Hand-coded hreflang vs plugins vs edge injection: failure modes compared

Hand-coded hreflang vs plugins vs edge injection: failure modes compared

Three common implementation tactics, three distinct ways they rot. We tracked which broke under what conditions.

Observations:

— Hand-coded tags give full control and are easiest to audit, but scale terribly. Beyond a few dozen URLs, humans miss return tags. We found reciprocity error rates climbing sharply past ~50 manually maintained pages.
— CMS plugins (Yoast-style multilingual, Polylang, Weglot) automate reciprocity, which kills the most common error class. Their weakness is opacity — when a plugin emits a wrong region code or self-referencing canonical conflict, debugging means reading generated output, not source.
— Edge or CDN injection (Cloudflare Workers, reverse proxy) decouples hreflang from the CMS entirely. Powerful for headless or multi-platform stacks, but adds a layer that audits and ordinary developers cannot see, and that no log review catches without specifically looking.

The data suggests the right tool tracks team capability more than site size. A team that cannot reliably edit a CDN worker should not own hreflang at the edge.

Our heuristic: plugins for single-CMS sites, edge injection only when you genuinely have multiple frontends, hand-coding only for small or unusually disciplined teams. Limitation — plugin behavior varies by version; treat specifics as snapshots.