Page weight and rankings: what the Backlinko study actually found

The idea that a bloated page drags down its own rankings is one of the oldest pieces of technical SEO folklore. It has an intuitive story behind it: heavier pages load slower, slower pages frustrate users, and Google is said to reward what users like. The largest publicly available correlation study to test that story directly found nothing.

What the study measured#

Backlinko, working with Ahrefs as a data partner, pulled ranking and on-page data for 11.8 million Google search results and ran correlation analysis across a long list of on-page and off-page factors. Two of those factors sit directly on the question this piece is asking: page HTML size, and page loading speed. The published article states the result for size under the heading “Page HTML Size Has No Relationship With Rankings,” and reports “we found no correlation between page size and rankings.” For loading speed, the finding is the same in kind: “Overall, we found zero correlation between site speed and Google rankings,” with first-page results averaging 1.65 seconds to load.

What it does not measure#

Two gaps matter for anyone tempted to read this as settling the mobile page-weight question. The study reports HTML byte size, not the full transferred weight of a page once images, scripts, fonts, and third-party tags are counted, so it is a narrower measurement than “total page weight.” It also does not report a count of HTTP requests per page anywhere in the public write-up, so there is no sourced basis here for a claim about request count and rankings, even though request count is a common companion metric to byte size in page-speed audits. Anyone citing this study for either of those two specific claims would be extending it past what it actually measured.

Why a null correlation is still worth reporting#

A finding of “no correlation” is not the same as “no effect never happens,” and the study’s own authors are explicit about that limit: “As this is a correlation study, it’s impossible to determine the underlying reason” behind any pattern in the data. An aggregate correlation across 11.8 million results, spanning every niche and query type Ahrefs’ index covers, can wash out an effect that is real but confined to a narrower slice, such as e-commerce category pages competing on load speed against each other specifically. The honest reading is that at this scale, page size and page speed are not acting as a broad, direct ranking lever, not that no page has ever been held back by its own weight.

What this means for prioritizing the work#

Do not sell a page-weight reduction project on the promise of a ranking jump; the largest public correlation study on the question found none. Do sell it on what it actually protects: Core Web Vitals pass rates, conversion rate, crawl efficiency on large sites, and the experience of users on constrained mobile connections, none of which need a ranking correlation to be worth fixing on their own.