Last updated: October 23, 2026
Websites filled with low-quality AI-generated content grew more than fivefold in the second quarter of 2026 compared with the same quarter a year earlier, and traffic to them quadrupled, according to ad-measurement data reported on October 17, 2026. The count covers ad-supported sites the measurement firm tracks, the comparison quarter falls before the past twelve months, and the firm hasn't published how it classifies a site, so the figure shows a trend and can't tell you how many such sites exist.
Mass-produced pages are cheaper to make than ever, and Google's spam policies are written to catch them. The spreadsheet-and-template method behind a useful integration directory is the same one behind a content farm. What separates them is the data on each page.
My verdict: programmatic SEO works for B2B SaaS only when every page carries data that no other page on your site has. Integration pages, template galleries and some glossaries can pass that test. Vs pages, pricing pages and "best [category] for [industry]" grids built from one template almost never do. This guide covers which pages to scale, the test to run before you build, how to index in stages, and how to prune.
A note on that opening figure: it comes from an ad verification company that sells fraud detection to advertisers, so it has a business reason to show the problem as large. Weigh it alongside Google's own policies below before you act on it.
Which SaaS page types can be programmatic, and which can't
Programmatic SEO means building many pages from a structured dataset and a shared template, instead of writing each page by hand. A template is fine as long as the dataset gives each page something real to say.
Page type | Can it be programmatic? | Why |
|---|---|---|
Integration pages ("[your product] + Slack") | Yes, if each has real setup data | Triggers, actions, fields synced, permissions and limits differ per integration |
Template or example galleries | Yes | Each template is a real artifact a user can open or copy |
Glossary or definitions | Sometimes | Only when entries include your own examples, screenshots or product context |
Location or currency pages | Rarely | Only when price, tax, compliance or data residency actually changes by location |
Use-case pages by industry | Rarely | Swapping "for healthcare" for "for retail" doesn't change the product |
Vs and alternatives pages | No | Each comparison needs its own research and judgment |
Pricing pages | No | One canonical page; variants split the signal and go stale |
The B2B SaaS pages that carry the most pipeline, comparisons and alternatives, are exactly the ones that can't be generated. The guide to comparison pages explains why each one needs hand-written criteria. The B2B SaaS SEO pillar covers programmatic pages in one short section, and this post is the longer version.
The unique-data test
Before you build a template, take three rows from your dataset and write the pages out by hand. Then put them side by side and ask:
Remove the page title and the entity name. Can you still tell the three pages apart?
Does each page answer a question that the other two can't? "Does the Salesforce integration sync custom objects?" has a different answer from the HubSpot one. "What is [product] for Salesforce?" usually doesn't.
Is the data on the page something a user can't get from your main feature page? Field mappings, sync frequency, a real template preview and setup steps count. A rewritten intro paragraph doesn't.
Would you send this page to a prospect who asked about that exact case? If your sales team wouldn't, a search engine shouldn't rank it either.
Who keeps it accurate? If the integration's API changes, does anything update the page, or does it rot?
If two of the three test pages fail, the dataset isn't rich enough yet. Fix the data first. Usually that means pulling from your product (actual integration metadata, actual template files, actual usage limits) rather than from a keyword list.
One quick check: count the sentences that change between two pages built from the same template. If it's mostly the entity name and a couple of adjectives, you're looking at a doorway pattern, whatever your intent was.
Where Google draws the line
Nothing in Google's rules bans programmatic pages as a method. Its spam policies, last updated August 28, 2026, define scaled content abuse as generating many pages mainly to manipulate search rankings instead of helping users, and add that this applies "no matter how it's created." The listed examples include using generative AI tools "to generate many pages without adding value for users" and stitching together content from other pages without adding value.
Doorway abuse is the other policy programmatic sites trip. Google's example is "substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy." A grid of "[category] software for [city]" pages fits that description almost word for word.
Google's guide to generative AI features in Search closes the loophole for AI answers too. It warns that writing a separate page for each way people might phrase a search, fan-out queries included, mainly to sway rankings or AI responses "violates Google's scaled content abuse spam policy." It also says "a high quantity of pages doesn't make a website higher quality or more relevant to users." Building pages to catch AI sub-queries is the same tactic with a new label.
Then there's the risk to the rest of your site. In September 2026, Google's John Mueller replied on Bluesky to someone describing a site that generated pages from combinations of domain names, technologies and attributes, as reported by Search Engine Journal. He wrote that "Programmatic SEO like this often leads to a site that's either spam, borderline spam, or low quality," and that Google's systems had "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." He was talking about one site, gave no timeline or threshold, and said "possibly," so read it as a warning from one thread. It's still a clear public hint that old thin pages can weigh on how Google judges the pages you're proud of.
As of October 20, 2026, Google's Search Status Dashboard lists four spam updates this year, in March, June, August and September, so enforcement passes come often.
Index in stages, not all at once
The safest way to launch a programmatic set is to treat it as an experiment you can reverse.
