One defect on 71 of 72 sites
Most platform pages report shares. This one leads with a near-absolute, so it is worth stating
the denominator first: 72 Divi sites, audited by us, on the snapshot date printed with the
figures below. That is not a sample of the web, and 72 is not a large number.
71 of those 72 sites shipped a stylesheet chain that blocks the first paint, and 68 of
them also shipped scripts that block the parser. On plain WordPress the same two defects sit
at 84% and 85% across 560 audited sites, which is high and still leaves room for a site to be
built otherwise. On Divi we have measured one site without the blocking stylesheet
chain, out of 72.
That is a property of how the builder assembles a page — not of how carefully your site was
built. Sections, rows and modules each carry their own styles and their own behaviour, and
the page they compose asks the browser for all of it at once.
What the builder puts in your head
A Divi page arrives as a stack of full-width sections. The visual weight of the first one is
usually a background image set in CSS — and that is the part that costs you. A CSS background
is invisible in the markup, so the browser discovers your hero only after the stylesheet has
been fetched, parsed and applied. On 81% of the sites we audited, the largest image on the
first screen was being found that late.
The rest of the head is volume. Divi generates per-module styles, your child theme adds its
own, plugins add theirs, and the result is a long list of small files. Small files are cheap in
bytes and expensive in round trips. On a phone the round trips are the bill.
None of this is reachable from the Builder. You can remove modules, and you can switch on
Divi’s performance settings, but the order in which a browser is allowed to use what it
received is a property of the response, not of the page you designed.
The share that does not gain
Across 37 Divi audits with a verified before and after (as of September 2026), the mobile
score rose without costing more than two points on desktop on 62% of them. On 24% the
mobile score went down. The median mobile gain was +3 points, and visible loading speed
(Speed Index, how quickly the page fills in) came out 1.17× faster.
That spread is not a Divi trait. A larger study, now repeated on
3,816 sites, found that what a site gains depends far
more on where it started than on what it runs: sites in the 30s and 40s gained a median of 13
points, sites in the 70s and 80s gained 3, and sites already above 90 were a coin flip. If your Divi site is slow, the numbers on this
page are the relevant ones. If it is already fast, the honest answer is that there is less here
for you.
Three things a proxy cannot reach
A slow server. We can serve a cached copy of a page quickly — but a request that takes
your host two seconds to compute still takes two seconds. That is a hosting conversation, and Divi
sites on shared hosting have it often.
Requests you asked for. The sites in our audit base load from a median of
12 hosts they don’t own. We defer third-party
scripts so they stop competing with your layout. We do not remove them, because which pixels
and widgets your business runs is not our call.
A layout that draws itself with JavaScript. A minority of Divi builds paint the first
screen after load rather than in the response. Deferring anything there moves the problem
instead of solving it, and we switch that optimization off for the domain rather than publish a
number that means something else.
What the audit reports back
It measures your site twice — as it is now, and served through WebSpeed — on mobile and
desktop, and lists every change that produced the second number. It takes a few minutes and needs only
the address. If your site turns out to have little to gain, the report says that too, which was
the outcome for about a quarter of the Divi sites we have measured.