Programmatic SEO Methodology: How We Built 3,949 Indexable URLs That Actually Rank

What programmatic SEO actually is (in one paragraph)

Programmatic SEO is the practice of generating large numbers of high-quality landing pages from a structured dataset and a small number of templates, then targeting the long tail of search demand at scale. The discipline is older than people think — Yelp, TripAdvisor, Zillow, G2, and Wikipedia are all programmatic SEO sites at their core. What's new in 2026 is that LLMs have made the underlying data enrichment and per-page customization step radically cheaper, which has caused both an explosion in good programmatic SEO and a much larger explosion in bad programmatic SEO that gets deindexed.

The single failure mode that defines whether a programmatic build succeeds or fails is per-page substance. A site can have 50,000 indexed URLs and earn meaningful traffic, or 50,000 URLs and earn nothing — the variable is whether each page actually contains something a real user searching that specific query would value. Everything in this playbook exists to enforce that constraint.

Our own site as the case study: 3,949 URLs across 26 sub-sitemaps

We don't write theoretical programmatic SEO content. The site you're reading is a programmatic SEO build that runs the playbook end to end — and we publish the sitemap so you can verify it.

The clusters break down roughly as follows — and the cluster taxonomy itself is the playbook. Each cluster targets a different search intent, uses a different template, and is monitored separately so we can deindex underperforming clusters without losing the rest.

Three of those clusters (best, canada-city-industry, legal-cities) account for roughly 1,500 of the 3,949 URLs and drive a disproportionate share of long-tail capture. Three other clusters (research, comparisons, answers) are smaller but produce most of the AEO citations and inbound links. The methodology below is what determined that mix.

The 7-step programmatic SEO playbook we run on every engagement

The data model: how we structure entities for programmatic generation

The single most useful framing we've found is that every programmatic page should map to one entity in a graph. Cities, industries, services, schema types, comparison-target tools, glossary terms — each is an entity with named attributes, related entities, and stable identifiers.

On this site, the data model is roughly:

• City entity

  Name, province, population, market characteristics, neighborhoods, top industries, GBP density, our written market commentary.
• Industry entity

  Name, search-volume signal, key sub-segments, regulatory notes, distinctive SEO challenges, named tactics that work.
• Service entity

  Name, scope of work, deliverables, pricing band, ideal-client archetype, related services.
• Schema-type entity

  Name, schema.org URL, recommended properties, common implementation mistakes, validator behavior.
• Glossary-term entity

  Term, plain-English definition, technical definition, related terms, named examples.
• Comparison-target entity (tools/agencies)

  Name, category, named strengths, named gaps, pricing posture, ideal use case, our verdict.

A page like /canada/toronto/seo-for-dental-practices is the join of three entities: City × Industry × Service. Each contributes its own attributes to the rendered page. Because each entity carries genuinely distinct content, the joined page is genuinely distinct — which is the whole quality bar.

Templates: the difference between thin and durable

The most common programmatic SEO mistake we see in audits is treating templates as Mad Libs — one paragraph of body copy with three nouns swapped per page. That pattern produces tens of thousands of near-duplicate pages and gets the entire cluster deindexed by Google's site-quality systems within two to three quarters.

A durable template is a structural skeleton with per-entity slots that produce real, asymmetric differentiation between pages. The skeleton is shared. The content under each H2 is genuinely different per entity because it pulls from genuinely different per-entity attributes.

The test we apply to every template before it ships: print two random rendered pages side by side. If a senior SEO can't tell within five seconds why each page deserves to exist independently, the template fails and gets rewritten. Most of the templated 'SEO content' shipped in 2024 and 2025 fails this test trivially.

Indexation strategy: why most programmatic sites get pruned

Google does not index every URL it discovers. On a 50,000-page programmatic site, indexation rates of 30–60% are common, and the unindexed pages are usually the lowest-quality cluster. The mistake is treating low indexation as a crawler problem; it's nearly always a quality-signal problem.

