The Data Behind WebP Adoption and Web Performance
Where Did WebP Actually Come From, and Why Does It Keep Winning Benchmarks?
Google published WebP in 2010 as a derivative of the VP8 video codec — the same compression logic used for video keyframes, repackaged for still images. The pitch was straightforward: smaller files at comparable visual quality. What nobody predicted was that it would take over a decade for the format to cross the threshold from "interesting experiment" to "deployed on most major web properties." That delay reveals something useful about how the web actually changes.
As of early 2024, WebP accounts for roughly 34% of all image bytes served on the web, according to data from the HTTP Archive's annual Web Almanac. JPEG still leads at around 43%, but its share has been falling by 3–4 percentage points per year since 2020. PNG's slice has been shrinking faster, down to about 15%, largely because WebP handles lossless compression well enough to replace PNG in many UI-asset contexts.
The Compression Numbers People Actually Quote
Google's original 2010 whitepaper claimed WebP delivered 25–34% smaller files than JPEG at equivalent SSIM scores. Independent testing since then has produced a narrower but still consistent range. Cloudinary's large-scale analysis across millions of real-world images found average savings of 26% over JPEG for photographic content and around 19% over PNG for graphics with transparency.
Those numbers shift depending on the image type. A flat-color illustration or screenshot compresses dramatically better in WebP lossless than in PNG — often 30–40% smaller. A high-ISO photograph with heavy noise is a harder case; the savings narrow to 10–15% because WebP's block-based DCT compression introduces artifacts in noisy regions at the same bitrate where JPEG holds up slightly better perceptually.
Where WebP consistently outperforms is in the mid-quality range — the 70–85 quality bracket most production sites actually use. At quality 80, a JPEG file that measures 120 KB typically encodes to 88–95 KB in WebP at visually indistinguishable quality. Multiply that across a page with 15 product images and you're looking at 375–480 KB shaved from a single page load.
What the Core Web Vitals Data Actually Shows
Chrome User Experience Report (CrUX) data from Q4 2023 gives us real-world performance signals broken down by image format usage. Pages that serve WebP for their Largest Contentful Paint element (the hero image, product photo, or above-the-fold graphic) show measurably better LCP times than equivalent pages serving JPEG — but the effect is more nuanced than a simple "WebP = faster" conclusion.
The median LCP for pages where the LCP element is a WebP image sits at 2.1 seconds on mobile (4G connection), versus 2.6 seconds for JPEG-LCP pages in the same CrUX cohort. That 500ms difference sounds clean, but there's a confounding variable: sites that have adopted WebP are also more likely to have optimized their delivery pipeline in other ways — CDN usage, proper caching headers, image resizing. WebP adoption is partially a proxy for engineering maturity.
Still, even controlling for CDN usage, Google's own PageSpeed research found that format modernization (switching to WebP or AVIF) was the single highest-impact image optimization a site could make, ahead of lazy loading, dimension specification, or compression quality tuning. The reasoning is straightforward: fewer bytes transferred directly reduces the time-to-first-byte for image resources, which feeds into LCP.
Browser Support: The Long Tail Is Finally Gone
The reason WebP adoption lagged for so long came down to Safari. Apple didn't ship WebP support until Safari 14 in September 2020. Before that, any site serving WebP without a JPEG fallback would break for Safari users — and with Safari's 18–20% browser market share, that was a real risk. Developers either maintained two image pipelines or skipped WebP entirely.
As of mid-2024, global WebP support sits at 97.5% of browsers according to caniuse.com data. The remaining 2.5% is almost entirely legacy IE11, older Android WebViews embedded in apps, and some niche enterprise browsers. For most public-facing websites, the `
- Safari 14+ (released Sept 2020) — full WebP support
- Chrome 23+ — supported since 2012
- Firefox 65+ — supported since January 2019
- Edge — supported from the Chromium-based Edge 18 onward
The practical implication is that the engineering overhead of running a WebP pipeline has dropped significantly. You no longer need to make nuanced decisions about which user segments get which format — you serve WebP to virtually everyone and JPEG only as a rarely-triggered fallback.
