Programmatic pages: the unfashionable acquisition channel that still works

A pragmatic guide to building programmatic page systems that don’t trip Google’s thin-content filters, and actually compound. The acquisition channel that everyone declared dead, still quietly producing the best ROI on the internet.

Programmatic SEO went out of fashion sometime around 2022. Google’s helpful content updates were going to kill it. AI-generated content was going to flood every long-tail query. The era of mass-producing comparison pages and city-template pages was over.

Three years later, the actual data tells a different story. Programmatic pages still drive enormous volumes of qualified traffic for the companies who do them well. They still compound. They still convert. What changed isn’t whether the channel works, it’s the bar for execution.

What killed bad programmatic, and why good programmatic survives

The kind of programmatic SEO that died deserved to die. Sites with 50,000 city-template pages where every page was 200 words of templated copy, no unique data, no functional utility — these were correctly downgraded. Google’s updates were precise about which signal they were targeting: thin content that exists only to capture a search query.

What survived — and continues to thrive — is programmatic content that meets a different bar:

Each page contains unique data or computed output a user genuinely couldn’t get elsewhere
The page provides a functional outcome beyond just answering the query
The template generates pages at a defensible scale — hundreds, sometimes low thousands, not tens of thousands
The system is structurally integrated into the site’s topical architecture rather than living as an isolated traffic pipe

This is a higher bar than the 2018 version of programmatic. But it’s also a much, much lower bar than building bespoke pages by hand. The economics still work — they just require thought.

Three programmatic patterns that still compound

Pattern 1: Utility programmatic

The strongest programmatic pattern in 2026 is what I call utility programmatic — pages where each variant is a working tool, parameterised by some attribute.

Example: a base “JSON formatter” tool gets variants for “JSON formatter for X”, “JSON to Y converter”, “JSON validator for Z framework”. Each variant is a real, functional tool with unique output. The query intent is satisfied by interacting with the page, not by reading templated text.

This is what BetterBugs ran to produce 40L+ monthly impressions. The 150 utility assets there are almost entirely a utility-programmatic surface.

Pattern 2: Data programmatic

The second pattern is pages built around unique, computed data. Each page surfaces information that doesn’t exist anywhere else on the web, computed from a dataset you control.

Example: “average response time for [framework] APIs”, computed from a benchmark you ran. Or “common error codes in [tool]”, extracted from your support corpus. The defensibility comes from data scarcity, not editorial quality.

This is harder than utility programmatic — you need a dataset to start with. But the moat is enormous because competitors can’t simply replicate it by writing better content.

Pattern 3: Structured editorial programmatic

The third pattern is what most people think of as programmatic but executed at the highest quality bar: structured editorial pages where the template is a frame for genuine editorial substance.

Comparison pages (“X vs Y”) are the canonical case. They’re programmatic in structure but contain real research, real opinion, and real differentiation. The pages take meaningfully longer to produce than utility or data variants — usually 1–2 hours per page even with strong templates — but they convert disproportionately well.

The indexing traps that kill most programmatic systems

Building the pages is the easy part. Getting them indexed and ranked is where 70% of programmatic projects fail. Three traps to know:

Trap 1: Index bloat

Shipping 2,000 pages at once is almost always wrong, even when the templates are good. Google’s crawl budget for new sites is small; even established sites get rate-limited when a sudden surge of new URLs hits the sitemap.

The pattern that works: ship in waves of 100–300, with 2–3 weeks between waves. Each wave gets crawled, indexed, and starts ranking before the next wave overloads the queue. Patience here is the cheapest performance optimisation available.

Trap 2: Internal linking that doesn’t scale

Every programmatic page needs internal links from outside the programmatic surface itself. Without this, the entire surface gets treated as an isolated subgraph — high-volume but low-authority. The fix is to seed cross-links from your editorial content and from cluster centres into the programmatic variants that are most relevant.

Trap 3: Cannibalisation between variants

If two variants compete for the same query — say “JSON formatter Python” and “JSON formatter for Python developers” — Google will pick one and downgrade the other. Worse, it may downgrade both, deciding the surface is duplicative.

The discipline is having an explicit canonicalisation strategy before generating the surface. Define which variants exist, what they map to, and which queries each one is meant to capture. If two templates produce overlapping variants, merge them.

Operator note
The single best test for whether a programmatic surface is structurally healthy: pick a random variant, run it through a search query, and ask “does this page demonstrably do something the top three results don’t?” If the answer is no, your surface is index bloat waiting to happen.

What changes in the AI search era

Programmatic pages have a counterintuitive advantage in AI retrieval. LLM citation systems prefer sources that demonstrate functional authority — and a working tool or a unique data table demonstrates that more cleanly than a written paragraph claiming expertise.

This means utility programmatic and data programmatic surfaces are particularly well-suited to AI search visibility. The pages get cited not because they argue better, but because they are the answer to the query in a way text can’t replicate.

Structured editorial programmatic (the comparison-page pattern) is the harder case in AI search — it competes more directly with whatever the LLM was going to summarise anyway. Still works, but the moat is thinner.

When not to do programmatic

Some honest signals that you shouldn’t pursue programmatic:

You don’t have a topical architecture yet. Programmatic without architecture produces a fast-growing low-quality surface that drags your entire domain down.
You don’t have engineering capacity. Real utility-programmatic surfaces require code, not just templates. If your team can’t build small tools, the programmatic path will be lower-quality than the equivalent editorial work.
Your category is too small. If your total addressable long-tail query volume is under ~20K monthly searches, the engineering investment doesn’t pay back.
You’re early-stage. Programmatic compounds over 12–24 months. If you need pipeline this quarter, build something else.

For the SaaS companies who pass these filters, programmatic remains one of the highest-leverage moves available. It just looks nothing like the 2018 version. Build for utility, not for templates. Ship in waves, not floods. Integrate with your architecture, don’t treat it as a side-channel. Do this well, and you get an acquisition surface that compounds for years.

Operator note
If you’re a SaaS founder thinking about your acquisition system and want to talk this through, book a call – I take a small number of these per quarter.
-Yash