TL;DR Summary
Programmatic SEO means building a large number of search landing pages from a database instead of writing each one by hand. It works when every page gives the searcher real data they can't easily get somewhere else. It goes wrong when a team takes a public dataset, pours it into a template, and publishes thousands of pages that all say roughly the same thing.
A couple of years ago I watched a founder do exactly that. He exported 4,200 rows from a public spreadsheet, connected them to a CMS template, and had 4,200 pages live by the next morning.
He got zero clicks. Not one.
About a month later, Google had dropped almost all of those pages from its index.
I see some version of this story all the time. Founders notice how much traffic Zapier gets from its integration pages and conclude that more pages means more rankings. It doesn't work that way. Without data of your own, a big page count is just spam with a nicer layout.
I've spent more evenings than I'd like to admit looking at how Google treats large sets of templated pages, on new domains and on older ones. When each page genuinely answers a question with useful data, programmatic SEO can bring in traffic for years. When you try to game the system, the damage usually doesn't stay on those pages. It drags the rest of the site down with it.
Below I walk through two companies that got it right, one type of site that got it badly wrong, and the questions I ask before I'd ever build a programmatic template.
What separates programmatic SEO from programmatic spam?
Good programmatic pages each answer a specific search with specific data. Spam pages swap a city name or a tool name into the same paragraphs, so you end up with thousands of pages that are almost identical and tell the reader nothing new.
It helps to know what Google actually objects to. Its policy on scaled content abuse doesn't ban automation. It goes after pages mass-produced mainly to rank rather than to help anyone, whether a person, an AI tool, or a template wrote them. Building pages from a database is fine. Building pages just to catch every keyword variation isn't. And if a big chunk of your site falls into that second group, the rest of the site can suffer too.
What makes the good version work is the depth of the data behind it. Think about someone searching "connect Stripe to QuickBooks." They don't want 2,000 words on the history of accounting software. They want to know which events trigger the sync, which fields map to which, how long it takes, and how to set it up. Give them that from a database you keep up to date and they'll stay. Give them padded paragraphs and they're gone in a few seconds, and they won't sign up.
New domains have an extra problem. If you launch thousands of templated URLs with no authority and no internal links pointing to them, Google will usually find them, decide most aren't worth indexing yet, and leave them there. That's why Content-led SEO for SaaS has to start with solving real problems, not with how many URLs you can publish.
Case 1: Zapier's 3-tier integration directory (When it works)
Zapier's pages work because each one points to an integration that actually exists and fixes an annoying workflow problem. You don't get filler text. You get the real triggers and actions, plus templates you can turn on in a couple of minutes.
I spent a few weeks going through how Zapier organizes these pages while I was studying how SaaS companies get distribution. The structure has three levels.
At the top there's a page for each app, like the Airtable page. Below that are pages that pair two apps, which is how Zapier shows up for searches like "connect Google Sheets to Slack." Then there are pages for individual workflow templates that handle one everyday task.
Every level earns its place. You won't find Zapier explaining how Slack helps teams communicate better. The Slack pages list what you can actually do, like start a workflow when someone posts in a channel, send a message to a channel, or look up a user by their email. That information comes straight from Zapier's own integration platform. It's accurate, and a competitor can't copy it without building the same integrations.
The reason it converts is simple. Someone shows up with a problem they want solved today, sees that it can be solved, and is one click away from a free account. That short path from search to using the product is what keeps the whole thing profitable, which I get into in my guide to SaaS Unit Economics.
Case 2: Nomad List's crowdsourced dataset (When it works)
Nomad List took a different route to the same result. Instead of rewriting travel articles, it built pages on top of data from its own community. Each city page pulls together cost of living, internet speed, weather, safety and a lot more, with much of it coming from remote workers who use the site.
Pieter Levels didn't grow the Nomad List by hiring writers to churn out travel posts. He built a database that members keep feeding, and the pages came out of that.
If you search "cost of living in Chiang Mai" or "best cities for remote workers in Portugal," you don't get an essay. You get a page packed with numbers like typical rent, what a coffee costs, internet speeds, air quality, and how other members rate the place.
