AI Product Content Optimizer for Shopify

Shopify merchants with 100+ products face an impossible choice. Either accept generic product descriptions that hurt conversion and SEO, or spend weeks rewriting each one manually. The AI tools that exist either generate slop that sounds nothing like the brand, or dump changes directly to the storefront with no review.

Beyond basic rewrites, there's a newer problem. AI search engines (Perplexity, ChatGPT, Google's AI overviews) rank content differently than traditional search. Product pages that rank fine on Google get skipped entirely by AI answer engines. Most SEO tools are still optimizing for the old game.

Merchants needed three things existing tools don't deliver together: brand voice fidelity, a review step before anything goes live, and scoring that accounts for how AI search engines evaluate content.

I built an embedded Shopify app with a six-factor SEO scoring engine, including a proprietary weight for AI-search readiness. Merchants get scored results on titles, descriptions, meta tags, and alt text, with specific recommendations for what to improve.

The brand voice system is the core differentiator. Merchants define their voice through examples and constraints. Protected regions (brand names, taglines, specific phrases) never get rewritten. The AI rewrites in the merchant's voice, not a generic e-commerce template.

Every rewrite goes into a draft queue. Nothing touches the live store until the merchant approves. One-click rollback if something does get pushed that shouldn't have.

The architecture needed to handle variable AI costs, multi-tenant isolation, and rate limiting per shop. Supabase with RLS on shop_id keeps tenants isolated. Upstash Redis handles sliding-window rate limits and caches entitlements. Every Claude API call logs cost in microdollars to an audit table, so margin analysis is always accurate.