What separates programmatic SEO from spam?
The unit economics of usefulness. Spam programmatic swaps city names into identical paragraphs; legitimate programmatic gives each page data that changes the answer — different pricing, different use-case specifics, different honest trade-offs. Google's March 2026 information-gain re-weighting made this distinction enforceable: pages that add nothing rank nothing, at any scale.
The test we apply to every page in this site's /feeds hub: if a buyer with this exact query lands here, do they get numbers, specifics and next steps they wouldn't get from the sibling page? If not, the page merges or dies.
The architecture that keeps quality enforceable
Structured data first: every page is a typed record (topic, intent, answer block, sections, stats, FAQs, related links) in a reviewed data layer — not freeform files that drift. One rendering template enforces the AEO layer mechanically: answer block above the fold, question-form headings, FAQ schema, breadcrumbs, canonical URLs. Interlinking is computed (hub-and-spoke plus siblings), so authority flows by design rather than after-thought.
This site runs exactly this stack — typed JSON records, one template, generated schema — which means quality review happens where it scales: in the data, not in a hundred rendered pages.
- Typed records with required stats/FAQ/answer fields — no empty pages possible
- One template enforcing AEO structure on every variant
- Computed interlinking: hub, siblings, related services
- Slug, title and canonical rules set once, enforced everywhere
Quality gates before and after launch
Pre-launch: minimum unique-content thresholds per page (we target 800+ substantive words including 3–4 citable stats and 4 FAQs), duplication checks across siblings, and a human read of a random sample from every batch. Post-launch: Search Console monitoring for index coverage and 'crawled — currently not indexed' patterns (the canary of thin content), pruning or merging underperformers quarterly, and refresh cycles on pages that win citations.
Scale is a multiplier, not a strategy: it multiplies quality if the gates hold, and multiplies evidence against you if they don't.
Questions we hear about this
There's no number — there's a ratio. Pages with genuine information gain per query: scale freely. The failure signal isn't page count, it's Search Console filling with 'crawled — currently not indexed', which means the model has decided your marginal page adds nothing.
AI can draft within a structured, fact-anchored pipeline — real numbers in, expert review before publish, named accountability on the output. AI spinning freeform variations of one seed article is precisely the spam pattern the 2026 update demoted.
No — consistent templates help users and crawlers. What must be unique is the information: the answer, the numbers, the FAQs. This page and its hundred siblings share one template and zero paragraphs.
Computed, capped and relevant: each page links up to its hub, sideways to 3–5 genuine siblings, and to the service page it feeds. Footer-stuffing hundreds of links per page recreates 2010 spam patterns and dilutes the graph you're trying to build.