SEO Operations
AI metadata jobs with validation, per-item results and recovery.

Operational PHP/Python workflow
Client project · Fine Touch Marketing
Med-spa marketing agency. SEO metadata automation across their clients' WordPress sites.
THE ENGINEERING
A protected operator interface queues WordPress metadata work through a PHP/Python runtime. Title normalization, industry and location constraints, deterministic repair and per-item results govern AIOSEO updates. September recovery work added supervised restart and boot recovery, separate frontend/API health checks and fault-injection verification. Twelve scoped recovery criteria passed in the retained record. Later headline, quality and article features were deployed, but their paid end-to-end generation test still needs a usable WordPress test target. The historical n8n implementation is separate from the current runtime.
BEFORE → AFTER
Real problems from this build, what I changed and the verified result.
- CASE 01Before
Interface availability and background-worker health are separate operational signals.
AfterTwelve scoped recovery criteria passed in the retained September record.
How I fixed itAdded separate frontend/API checks and supervised restart and boot recovery.
- CASE 02Before
Titles failed to match because the site and the operator encoded the same characters differently.
AfterUpdates land on the intended page instead of silently skipping.
How I fixed itNormalized both sides with one shared contract covering entities, punctuation and dash variants.
- CASE 03Before
AI examples leaked the wrong business context, and output sometimes hardcoded a city.
AfterCopy stays on-industry and location-correct, checked per item.
How I fixed itLocked the prompt to the supplied industry, derived the city per page, and added deterministic repair after generation.
Evidence reviewed: September 4 recovery records · October 10 public case study