A few things make this hard to compete with:
- Numbers nobody else has: a general travel site can't just reproduce data that came from Nomad List's own members.
- Data that stays current: members keep adding and updating information, so the pages don't go stale and nobody has to rewrite them by hand.
- Lots of useful ways in: the same data powers pages filtered by weather, budget, region and community rankings.
These pages last because people use them like a tool. They filter, sort and compare cities, which is a pretty good sign the page is doing its job. Google doesn't say exactly how much that kind of behavior counts, but pages people genuinely use tend to pick up links, repeat visits and branded searches over time. If you want to see how I test page formats like this, my notes on Growth Experiments & Analysis go into more detail.
Case 3: The synthetic AI directory collapse (When it backfires)
This is what the bad version looks like. Someone scrapes a list of products, has an AI tool, writes a short summary of each one, and publishes thousands of pages. During Google's 2024 core and spam updates, a lot of AI tool directories built this way lost most of their search traffic. None of their pages had real testing, screenshots or confirmed pricing.
From late 2023 into early 2025, hundreds of people launched "AI tool directories." Almost all of them followed the same recipe. Pull a few thousand product names from Product Hunt or GitHub, run each one through a prompt that writes about 300 words, and hit publish.
I kept an eye on several of these sites, mostly out of curiosity. For the first month or two things looked good. A few of them were getting tens of thousands of impressions a month while Google tried the new pages out.
Then the updates rolled out.
Nothing on the sites I followed suggested anyone had actually tried the tools. There were no screenshots of their own, no checked prices, and no notes from real use. Once Google's updates targeting mass-produced, low-value content arrived, most of those sites lost the majority of their indexed pages and traffic within a few weeks.
The takeaway is uncomfortable but simple. If someone with a basic script and an API key could rebuild your site in a weekend, you haven't built anything that will last. The next update will probably prove it.
The programmatic evaluation framework: 4 questions before building
Before you build a template, ask four questions. Is the data yours? Do people search for these specific things? Can your site get the pages indexed? Does each page lead somewhere in your product? If the answer to any of them is no, write the pages by hand instead.
These are the questions I go through whenever someone brings me a programmatic idea:
Is the data yours?
If it came from Wikipedia or a free CSV that five of your competitors have already downloaded, stop here. You need something only you have, like your own metrics, reviews, integration data or product usage.
Are people searching for these exact things?
Programmatic pages do best on narrow, lower-volume searches from people who are close to buying. If your audience mostly searches broad topics, one solid guide will beat a hundred thin pages. Check this before anyone writes a line of code, or you'll end up with thousands of pages that never rank.
Can your site get these pages indexed?
A new domain with no links and no internal structure will struggle to get thousands of URLs into Google. Some pages will get crawled, and most will sit and wait.
Does each page lead into the product?
Traffic that never turns into active accounts just costs you hosting. Every template needs a natural next step into signing up.
| Audit Factor | Sustainable pSEO (Zapier / Nomad List) | Fragile pSEO (Thin AI Directories) |
|---|---|---|
| Data Source | Their own integration data or records contributed by users | Scraped public listings or AI-written summaries |
| What the Page Adds | A lot: real triggers, field mappings, current numbers | Very little: generic definitions reworded over and over |
| How People Use It | They set things up, apply filters, compare options | They leave quickly and rarely click anything |
| Next Step | Straight into a working product | Generic affiliate links or vague sign-up forms |
| How It Holds Up | Still visible after years of Google updates | Hit hard in the 2024 core and spam updates |
If your pages can't offer your own pricing records, real integration details or benchmarks your users verified, expect trouble. Those are the pages people can't find on Wikipedia, and pages that add nothing new are exactly what Google has been clearing out. If you want to model conversion rates and acquisition costs before building, my SaaS Metrics Guide walks through it.
Indexation limits, crawl budgets, and legal boundaries
Growth mechanics, zero fluff.
Join SaaS founders and operators getting weekly notes and tear-downs on organic acquisition and churn reduction.
