How We Built 24 Programmatic SEO Pages From 1,093 Verified Companies

Programmatic SEO has a reputation problem. Most pSEO pages are thin wrappers around AI-generated filler: “[City] plumbers,” “[Keyword] in [State],” templated text with a find-and-replace on the location name. They rank briefly, then get nuked in the next useful-content update. Visitors land, see the template, and bounce.

There is a better way. If the data is the moat and the pages are just formatting, the SEO value comes from the dataset, not the template. That is what we built this week: 24 pages across two languages, generated from 1,093 live-verified public companies in 11 countries. Zero AI-generated text in the data layer. Every row has a ticker, exchange, and source attribution.

Here is how we built it, what we learned, and what we would do differently.

The data: verified, not scraped

The foundation is a dataset of 1,093 public companies in 11 countries, pulled from index constituent lists (Nasdaq Stockholm, Nikkei 225, FTSE 100, CAC 40, DAX 40, IBEX 35, OBX, OMX Helsinki 25, SMI 20, TA-35/TA-125, MOEX). Every ticker was live-verified against quotes on 2026-08-30.

The CSV has seven columns: country, index, company, ticker, exchange, shortname, verification, source. The “verification” column is the audit trail. The “source” column links back to the public index list.

This matters because Google’s useful-content system rewards pages where the data layer has independent verification signals. A page that says “377 companies, verified 2026-08-30 against Nasdaq Stockholm” carries more weight than one that says “list of Swedish companies” with no source.

The build: one data module, two languages, 24 pages

We use Astro and host on Cloudflare Pages. The architecture is simple:

src/
  lib/
    nations.ts          ← 1,093 rows, hardcoded from CSV (one source of truth)
  pages/
    robotics-stocks/
      index.astro       ← EN hub (11 country cards)
      [slug].astro      ← EN subpage per country (dynamic route)
    sv/
      robotics-stocks/
        index.astro     ← SV hub
        [slug].astro    ← SV subpage

The nations.ts module exports an array of NationEntry objects and a NATIONS_MAP grouped by country slug. Both language routes import from the same module. There is no translation pipeline: the Swedish pages use the same company names and tickers (universal) and translate only the page chrome (labels, methodology, CTAs).

This gives us 24 pages total:

• EN hub + 11 country subpages
• SV hub + 11 country subpages

Each page carries:

1. A data table with ticker, company, exchange (12 sample rows, enough for SEO context without giving away the full dataset).
2. Dataset JSON-LD (schema.org/Dataset) with row count, verification date, license, and distribution link.
3. Hreflang alternates: <link rel="alternate" hreflang="sv"> on every EN page and vice versa. Both canonicalize to themselves.
4. A CTA to the full dataset on Gumroad (the revenue path).

What worked

Hreflang from a single data module. Both language routes consume the same nations.ts. The only difference is which labels and chrome strings get rendered. This eliminates the most common pSEO translation bug: data drift between language versions. There is one source of truth for the company data; the pages are views over it.

Sample tables as conversion levers. Showing 12 of 377 Swedish companies creates a natural information gap. The visitor sees the structure, trusts the verification, and wonders what they are missing. The CTA answers that question. Showing all 377 rows removes the gap and the conversion reason.

JSON-LD on every page. Google’s dataset schema is underused in programmatic SEO. Most builders skip it because their data is scraped or AI-generated and would fail validation. Verified data makes the schema credible. Each page claims @type: Dataset with a datePublished, license, and distribution link. These are real values, not placeholders.

Swedish as a low-competition lane. Swedish-language robotics/nation pages have near-zero competition compared to English. The /sv/robotics-stocks/ mirror is the sleeper lane: same data, different audience, less noise.

What we would change

The Swedish translation needs a native rewrite. An LLM-generated Swedish pass sounds literal and archaic. A native speaker can fix the flow in an hour. We should have budgeted that time before deploying.

The sitemap regeneration is currently manual. We run astro build and the sitemap auto-generates from the page tree, but there is no pre-deploy verification step. A CI check that compares expected page count (hub + N subpages) against the built sitemap would catch missing pages.

Data freshness decays. Every week, a few tickers change, a few companies delist. A 5% random re-verification batch against live quotes, run monthly, would keep the “verified” claim honest. We have not automated that yet.

The numbers

Metric	Value
Pages generated	24 (12 EN + 12 SV)
Data rows	1,093
Countries	11
Sitemap URLs	60 (incl. all other site pages)
Build time	< 30 seconds (Astro SSG)
Deploy	git push → Cloudflare Pages auto-deploy
Time from CSV to live	~4 hours (incl. hreflang + JSON-LD + CTA)

The generator

We open-sourced the generator that builds these pages from any CSV. It is a single Python script with zero dependencies:

python3 pseo-gen.py data.csv --category nation --slug-col slug \
    --title-col country --data-cols ticker:company:exchange \
    --hub-title "Robotics Stocks by Country" \
    --lang en --out-dir src/pages/robotics-stocks/

The script handles hub page generation, per-entity subpages, JSON-LD dataset markup, dual-language labels (EN/SV), and configurable sample row counts. It is part of the vhsgreed SEO toolkit.

The takeaway

Programmatic SEO works when the data is the product and the pages are just formatting. Verified data beats AI-generated filler because it survives useful-content updates, earns dataset schema credibility, and converts visitors who recognize real research behind the template.

The full 1,093-company dataset is available on Gumroad. If you have a dataset and want pages built from it, we do that too.

---

Published 2026-08-31 by vhsgreed. The dataset, generator, and audit checklist are part of the SEO toolkit on GitHub.