We fetched 207 Canadian dental homepages as an iPhone and measured what they send. The median page transfers about 929 KB, which is reasonable. The distribution behind it is not: the 90th percentile is 6.3 MB, the 95th is 16.3 MB, and the heaviest single page sends 56 MB. 8.7% of these clinics autoplay a background video, and the median autoplay video is 13.1 MB — with the largest at 54.9 MB. On an iPhone in Low Power Mode those videos are widely reported not to play at all, which means the patient pays for the download and sees a still frame.
Why we measured this
Because we got it wrong on our own site first. Our hero carried a 5.9MB autoplay background video that looked fine in every desktop browser we tested at a phone viewport, and did not play on a real iPhone. It had been downloading and decoding the whole time — the bytes were spent, the motion never arrived.
That is a specific enough failure that we wanted to know how common it is. So we fetched every homepage in our 213-clinic Canadian dataset with an iPhone user agent, listed the assets each one references, and asked each asset how big it is.
What a Canadian dental homepage weighs
The median is 929 KB across 207 reachable sites. For a page with photographs on it that is a perfectly ordinary number, and most of this market is fine.
The tail is where it goes wrong. The 75th percentile is 1.9 MB, the 90th is 6.3 MB, the 95th is 16.3 MB and the heaviest page sends 56 MB to a phone. 15.9% of these homepages transfer more than 3 MB and 12.6% more than 5 MB.
A patient on a mobile connection outside a city does not experience the median. They experience whichever clinic they tapped.
The autoplay video problem
27.5% of these homepages contain a video element or background-video markup of some kind. 19.8% carry an autoplay attribute. 8.7% — 18 clinics — autoplay a video file we could measure.
Those 18 files have a median size of 13.1 MB. The largest is 54.9 MB. For comparison, the one we just removed from our own mobile hero was 5.9 MB, which puts us comfortably under the median and still too heavy to defend.
The part that makes this worse than ordinary bloat is that the download and the effect are not linked. Apple documents that a video must be muted and on-screen to autoplay at all. Beyond that, developers consistently report that iOS in Low Power Mode replaces autoplay with a play button, and the WebKit issue raised about it was closed as intended behaviour. Apple does not document that one, so we state it as the widely-observed behaviour it is rather than as policy — but it matches what we saw on a real handset with Low Power Mode on.
Low Data Mode, Reduced Motion and a slow connection each produce their own version of the same outcome. A background video is therefore the one design element that bills every visitor and performs for only some of them.
We overstated it first, and the correction is the useful part
Our first pass reported a median homepage of 977 KB and a heaviest page of 85 MB, counting every video file at full size. Then we opened the 85 MB one.
It is a promotional video with a controls attribute and preload="metadata" — a click-to-play film that a browser fetches a few kilobytes of and leaves alone. Counting it as transferred weight was simply wrong. The same error applied across every site with a click-to-play video, so we recalculated with non-autoplay video bytes excluded, and the figures in this article are the corrected ones.
This is worth saying out loud because it is the standard failure of automated page-weight audits, including the ones sold to dental practices. A tool that lists asset sizes without checking whether the browser requests them will hand a clinic a frightening number and an invoice. Ask any audit that flags a heavy video whether the attribute says autoplay or preload="none" before paying anyone to fix it.
What we would actually do about it
If the video autoplays, replace it on mobile. Not compress it — replace it. A CSS animation keeps running in Low Power Mode, costs nothing to download, and honours prefers-reduced-motion, which a video does not. That is what we moved our own hero to, and it removed 5.9MB from every phone visit.
If the video is click-to-play with controls and preload set to none or metadata, leave it alone. It is not costing anything and an audit flagging it is wrong.
And check the actual transfer rather than the asset list. The gap between those two numbers was the whole difference between our first draft of this article and this one.
- Autoplay background video on mobile: remove it, and use CSS for the motion.
- Click-to-play video with preload="none" or "metadata": no action needed.
- Over 3 MB of real transfer: look at images before anything else — median image payload here is 487 KB.
- Any audit quoting asset sizes: confirm the browser actually requests them.
Limits
We listed up to 25 assets per homepage and measured each by its reported content length, so these figures are a floor for sites with more than 25 assets, and they ignore compression applied in transit for text assets.
We measured what the page references, not a real page load. We did not run a browser against 207 sites, so lazy-loaded images below the fold and assets injected by script are not counted — another reason to read these as a floor.
We did not verify on a physical iPhone that any specific clinic video fails to autoplay. We verified that on our own site and measured file sizes on theirs; the Low Power Mode behaviour is reported by developers rather than documented by Apple.
6 of the 213 clinics did not respond.
What 207 Canadian dental homepages send to a phone (ScaleUp Dental, November 2026). Non-autoplay video excluded, since browsers do not fetch it.
| Percentile | Homepage transfer |
|---|---|
| 25th | 509 KB |
| Median | 929 KB |
| 75th | 1.9 MB |
| 90th | 6.3 MB |
| 95th | 16.3 MB |
| Heaviest | 56 MB |
FAQ
Under about 1 MB of real transfer puts you at the median of the Canadian market, which is a reasonable target for a page with photographs. The problem is not the median — it is the tail. 15.9% of the clinics we measured send more than 3 MB and 12.6% more than 5 MB, almost always because of an autoplay video or uncompressed images.
Not one that autoplays on mobile. 8.7% of Canadian dental homepages do, at a median of 13.1 MB, and the visitor pays for that download whether or not it plays. iOS requires a video to be muted and visible to autoplay, and developers widely report that Low Power Mode replaces autoplay with a play button regardless. A CSS animation gives the same impression, runs in every power state and downloads nothing.
Less than it is sold as doing. Our own audit of these clinics found technical health scores correlated with organic traffic at minus 0.00 while keyword coverage correlated at plus 0.88. Weight matters for whether a patient on a weak connection waits for the page, which is a conversion question, not a ranking one.
Because most audits list asset sizes without checking whether the browser requests them. A click-to-play video with preload="metadata" fetches a few kilobytes, not its full size. We made this exact mistake in the first pass of this study and reported an 85 MB page that in reality transfers under 2 MB. Check the attribute before paying anyone to fix it.
Remove any autoplay video from mobile first — it is usually larger than everything else combined. Then images: the median image payload across these homepages was 487 KB, and most of it is uncompressed JPEG or PNG that would be a fraction of the size as WebP.
