JPEG XL is objectively better than WebP in almost every way - so why are most browsers still ghosting it? And should we start a petition?

A bit of context first. I run a service that caches images from paywalled sites so users don't have to load them fresh on every visit. The overwhelming majority of what we cache is PNG - huge, bloated, uncompressed PNG. Naturally, I started looking into smarter storage and serving strategies, and JPEG XL kept coming up as the obvious answer. The compression gains on PNGs especially are remarkable: you can cut file sizes by 50–60% compared to JPEG with minimal perceptible quality loss at equivalent settings. So the plan seemed straightforward: Convert everything to JXL Detect browser support via the Accept header Serve JPEG as a fallback on the fly for unsupported browsers Here's what the numbers actually looked like: Strategy Total Size Savings Do nothing ~51 GB - WebP Q85 (universal) ~12 GB −39 GB JPEG Q92 (universal) ~21 GB −30 GB JXL d=1 + JPEG fallback ~16 GB / ~5 GB −46 GB (85% of users get 76 KB avg) The JXL route has the best savings on paper - but it means storing two versions of everything, or doing on-the-fly conversion, which adds latency. WebP Q85 just wins. Universally supported (~97–98% of browsers globally), −39 GB in savings, no fallback needed. I hate that this is the conclusion, because JXL is better across most technical dimensions that matter Chrome removed JXL support in Chrome 110 in October 2022 - and that removal was the real killer, given Chrome's ~65% global market share. The stated reasons were actually fourfold: experimental flags shouldn't remain indefinitely; insufficient ecosystem interest; insufficient incremental benefits over existing formats; and maintenance burden reduction. Critics, including engineers from Intel, Adobe, Cloudinary, Meta, and Shopify, disputed all of these claims vigorously in what became one of the most contentious threads in Chromium history. In 2026: Google has reversed course. Chrome 145 (released February 2026) ships with a JPEG XL decoder - currently behind a flag, but back in the codebase for the first tim