Start with a batch. Publish 50 to 100 of your strongest pages, the ones with the richest data, and leave the rest unpublished or set to noindex. Watch them for four to eight weeks in Search Console. If the batch gets indexed and earns impressions, release the next batch. If most of it sits unindexed, adding more pages won't help.
Use the right control for each job:
noindex keeps a page out of the index while it stays live for users. Google's noindex documentation warns that the page must not be blocked in robots.txt, or the crawler never sees the rule.
rel="canonical" tells Google which of several near-duplicates you prefer. Google's canonicalization guide calls it a strong signal of which URL you prefer, so it won't rescue a set of pages that are mostly duplicates.
Separate sitemaps per page type make monitoring easier. Put the integration pages in their own sitemap so the Page indexing report shows their status on its own.
Watch two statuses in the Page indexing report. "Crawled, currently not indexed" means Google fetched the page and chose not to index it, which usually points to thin or duplicate content. "Discovered, currently not indexed" means Google knows the URL but hasn't crawled it yet. Google's crawl budget guide calls out sites where a large share of URLs sit in that second state, alongside very large sites, as the ones that need to manage crawling. Most B2B SaaS sites are far below the size thresholds it gives (roughly a million pages that change weekly, or 10,000 that change daily), so if your URLs pile up in "Discovered," the likely cause is quality or internal linking, not server capacity.
Index coverage problems hide in places you don't expect. In one Miraa engagement with an email automation SaaS, a technical audit in the first week found crawl budget issues that were halving the site's index coverage. That was a site without a large programmatic set. Add thousands of template pages to a site like that and the important pages compete with the new ones for crawling.
Build templates that AI answers can quote
A programmatic page earns AI citations the same way any page does: it has to be indexed and eligible to show with a snippet, per Google's AI features guide. After that, the page needs facts an answer can lift.
For an integration page, that means a short summary of what syncs and in which direction, a table of triggers and actions, the plan the integration requires, setup steps, known limits, and a date the page was last verified. Each of those is a self-contained fact. "Syncs contacts and deals both ways every 15 minutes, on the Pro plan and above" can be quoted. "Seamlessly connect your favorite tools" can't.
Write the intro paragraph once per page type, with variables that pull real data, and keep anything generic below the specific facts. If you use AI to draft the per-page copy, feed it the structured data and have a person review a sample of every batch before it's indexed. Google's policy is about added value whatever tool wrote the page, and generated filler is where most templates fail. For more on what makes a passage quotable, see how to get recommended by ChatGPT.
Prune what doesn't earn its place
Programmatic sets decay. Integrations get deprecated, templates go stale, glossary terms stop mattering. Put a pruning review on the calendar, quarterly for big sets.
Pull every URL in the set with its Search Console impressions, clicks and indexing status for the last six to twelve months. Then sort pages into four groups:
Earning impressions and accurate: keep and refresh.
Accurate but invisible: improve the data, merge into a stronger page, or noindex.
Outdated (deprecated integration, retired template): redirect to the closest live page, or remove with a 404 or 410 if nothing fits.
Duplicates of another page: consolidate with a redirect.
There's no published threshold for how many weak pages is too many. Set your own rule, write it down, and let the set shrink when the data says it should.
What to do, by situation
If you have more than ten real integrations, build integration pages first. They pass the unique-data test most easily, and buyers search for them right before they buy.
If your product produces artifacts (templates, workflows, dashboards, reports), a gallery is your next best programmatic set. Show the real artifact on each page.
If you're tempted by industry, role or city pages, write five by hand first. If you can't make five genuinely different, you can't make five hundred.
If you've already launched a big set that isn't indexing, stop adding pages. Noindex or remove the weakest half, improve the rest, and give it time. Mueller's comments suggest recovery is slow.
If you're a small team with a young domain, skip programmatic SEO for now. A dozen strong comparison and integration pages written by hand will do more for pipeline. Before you scale, get the technical SEO basics in order: crawling, indexing and internal links.
What to watch out for
Spinning up vs pages from a template. You'll publish claims about competitors that nobody checked, and buyers will notice.
Location pages for a product that works the same everywhere. That's a doorway pattern.
Indexing everything on launch day. You lose the chance to learn from a small batch.
Treating canonical tags as a fix for duplicate content. They tell Google which URL you prefer and leave the duplicates live, so remove or merge them.
Building pages for AI fan-out queries. Google has named this as scaled content abuse.
Forgetting the pages exist. A deprecated integration page with wrong setup steps hurts trust and wastes crawling.
Final thoughts
Code generators made programmatic pages cheap to build, and the test for whether one deserves indexing is the same as before: each page needs data the others don't have, and a reason for a buyer to land on it.
Start with the page type your product data supports best, launch a small batch, and let Search Console tell you whether to continue. For the rest of the technical floor these pages sit on, read the B2B SaaS SEO guide, and see measuring AI search visibility for tracking whether those pages show up in AI answers.