Adoption Rates Across Industry Segments
HTTP Archive crawl data from January 2024 (covering roughly 9 million desktop URLs and 10 million mobile URLs) shows adoption isn't uniform across sectors. E-commerce sites have the highest WebP uptake — around 54% of product images on the top 10,000 retail domains are served in WebP. This makes sense: product photography is JPEG-heavy, the savings are predictable, and page performance directly affects conversion rates.
News and media sites lag behind at around 28%, despite having large image inventories. The reason is partly CMS friction — legacy media organizations run editorial workflows tied to JPEG delivery, and retooling asset pipelines is an infrastructure project, not a quick switch. Publishing platforms like WordPress (via Imagify, ShortPixel, or the native image editor in WordPress 6.1+) have pushed automatic WebP conversion, which is slowly moving that number.
Interestingly, social media embeds and third-party widgets account for a disproportionate share of JPEG bytes still served — these are images fetched from external domains the site owner doesn't control. Even a fully WebP-optimized site can have JPEG images in its waterfall from embedded content.
AVIF Is Competing, But WebP Holds the Practical Edge Today
AVIF — the still-image format derived from the AV1 video codec — delivers better compression than WebP at high quality levels, typically 15–20% smaller files than WebP at comparable quality. Google's Squoosh tool benchmarks consistently show this advantage, particularly for photographic content above quality 85 where WebP starts to struggle more with fine detail.
But AVIF's encoding speed is a serious production concern. Encoding a single high-resolution image in AVIF can take 5–15 seconds on typical server hardware, compared to under a second for WebP. For sites that generate image variants on-the-fly (like most e-commerce platforms with dynamic cropping), AVIF's encode time makes it impractical without dedicated image processing infrastructure. Cloudflare's Images product and Cloudinary handle this by pre-generating AVIF at upload time, but smaller operations can't easily replicate that.
The current pragmatic answer for most sites: WebP as the primary format, AVIF served where encoding infrastructure allows it and where the images are static (logos, hero banners, editorial photography). HTTP Archive data shows AVIF at roughly 5.4% of image requests as of early 2024 — growing fast from near zero in 2022, but still far behind WebP's 34%.
What the Byte Savings Mean at Scale
The HTTP Archive's Web Almanac 2023 calculated that if every JPEG on the web were converted to WebP at equivalent quality, the total data transferred for images globally would fall by approximately 25–30 petabytes per day. That's an estimate with wide error bars, but it captures the order of magnitude involved.
At the level of an individual site, the impact on bounce rate and conversion is well-documented. Google's case studies with retailers show that 100ms improvements in LCP correlate with 1–2% conversion rate improvements — and format optimization is one of the most reliable paths to shaving 200–400ms off LCP on image-heavy pages.
The Akamai State of Online Retail report from 2023 put a dollar figure on this: for a mid-size e-commerce site doing $10M annual revenue, a 200ms improvement in page load time on mobile is associated with roughly $80,000–$160,000 in incremental annual revenue, depending on the vertical and mobile traffic share.
The Practical Takeaway From the Research
The data converges on a few actionable conclusions:
- WebP adoption is mainstream but not universal — 34% of image bytes served in WebP means the majority of image weight on the web is still in legacy formats, leaving significant optimization headroom for sites that haven't migrated.
- The browser support problem is solved — with 97.5% coverage, there is no meaningful audience risk in defaulting to WebP delivery today.
- LCP impact is real but requires proper implementation — preloading the LCP image, using `fetchpriority="high"`, and ensuring the WebP file is properly sized for the viewport all matter as much as the format choice itself.
- AVIF is worth monitoring but not worth blocking on — serve it where your infrastructure makes it easy, treat WebP as your primary format otherwise.
The most underappreciated finding from the CrUX and HTTP Archive data is how much variation exists within the same format. Two sites both serving WebP can differ by 3x in their image-related LCP times depending on compression settings, lazy-loading implementation, and CDN proximity. Format choice matters, but it's the starting point of image optimization, not the finish line.