HTTP GET requests should never have side effects

Using GET requests to trigger actions (delete, create, modify) causes accidental execution: browser prefetch, search engine crawlers, browser back/forward cache, and link previews all trigger GET requests.

Follow HTTP method semantics: GET is safe (no side effects) and idempotent. POST/PUT/DELETE for state changes. If using GET for mutations: prefetch will trigger deletions, search crawlers will trigger actions, browser preloading creates duplicates. Rails had a famous incident where Google Web Accelerator prefetched all links including 'delete' links.

The HTTP spec defines GET as safe (no side effects). Infrastructure (caches, CDNs, browsers, crawlers) relies on this contract. Breaking it causes unpredictable behavior.