Programmatic SEO without the AI slop
Programmatic SEO has a bad reputation right now and it deserves it. Most of what's being shipped under that label is AI slop — thin, templated, generated, unreviewed pages that exist to spam Google's index and that Google has rightly started penalising. The problem isn't the technique. The problem is the execution. This piece walks through how to do programmatic SEO in 2026 in a way that ranks, doesn't get penalised, and is sustainable for a small team.
What programmatic SEO actually is
Programmatic SEO is the practice of generating many indexable pages from a single content template plus a structured data source. Classic examples: Zapier's 'connect [X] to [Y]' integration pages, Airbnb's location pages, G2's category comparison pages. None of these are AI-generated. They're human-written templates filled with structured data. They rank because each page genuinely answers a specific query that the structured data uniquely positions you to answer.
Why AI changed the math
Until 2022, the bottleneck on programmatic SEO was either having the structured data (most companies don't) or hand-writing the template variations to keep pages from feeling identical (expensive). AI didn't change the structured data part. It changed the variation part — and the editorial review part, and the keyword research part. What used to require a small team of writers can now be done by one editor with a Claude pipeline.
The four-stage pipeline that ranks
Here's the architecture we use for every client. Each stage is a checkpoint with a clear input, a clear output, and a human at the gate.
Stage 1 — keyword research
We don't use volume-based keyword tools. We use Perplexity and Claude to map the actual question landscape in your category — what people search, what they're trying to learn, what their commercial intent is. The output is a topic graph with hundreds of long-tail keywords clustered by intent. Hours per client: ~6.
Stage 2 — page templates
A human editor writes a small set of page templates — usually 3 to 5 — that fit different intent types. Comparison pages, integration pages, how-to pages, location pages, alternative pages. Each template has placeholders for the structured data. This is the part nobody can skip. If your templates are bad, your pages will be bad, no matter how good the AI is at filling them in.
Stage 3 — generation with review gates
Now Claude fills the templates with the structured data, generating one page per intent target. But — and this is the part most teams skip — every generated page passes through a human editor before publication. Not for typo-checking. For taste. The editor's job is to catch the moments when the AI wrote something that's technically correct but that a human would never say. Hours per page: 2–4 minutes of editor time on average.
Stage 4 — monitoring
We deploy a monitoring agent that watches each indexed page for a) ranking changes, b) traffic anomalies, c) Search Console errors. When anything moves, the editor gets a Slack message and decides whether the page needs a refresh. This is what takes programmatic SEO from a one-time push to a compounding asset.
What Google actually penalises
Google has been explicit about this in their helpful content guidance. They penalise content that:
- —Is created primarily for search engines, not humans
- —Provides no original information, expertise, or insight
- —Is duplicative or thin to a degree that doesn't satisfy the query
- —Was generated without human review for quality and accuracy
Every one of these is solvable. Original information comes from your structured data and your category expertise. Quality and human review come from the editor gate. Duplicate avoidance comes from the template variation work. Content for humans comes from writing templates that actually answer questions, not just stuff keywords.
What this looks like in practice
For a typical client, our content engine generates 200–400 indexable pages per month, with an editor spending roughly 20 hours per month on review. That's $4k/mo in retainer fees and produces something that would have cost $25k/mo at a traditional content agency. Rankings start showing in 30–60 days. Material traffic lift in 90 days. Compounding traffic — the actual reason you're doing SEO — in six months.
Done right, programmatic SEO is the closest thing to free distribution a small business can get. Done wrong, it's a fast way to get your domain demoted. The difference is the editor gate. Don't skip the editor gate.