Why Your Insurance Website Fails Google's Speed Test

Why it passes on your laptop

It passes because you almost certainly tested it on the wrong device: Google's speed tools check your site twice, once simulating a mid-range phone and once on hardware closer to your own laptop, and only the phone version is the one most of your prospects actually see. Here is the scenario, almost word for word, from a dozen agency owners we have talked to. You check your own website. It opens fast, the quote form loads, everything looks clean. Then a client mentions the site felt slow on their phone in the parking lot before an appointment, or you run it through a speed checker on a whim and the mobile number comes back in the 40s or 50s while desktop sits in the 90s. Nothing changed on the site between those two tests. What changed is which version of "your site" got tested.

This is not a rare glitch. It is the predictable result of testing a site on the one device that almost never struggles with it, your own workstation, instead of the one Google actually uses to decide how your site indexes and ranks. Since the shift to mobile-first indexing, Google's Search Central documentation states plainly that it "uses the mobile version of a site's content, crawled with the smartphone agent, for indexing and ranking"¹⁴. Your desktop experience is not the primary version anymore, in Google's own systems. The mobile rendering of your site is.

Google's own speed tooling does not test your site once. It tests it twice, once simulating a mid-range Android phone on a throttled connection, and once on hardware closer to what you are actually sitting in front of. Google's PageSpeed Insights documentation is specific about this: the lab test uses "predetermined network and CPU throttling settings" run against fixed device specifications, historically a Moto G4, in a Google data center, precisely so the test reflects a realistic phone rather than your Chrome browser on an office connection⁷. Your laptop, on your office wifi, is not a fair proxy for that test. It is closer to the desktop condition, which is the easier of the two.

There is a second reason the gap shows up specifically on mobile. Chrome UX Report, the dataset behind both PageSpeed Insights' field data and Search Console's Core Web Vitals report, requires a page to be "publicly discoverable" and "sufficiently popular" to be included at all, and it reports mobile and desktop as separate populations of real visits, not a blended average⁶. A site that is fine on desktop and weak on mobile is not an edge case in that data. It is one of the more common shapes a report takes, because the two platforms fail for different reasons: desktop machines are fast enough to absorb a lot of inefficiency, and phones are not.

What Core Web Vitals measure

Core Web Vitals is Google's name for three specific measurements, and Google's Search Central documentation states plainly that they "are used by our ranking systems," alongside other page experience signals¹. None of the three is exotic once you see what each one is actually counting.

Largest Contentful Paint (LCP) is the time until the biggest visible element on the page, usually your hero image or headline, actually renders. It is loading speed, measured the way a visitor experiences it: not when the server responds, but when something worth looking at shows up. Interaction to Next Paint (INP) measures responsiveness across the entire visit: how long the page takes to visibly react every time someone taps a button, opens the mobile menu, or fills in a form field, not just on the first click¹. Cumulative Layout Shift (CLS) measures visual stability, whether the page jumps around while it loads, which is the experience of trying to tap "Get a Quote" and hitting an ad or a different button because the layout moved under your thumb.

Every one of those thresholds is measured "at the 75th percentile of page loads, segmented across mobile and desktop devices," per Google's own methodology⁶. That detail matters more than it sounds like it should. A 75th percentile threshold means three out of every four of your real visits need to hit that number, not your best case, not an average that a handful of fast visits can drag up. One good test on your own phone on wifi proves nothing about the other 75% of visits Google is actually grading.

Why mobile is graded on its own

It is worth being precise about why Google bothers to split this at all instead of reporting one blended score. Search Console's Core Web Vitals report shows mobile and desktop as separate tables. PageSpeed Insights runs them as two separate tests you have to trigger independently. The Chrome UX Report, the real-user dataset behind both, segments every metric by device category because a phone on a cellular connection and a desktop on broadband are not comparable environments, and blending them would hide exactly the gap that matters most for a prospect standing in a parking lot with one bar of signal⁶.

Google is explicit that Core Web Vitals sit inside a broader page experience signal, and that "there is no single signal" deciding a ranking, since core ranking systems weigh a variety of factors alongside it². Speed will not out-rank a page with thin, irrelevant content. What Google's documentation does say is that when two pages are otherwise close on relevance and quality, page experience "can contribute to success in Search"². For an agency competing against three or four other local producers whose content covers similar ground, that tiebreaker condition is closer to the normal case than the exception.

