// Short answer
The problem
The client wanted a sports news site that stayed fresh daily without a newsroom — and without tripping Google's penalties for AI-generated, scaled content. The hard part isn't generating text; it's generating accurate, publishable content safely, on a schedule, with a human still in control.
What we built
- A content engine: one model drafts from structured data (facts and stats), a second model judges for quality and factual drift before a human ever sees it.
- A human approve-to-publish gate — nothing goes live without one click, which keeps the site clear of scaled-content-abuse penalties.
- Data-grounded player profiles generated from live football data, not invented stats.
- Automated social fan-out across multiple platforms on publish.
- Satellite SEO microsites and a Google News sitemap with article schema to accelerate indexing.
The stack
Next.js and Tailwind for the site, Supabase for data, Inngest for the scheduled jobs (five recurring crons handle fixtures, rosters, digests, and recaps), and a Claude content engine for drafting and judging. Generation runs on a subscription rather than per-call billing, which is why the marginal cost of new content is effectively zero.
The hard part: safety, not generation
The engineering that matters here is the guardrails. Articles are written from facts, never scraped-and-rewritten — the exact pattern Google penalizes. A second model checks every draft for factual drift (it caught and rejected cases where the first model miscalculated a stat). And a human approves everything before publish. That's the difference between an automation that quietly works and one that gets a site deindexed.
Outcome
The site went live on its target date, reached 93% Google indexing during domain warm-up, and the client reported ranking on the first page for several competitive searches. It runs its own scheduled jobs, drafts its own content, and waits for a human — exactly the kind of system we build: one that removes the daily work and keeps running.
Frequently asked questions
- What is Atlas Lions?
- Atlas Lions (atlaslions.com) is an automated sports content site for fans of the Morocco national team. A Claude-powered engine drafts articles from real data, a second model judges them for quality and factual accuracy, and a human approves before anything publishes.
- How much content did the system produce?
- 38 auto-drafted articles and 26 data-grounded player profiles, all published behind the human approval gate. The site reached 93% indexing on Google during its domain warm-up.
- How does it avoid AI content penalties?
- Every piece passes a human approve-to-publish gate — nothing auto-publishes. Articles are written from structured data (facts and stats), not scraped-and-rewritten RSS, which is what Google penalizes as scaled content abuse. A second model judges for factual drift before a human ever sees it.
- What did it cost to run?
- Near-zero marginal cost. Generation runs on a subscription rather than per-call API billing, so producing new articles and profiles costs effectively nothing beyond hosting.
Want a system that runs itself?
Book a 30-minute intro call. We'll scope what we'd automate first — zero pitch, no cost.
// Keep reading