Legacy Image Formats, and the Credit You May Already Have
Four images on our own site are still PNG files, and the report lists them as legacy image formats. The finding is called images using legacy formats, and it is one of the few where the sensible first question is whether you need to do anything at all.
The card counts the raster images served in an older format and names the breakdown, so one glance tells you whether you are looking at two files or forty.
It does not say whether a modern format appears anywhere on the page, and that detail decides how much this is worth doing.
Nothing about the flag is a criticism of your site. Most WordPress media libraries are years old, and every image uploaded before WebP became routine is still sitting there in the format it arrived in.
Where the flag draws the line
The check looks at the format of every image the page loads. PNG, JPEG and GIF count as legacy. WebP and AVIF do not, and SVG never enters the count because it is not a raster image at all.
A single old image is enough to raise the finding, and it stays at low severity below eight. From eight upward the report marks it medium, which is where a page built on a large photo library usually lands.
One site we test serves sixteen JPEG files and a GIF, with only two WebP images on the page. That is a medium severity flag and a genuine amount of avoidable transfer on every visit.
Our own site sits at the other end. Four PNG files, eight WebP images, low severity. Same finding, very different amount of work behind it.
When legacy image formats cost you nothing
The performance score includes a four point credit for serving modern image formats. It is one of only two things that adjust the performance score directly, and it is the generous one, because it is awarded on a single image.
One WebP anywhere on the page earns the whole credit. The check does not measure how much of your library is modern, it asks whether the page serves a modern format at all.
Your own report answers that question quickly. Its waterfall lists every image request by URL, so looking for webp in that list settles whether the credit is already yours.
That makes the arithmetic easy. If your site already serves WebP, the four points are banked, and converting the remaining PNG files will not move your score by a single point.
If your site serves none at all, the picture changes. Converting even one image earns the credit, so the first conversion is worth four points and everything after it is worth the bandwidth.
Either way the finding is advisory. It points at room to improve rather than a failure, which is why it can sit on a report with a healthy performance score and no contradiction in sight.
What the conversion actually saves
The saving depends on which format you are leaving. Google’s own figures for WebP put lossless images about 26 percent smaller than PNG, and lossy images 25 to 34 percent smaller than a comparable JPEG at matched quality.
Transparency matters here too. Lossy WebP still supports it, and Google measures that combination at roughly three times smaller than the equivalent PNG, which is where a logo or a product cut-out stops hurting.
Those percentages apply to the image itself, so the page level saving depends on how much of your weight is pictures. On a photo heavy page it is the single easiest reduction available.
The performance effect is mostly on Largest Contentful Paint, the metric that times the biggest element on the screen. Smaller image bytes arrive sooner, so the main picture lands earlier, and on a phone on a slow connection that gap is the whole experience.
Weighing the effort against the gain
This is an optimisation rather than a repair. Nothing is broken when a page serves a PNG, and if your images are already sensibly sized, converting them will not change how the page feels.
The gain scales with how many pictures you serve. Swapping a dozen small thumbnails saves a rounding error, while converting the large photographs that dominate a page is where the transfer actually falls.
That is why the finding lands differently from site to site. An online store with product photography on every page has real weight to win back. A blog or a small business page with one or two images will barely notice the difference.
So weigh the effort against the return before starting. If the report also shows a heavy total page weight, the conversion is worth the afternoon. If it names four small files and the page loads well, leave them alone.
Converting without breaking the design
On WordPress, let the conversion happen for you. The media library accepts WebP uploads directly, and an image optimisation plugin can convert what is already there or serve a modern format from the original file.
That second option is the one to look for, because it leaves your library alone. The plugin keeps the PNG and hands the browser a WebP copy, so nothing in your content breaks if you later turn it off.
Check whether your CDN already does it. Several edge networks will convert images on the way out and cache the modern version, which means one setting instead of a library wide cleanup. Cloudflare calls this Polish, and it is a toggle on the dashboard.
A CDN conversion is the cheapest route to try first, because it is reversible. Turn it on, test the page, and turn it off if a picture looks wrong.
Know when PNG is still the right answer. Sharp edged graphics, screenshots with text in them, and anything needing lossless transparency can look worse after a lossy conversion. Convert a copy and compare before replacing the original.
Logos and icons are the exception worth acting on regardless. An SVG is smaller than either raster format, scales cleanly on any screen, and removes the file from this count entirely.
For a hand built site, the conversion belongs in the build. An image pipeline can produce WebP at the sizes the design actually uses and drop the originals, which is maintainer work rather than something you do by hand in a dashboard.
One caution before you convert anything. Shrinking the format does nothing about an image that is simply too large, so if the report also flags an oversized individual asset, fix the dimensions first and the format second.
Checking the next report
Run another test and the card prints the new breakdown, so you can watch the PNG count fall and the WebP count rise in the same line of text.
If the modern count is already above zero, the score will not change and that is expected rather than a broken test. The report-reading walkthrough covers how the performance score is assembled if the credit is unclear.
Aim for the images people actually see. The hero picture, the product shots, the ones sitting above the fold. Converting the last thumbnail in a gallery is tidier than it is useful.