And the tiebreaker only fires on the metric that actually reflects your real traffic mix. Pew Research's most recent mobile fact sheet, drawn from a survey of 5,022 U.S. adults run from February through June 2025, puts smartphone ownership at 91% of Americans⁸. A prospect comparing Medicare Advantage plans or an ACA subsidy estimate at ten o'clock at night is doing it from a phone far more often than from a desk. If your mobile score is the one nobody is watching, it is also the one deciding what the median prospect actually experiences.

It is worth naming the specific moment where this bites hardest for an insurance site: the quote or intake form. A brochure page that loads two seconds slower is annoying. A quote form that stutters while someone is entering a ZIP code and a birthdate, mid-thought, on a phone held in one hand, is the moment an agency actually loses the visitor it paid to acquire or waited months to earn organically. That page deserves its own speed check, not just the homepage, for exactly that reason.

Why page builder sites run slower

They run slower because the browser has to assemble the page after it arrives, instead of receiving a finished one. That difference is architectural, not a setting someone forgot to flip, and it is worth explaining plainly rather than waving at.

A static site, the kind we build on Astro, renders the page to finished HTML at build time. When a phone requests the page, the server hands back a complete document. The browser paints what it received. There is no framework to download and boot before the content shows up, because Astro ships zero JavaScript to the browser by default. A page builder like GoHighLevel or a WordPress site running a visual page builder works the other way around: the browser downloads a JavaScript bundle, boots a rendering framework, fetches the page's content and styling, and only then assembles and paints what the visitor sees. Every one of those extra steps costs time, and on a phone on a slower connection, the cost compounds instead of staying fixed.

You do not have to take our word for the weight of that default. GoHighLevel's own help documentation, last updated March 26, 2026, describes a recent internal optimization where the platform "removed unnecessary fields from the page data response," cutting "the average final HTML document size by about 10%" to improve load performance⁹. That same article recommends agency users limit animations and widgets, avoid stacking third party tracking scripts, and remove unused page sections to help their own pages load faster⁹. A platform does not publish that advice, or ship an internal optimization to shrink its own HTML output, unless the default weight was real enough to be worth fixing on their end too.

There is a second, quieter cost to the page builder model that has nothing to do with the framework itself: who is allowed to touch the page. A drag and drop builder makes it easy for anyone on the team to add a testimonial carousel, a countdown banner, or a chat widget from the app marketplace, and each addition is a small, individually reasonable decision. Eighteen months later the homepage is carrying a dozen widgets nobody remembers approving, each one adding its own script to the load. A static site does not remove human judgment from the process, but it does mean every addition goes through a build step where its weight is visible before it ships, rather than accumulating silently inside a visual editor.

None of this means a page builder site is doomed to fail Core Web Vitals forever, and it is worth saying so plainly. GoHighLevel's own changelog shows real engineering work going into render-blocking CSS and initial payload size on their end. A disciplined editor who compresses every image, limits third party scripts, and keeps pages lean can get a page builder site into a respectable range. What that editor cannot do through settings alone is remove the JavaScript boot step the platform's architecture requires before the page paints. That is the ceiling, not the floor, and it is the reason two sites with equally careful content teams can land in very different places on a mobile test.

What slow mobile speed costs you

None of this is abstract once you put a number on it. Google commissioned 55 and Deloitte Digital to study mobile site speed across 37 leading brand sites, tracking over 30 million user sessions hour by hour across 30 days¹⁰. The headline finding: a 0.1 second improvement in mobile load time increased retail conversion rates by 8.4% and average order value by 9.2%¹⁰. The lead generation breakdown is the more relevant number for a quote form. Moving visitors from the homepage to the first step of a form improved by 5.5% for every 0.1 second gained, and completions from that first step through to a finished submission improved by 21.6%¹⁰. That is not traffic you would have to go acquire. It is the traffic you already have, converting at a higher rate because fewer people gave up waiting.

The other cost is quieter and shows up before a visitor ever converts. Semrush's Q1 2025 State of Search data puts zero-click search traffic at 27.2% of US searches, up from 24.4% in March 2024¹¹, meaning a rising share of buyers never leave the search results page at all, and answer engines are doing more of the reading on a page instead of a human clicking through. Google's own crawl budget documentation is direct about how site speed factors into that reading: "If Google can load and render your pages faster, we might be able to read more content from your site"¹². A slow, script-heavy page is not just losing human visitors to impatience. It is more expensive for the same crawlers feeding AI answers to process, at the exact moment that channel is taking a larger share of how buyers find an agent.

