Bharat AI Saathi: Productised AI-Setup Catalog Site
A static company site with a 1,000-setup India catalog, a 39-system global catalog and a 10,000-guide content pipeline.
Custom design system, build-time currency twin, lawyer-critiqued legal pages, structured data everywhere, an independent AI review loop that scored the site from 28 to 58, and an indexing policy that recovered from an AdSense rejection.
- 1,000 India setups + 39 global systems rendered from one content model
- Independent model review loop raised the score 28 to 58
- Full structured-data and legal-page coverage, lawyer-critiqued terms
- Noindex policy keyed to curated unique content after a review rejection
Bharat AI Saathi sells productised AI setups (WhatsApp receptionists, document bots, lead qualifiers) to Indian SMEs across 20 services and 50 business types, plus a USD catalog of 39 systems for a global audience. The site is generated by a Python build that reads JSON content and writes a complete static tree: service, industry and nested setup pages, a how-we-work page with locally generated images, a founder page, FAQ, seven legal documents and thousands of long-form guides.
The design system, called Laterite, uses warm paper neutrals, one accent, hairlines instead of shadows and light/dark tokens, with a rupee/dollar switch computed at build time. Quality was enforced by machines rather than taste: an independent review loop fetched the live URLs with a separate model and wrote scored reports (28 for the old one-pager, 41 with placeholders, 58 after they were resolved), a dist-wide link crawler, one-H1-per-page checks, BreadcrumbList/Product/FAQ/Organization JSON-LD, llms.txt and IndexNow submission.
When the first review flagged 4,400 machine-generated guides as low value, the build gained a noindex policy keyed to a curated set of unique setups, cutting indexable pages to about 1,000 while keeping the rest reachable.
How it works
- Content JSON and catalog YAML define services, industries and setups
- Guide generator writes long-form articles in batches with a strict prompt
- Build script renders every page, sitemap part and OG image to a dist folder
- Link crawler and SEO checks gate the build
- Zip parts deploy to hosting; CDN cache is flushed and IndexNow pinged
- Review script scores the live site and feeds the next fix list