Merchants assume a single favicon file covers every browser and device, but the accessibility and mail-industry data used here to illustrate measurement practice shows how often published specs drift from what actually gets checked, so a favicon guide that only lists one pixel size is already out of date.
Most merchants treat a favicon as a single square image dropped into a theme folder and forgotten, and that assumption is exactly what causes the icon to look broken on half the surfaces it appears on. The truth is that no favicon spec exists in isolation from the browser, the device, and the markup around it, and the only way to know whether a site’s small details are correct at scale is to measure a large sample of real homepages and see what actually holds up. That is what the WebAIM Million project does for accessibility markup broadly, and its numbers are useful here not because they mention favicons directly but because they show, precisely, how a “settled” web standard drifts over time even when nobody announces a change.
The WebAIM Million dataset covers one million home pages and reports a WCAG conformance rate of 95.9%, which is up from a prior reading of 94.8%. That is a real increase, and it looks like progress until you set it against the second number from the same crawl: the count of detected errors on those same pages increased 10.1%. Both things happened together. Conformance rose and errors rose. That is not a contradiction once you accept that a bigger, more thorough crawl finds more of everything, both passes and failures, and it is the same caution that applies to any claim about favicon files being “handled” by a theme or a platform by default. A conformance percentage without an error count next to it tells only half the story.
Nothing in the data reviewed for this article isolates favicon markup as its own line item. What is measured, at the same scale, is how often other small, easy details get skipped. Low contrast text appears on 83.9% of the pages in the WebAIM Million sample. Missing alternative text for images sits at 53.1%. Missing form input labels sits at 51%. None of those three numbers is about favicons, but all three describe the same category of problem a favicon falls into: a detail that is cheap to fix, invisible until it fails, and skipped on more than half of live commercial pages anyway. If a site cannot consistently label a form input, there is no reason to assume it is checking its icon file across every browser surface either.
USPS Notice 123 is a useful contrast piece precisely because it is not about the web at all. It is a postal rate and package size schedule, and it states its adjustments in exact, checkable numbers: authorized publications receive a 5% rate factor, and a nonadvertising adjustment factor of 1% applies for each qualifying measure. Nobody reading that document has to guess what the numbers mean or when they apply, because the document states them plainly and ties them to specific categories. That is the standard a favicon specification should meet and often does not, since so much favicon guidance circulating online repeats old pixel dimensions without saying which browser, which year, or which document they came from.
| Source | Measure | Value |
|---|---|---|
| WebAIM Million | WCAG conformance | 95.9% |
| WebAIM Million | Prior conformance reading | 94.8% |
| WebAIM Million | Increase in detected errors | 10.1% |
| WebAIM Million | Low contrast text | 83.9% |
| WebAIM Million | Missing alt text for images | 53.1% |
| WebAIM Million | Missing form input labels | 51% |
| USPS Notice 123 | Authorized publication rate factor | 5% |
| USPS Notice 123 | Nonadvertising adjustment factor | 1% |
The WebAIM figures come from a crawl of one million home pages, checked as loaded on September 9, 2026, and the postal figures come from USPS Notice 123 as loaded the same date. Reading the two side by side is the point: one dataset tracks how often the web gets small details wrong at scale, and the other shows how a rate document states its own adjustments without ambiguity.
Treat any favicon guidance you find the way you would treat an unsourced accessibility claim: check the date it was published and check whether a newer measurement has moved the number since. Until a source publishes an actual measured favicon dimension, the safest move is to test the current file directly in the browsers and devices that matter to the store rather than trust a remembered pixel size.
There is no single measured favicon pixel dimension in the data reviewed for this piece, so nothing here specifies one. What is measured is how consistently sites keep basic markup correct at scale, and the WebAIM Million sample shows WCAG conformance at 95.9%, up from 94.8%. Treat any specific favicon pixel number you see elsewhere as a claim to verify against the actual document that published it, the same way every number in this article traces back to a named source.
The facts available here do not break out favicon markup as its own error category. What is measured, across the WebAIM Million homepages, is a 10.1% increase in detected errors overall and low contrast text present on 83.9% of pages. A favicon problem would fall under general markup or link issues, not a category the current data isolates on its own.
Very often, based on the scale checked. Missing alternative text for images shows up on 53.1% of the pages in the WebAIM Million sample, and missing form input labels on 51%. Those are not favicon-specific but they are the closest measured proxy for how easily small markup details, including icon references, get overlooked across a million homepages.
Not for favicons specifically. The only adjustment figures available here come from USPS Notice 123, where authorized publications receive a 5% rate factor and a nonadvertising adjustment factor of 1% applies for each qualifying measure. Those numbers describe postal rate mechanics, not web asset formats, so they are cited here only to show how a published rate sheet states its adjustments precisely, which is the standard a favicon spec sheet should also meet.
Because the only measured facts available for this piece come from the WebAIM Million accessibility crawl and from USPS Notice 123, and the rule for this article is that every number printed must trace to a fact already on record. No favicon pixel dimension appears in either source, so none is invented. What the two sources do establish is a model for how a specification should be read: check the conformance rate, check the date it was measured, and check whether the number moved since the last read.