Autonomous Website Content Audit
A recurring audit of Blenheim Chalcot's and ClickZ's websites, built with Claude Code. Scores each site out of 100 across 50 metrics — from broken links and stale claims to how well content is optimised for AI/LLM search (GEO) — then suggests a specific action per metric and the score improvement it would deliver.
Corporate websites accumulate stale and contradictory content across dozens of pages — and most teams have no consistent way to score how a site is actually performing, for human readers or for AI search engines.
One instruction gets decomposed by the agent itself: the site is split into logical sections, parallel research agents comb each one, and everything is scored out of 100 across 50 metrics. Each metric comes with a specific suggested action and the score improvement it would deliver — a human decides what to action, rather than the agent pushing changes unsupervised. Runs monthly on both sites.
Surfaced real cross-page contradictions, broken links and stale compliance-relevant language a manual pass would likely have missed, with every finding scored and prioritised rather than left as a flat list. Less a linter for websites, more an auditor that reads the whole site as one coherent story, scores it, and says exactly what to fix first.