.5 million indexed URLs, about 85% of them programmatic: a photo plus generic text, with heavy ads. The hit spread to the whole site, including sections with legitimate user-generated content.
4. A site with 250,000+ URLs that redirected users to riskier sites. It lost close to 25,000 queries.

Case 3 is the one to remember. Mueller described the same mechanism: systems may have “lost faith in your site providing good value to users based on the old pages.” Per SEJ’s report, he also said that adding more generated content to each page wouldn’t fix it. A weak template section can drag down pages that were never part of it.

Does programmatic SEO still work? A three-question test

Yes, when every page passes three checks. These come from the policy text and Solis’s analysis, and each of Gabe’s failures breaks at least one of them.

1. Unique data. Does the page show data the searcher can’t get on the next page or from the source that already ranks?
2. Intent match. Is it the result type Google wants for this query: a tool, a listing, a comparison table, a guide?
3. Usefulness. Would someone who landed on it from a link, with no search involved, find it worth reading?

If the only thing that changes between two pages is the city name, the page fails the first question. More on sourcing that data is in the data hub.

Which programmatic pages still rank, and why?

• Integration pages. Zapier is the usual example because each page documents a real product capability. Third-party traffic estimates disagree widely, from about 2M to about 16.2M monthly visits (Shno, Aurelius Media), so treat them as rough.
• Comparisons. These hold up when they are built on current data, as with Pricerunner’s March gain. Thin comparison pages are “a weaker echo.”
• Directories and marketplaces. These hold up when they have real inventory and reviews, like trip.com and indeed.com in May. Directories that only aggregate and add nothing were the losers in March.
• Location pages. These hold up only with real local data behind them. A city-swap page that funnels users to one booking page matches Google’s own doorway example almost word for word.

Is programmatic SEO bad?

Not inherently. The doorway policy explicitly accepts “a clearly defined, browseable hierarchy.” Mueller’s 2023 line was that pSEO “is often a fancy banner for spam.” That describes how these builds usually turn out. It isn’t a ban.

How can you tell if a traffic drop comes from an update or a quality problem?

Search Console won’t tell you directly. Google’s debugging guide lists six causes: algorithmic update, technical issue, security issue, manual action, seasonality and site move. Only manual actions appear in Search Console. This is the order I’d check them in:

1. Match the dates. Does the drop start inside one of the windows in the table above?
2. Open the Manual Actions report. If it’s empty, a human reviewer didn’t act on your site. An algorithmic demotion is still possible.
3. Wait a week. The core updates doc says to wait “at least a full week” after completion, then compare the week after with the week before.
4. Measure the size. Google calls a move from position 2 to 4 not worth “drastic action.” A move from 4 to 29 across many terms needs a “deeper assessment.”
5. Segment by template. Check whether only the programmatic section fell or the whole domain fell with it.
6. Rule out the rest. Use Google Trends for market-wide demand and Page indexing and Crawl stats for technical problems. Also look for a CTR-only drop (impressions flat, clicks down), which usually points to the snippet or to SERP features such as AI Overviews.
7. Compare against the spam policies. If the pages that lost traffic match the scaled content, doorway or thin affiliation examples, and the losses are complete rather than gradual, treat it as a spam-update hit.

Recovery is slow either way. The spam updates doc says improvements show up as systems learn “over a period of months.” Gabe cites one recovery that took about five months. The measurement hub covers how to set up template-level reporting before you need it.

When a programmatic build is the wrong call

I wouldn’t start one if:

• you have no data source of your own, and pages would differ only by a swapped variable
• the pages would echo a source that already owns the query
• you can’t fact-check generated text at the volume you plan to publish
• demand isn’t validated (Zapier’s own guide suggests testing with affiliates or paid ads first, and warns about tech debt)
• the domain already carries thousands of old low-value pages, which you should fix or remove first

Is SEO dead or evolving in 2026?

Google has shipped two core updates and four spam updates this year, so the rules are being enforced, and SEO is still very much alive. Local SEO isn’t dead either, but city pages without local substance are exactly what the doorway policy names. One clarification, because the searches overlap: programmatic advertising (automated ad buying) is a separate industry from programmatic SEO, and nothing here applies to it.

What I’d check before shipping a template

• Pick ten pages at random. Each should show at least one fact the other nine don’t have.
• Search the target query and check that the page type matches what ranks.
• Read one page aloud as if a friend sent you the link. If you’d close the tab, don’t publish it.
• Ship a small batch first, and only scale once those pages get indexed and earn clicks.