Our default indexation strategy on every programmatic engagement:

• Index every page that can defend itself

  Pages that pass the quality gate ship indexable, in the sitemap, with a self-canonical.
• Noindex (don't 404) borderline pages

  Cluster pages that exist for completeness but don't yet have substance — noindex,follow with a planned content upgrade. Removes them from quality assessment without dropping internal-link equity.
• 404 obvious zero-value pages

  Pages whose existence is harder to defend than not — 404 with a 410 if the URL has been seen by Google. The sitemap-pruned count on this site is 1,338. Most are deliberate.
• Quarterly consolidation review

  Cannibalization audit per cluster. Two pages competing for the same query get consolidated to the stronger URL with a 301.
• Search Console saved views per cluster

  Indexation, impressions, CTR, average position — tracked separately for each programmatic cluster so deteriorating clusters get found in week one, not quarter four.

The asymmetric trade is this: removing a weak page is nearly always a net win for the cluster's overall ranking. A small reduction in 'reach' is dramatically outweighed by the quality-signal lift to the surviving pages. Most agencies don't do this because it's psychologically uncomfortable to delete pages they billed for. We do it because it's how programmatic SEO actually works.

The quality gate: 12 rules every programmatic page must pass

A programmatic page can ship to the production sitemap on this site only if it passes every rule below. Pages that fail are flagged in the build output and either rewritten or held back. This list is roughly the most useful artifact this guide will produce for you — copy it, adapt it, enforce it in CI.

• 1. Unique H1 derived from real entity attributes

  Not 'SEO for {city}' but a specific noun phrase that exists nowhere else on the site.
• 2. Self-canonical with a stable, lowercased URL

  No trailing-slash inconsistency, no UTM in canonical, no parameterized variants competing.
• 3. ≥ 700 words of unique-per-page content

  Not 700 words of swapped boilerplate — 700 words drawn from genuinely per-entity data.
• 4. ≥ 3 entity-specific data points above the fold

  Local market data, named competitors, regulatory notes — visible in the first viewport.
• 5. Hub up-link + ≥ 5 sibling links

  Internal linking that survives Google's render. No JavaScript-only links to programmatic pages.
• 6. JSON-LD schema appropriate to the page type

  LocalBusiness, Service, Product, ItemList, FAQPage — whichever fits. Validated, no errors.
• 7. Image with descriptive alt + width/height

  No layout shift, no decorative-only stock photos.
• 8. Title + meta description specific to the entity

  Not duplicated against any other URL on the site, length within bounds.
• 9. ≥ 1 outbound citation to an authoritative source

  Government data, industry body, primary research. Citations earn citations.
• 10. Page passes Core Web Vitals on the production template

  LCP < 2.5s, CLS < 0.1, INP < 200ms. Tested on the actual template before scaling.
• 11. Last-updated date present and honest

  Either the dateModified is real or the page is being misrepresented. Lying about freshness is detected.
• 12. Defensible reason to exist, in writing

  Each cluster has a one-paragraph rationale. If we can't write the rationale, we don't ship the cluster.

Anti-patterns we refuse to ship

The programmatic SEO patterns below produce short-term traffic spikes and predictable medium-term collapses. We've audited dozens of post-collapse sites and the failure modes are remarkably consistent.

Measuring success: the four metrics that matter

A programmatic SEO build produces a lot of dashboards. Most of them are vanity. The four metrics we report on monthly for every programmatic engagement:

If the indexed-URL ratio is below 50%, the cluster needs pruning. If cluster-level CTR is below 1%, the title and meta need rewriting. If average position is flat, the templates need depth, not the dataset more rows. The trap most operators fall into is responding to soft metrics by adding pages; the actual fix is almost always to deepen the existing pages.

The honest summary

Programmatic SEO done well is one of the highest-leverage tactics in modern SEO. Done badly it's the fastest path to a deindexation event we know of. The difference is per-page substance, enforced by a quality gate and a willingness to prune. The site you're reading runs the playbook end to end — 3,949 URLs in production, 1,338 deliberately pruned along the way, the entire cluster taxonomy public.

If you're considering commissioning a programmatic build, bring your dataset and your intent map. We'll review the architecture honestly and tell you whether programmatic is the right tool for your situation, before you invest in tens of thousands of URLs you'll spend the next year deleting.

Frequently Asked Questions

The questions Canadian operators ask most often about programmatic SEO before deciding whether to commission a build.