The two things most likely to trip you up are getting the pages indexed and staying on the right side of the law. Google is cautious with new sites. Copying protected content, ignoring a site's terms or using brand assets without permission can bring takedown notices or legal letters.
Indexing first. Google has said crawl budget is mostly a concern for very large sites. New sites run into something similar, though, because Google decides how much of a site is worth crawling and indexing based on quality and demand. Put 15,000 templated pages with thin content and no links on a young domain and a lot of them will end up marked "Discovered - currently not indexed" in Search Console. In plain terms, Google knows the pages exist and has decided they can wait.
What works better is publishing in small batches:
Start with the 50 to 100 records you're most confident about. Watch Google Search Console for about four weeks. See how fast those pages get indexed and whether they start getting impressions. Only publish the next batch once the first one is reliably indexed and showing results. It's slower, but it works.
Now the legal side, which people skip far too often. I spent years during my LL.B. and LL.M. studies looking at how copyright, software licenses, scraping restrictions and fair use apply to software marketplaces and to republishing other people's data. A lot of founders assume that if something is visible on a public website, they can scrape it and put it on their own site.
That assumption can get expensive, and the risk comes from more than one place.
Copyright mostly protects the way information is written and presented, not the bare facts. Copying someone's descriptions, reviews, screenshots or page design can get you DMCA takedown notices. In the EU and UK there's also a separate database right that can protect a large part of a database even when the individual facts aren't protected. On top of that, scraping a site against its Terms of Service can lead to a breach of contract claim, and very aggressive scraping can run into computer misuse laws depending on where you are.
Trademarks are a separate issue. In the US, nominative fair use generally lets you use a brand name to say your product works with it, as in "Integrates with Slack." Showing logos, hinting that a brand endorses you, or copying its look goes further than that and can start a dispute, so read each company's brand guidelines first. The safest thing to build on is data you collected yourself, APIs you're licensed to use, or data that's openly licensed or in the public domain. This is general information, not legal advice for your situation. You can find more practical checklists in our SaaS growth resources.
Frequently Asked Questions
How many pages do I need for programmatic SEO to be worth it?
Far fewer than most people think. Fifty to 200 pages built on data only you have can outperform a directory with thousands. Go after searches from people who are ready to buy, and don't worry about the page count.
Can I use AI to generate the text for programmatic landing pages?
You can use AI to format, organize or summarize data you actually have. Where it goes wrong is letting AI make up the descriptions. If every page leans on generated paragraphs instead of checked facts, Google is likely to treat the whole set as low-value content. Build every template on accurate data, and have a person read a sample of pages before anything goes live.
How do search engines detect low-quality programmatic pages?
Google doesn't share the details. What we do know is that it can spot duplicate and near-duplicate pages, group similar pages together, and judge whether a page tells people anything they couldn't already find. Thousands of pages with the same layout and wording, where only one keyword changes and nobody links to or uses them, are very likely to get flagged as mass-produced.
What is the biggest technical mistake in programmatic SEO rollouts?
Putting your whole database live on the first day with no internal links. If you don't have breadcrumbs, category pages and sitemaps split into sections, Google has a hard time finding your deeper pages. Thousands of them can stay unindexed for good.
Conclusion and Actionable Next Steps
Programmatic SEO doesn't save you from doing the work. It moves the work into engineering and data.
If you build a database that genuinely helps people solve a specific problem, templated pages can bring in customers for years, and competitors can't easily copy that. If you scrape public directories or fill pages with generated text, Google will eventually clean them out.
So start small. Build one template for your 30 best records. Watch how fast those pages get indexed, see how people behave on them, and check whether any of them actually sign up. Only add more pages once that first batch has earned it.
Join My Growth Experiments Newsletter
If you're planning your first programmatic project or trying to figure out why your pages won't get indexed, join my newsletter below. Once a month I send a write-up of a SaaS growth experiment, the ranking data behind it, and what I learned from my own tests, including the ones that didn't work. Drop your email below and you'll get the next one straight to your inbox.