Put a small number on it so the lift feels concrete instead of abstract. Say your quote page gets 500 mobile visits in a typical month, and 120 of those visitors start filling out the form. At the study's measured 21.6% lift in completions per 0.1 second of mobile speed gained¹⁰, closing just one 0.1 second gap would be worth roughly 26 additional completed forms a month from that same traffic, no additional ad spend, no new content, nothing except the page responding faster than it did before. That is an illustration built from the study's measured rate applied to a round hypothetical visit count, not a promise about your specific site, since the actual lift depends on where your page starts and how your visitors behave. The direction of the effect is the part that is not in question.

If you want to know where your own site stands rather than estimate it, the free Audit scores this alongside your AI citation readiness in about a minute. Run the free Audit.

Check your own score in 5 minutes

You do not need to guess at any of this. Here is the actual check, in order.

Do this on your phone's actual mobile data at some point too, not just wifi, since PageSpeed Insights simulates a connection but nothing replaces watching your own quote form load on a real cell signal. If you have never seen your site stall on a weak connection, you have not seen it the way a meaningful share of your prospects have.

What you can fix, and what you can't

If you are running a page builder site today, you are not stuck, and it is worth being honest about what real work can and cannot buy you. None of the fixes below require touching code or switching platforms. They require the kind of housekeeping that gets skipped when a site has five years of accumulated widgets, funnels, and half-finished sections nobody has cleaned out since the agency was smaller.

You can do all of this yourself, this week, without touching a line of code. Most agents who try it get a real improvement and a mobile score that moves from the 40s into the 60s or 70s. Some read this list, decide the DIY path is worth the afternoon, and stop there, and that is a perfectly reasonable place to land if your site is otherwise working for you. Others get that far, watch the score plateau below 90 no matter what they trim, and realize the remaining gap is the architecture itself, not a setting they haven't found yet.

How we build differently

This is the problem we build agency sites to avoid from the start rather than optimize around later. We build on Astro, which ships zero JavaScript to the browser by default, so a phone receives a finished page instead of a framework to boot. Every site we ship is served from the edge on Cloudflare, close to wherever the visitor actually is. That architecture is why our builds score 90 or above on Lighthouse performance for both mobile and desktop, and why a number of pages score 100.

The speed work is only half the build. Every site also ships with a sitemap.xml, an llms.txt, and a robots.txt that regenerate themselves the moment a new page or post goes live, with the full absolute URL, so nothing has to be remembered or hand maintained. The robots.txt is written to welcome every crawler, including the named AI crawlers, rather than quietly blocking the traffic that would cite you. Every page carries the schema stack an answer engine actually reads: Organization, Service, Article, FAQPage, and BreadcrumbList, matched to what is really on the page. If you want the fuller picture of what changed in how agencies get cited, we wrote about that separately in what actually moved in AI search this year.

Digital Foundation starts at $247 a month for a complete, compliant website with AEO optimization, Google Business Profile management, and an AI chat widget. The Pro tier, $497 a month, adds a weekly blog post, a weekly location page, and a 24/7 AI receptionist. The Scale tier, $997 a month, moves the publishing cadence to one new blog post every business day. All three tiers include a 14-day free trial and month to month billing, live in days, no long term contract¹³.

What a fast foundation gets you

Put the pieces together and the payoff is concrete rather than theoretical. A phone visitor who lands on a page that paints in under 2.5 seconds and responds the instant they tap something is a visitor who is still there when your quote form asks for a ZIP code. If you run paid traffic, that same speed gain is the one Google's own research ties directly to more completed lead forms from the same ad spend¹⁰. And a site a crawler can read cheaply and quickly is a site more likely to be the one an AI answer actually cites when a prospect asks who handles Medicare enrollment in their county, rather than a competitor's slower page the crawler gave up on halfway through.

None of that is a guarantee of a ranking, a lead count, or a citation. It is the foundation those outcomes are built on, and it is the one part of the stack that is entirely within your control before you spend a dollar on ads or content.

It also compounds in a direction most agency owners do not think about until it happens to them: every new blog post, location page, or quoter you add sits on top of the same foundation. On a slow, script-heavy platform, each addition is one more thing competing for the same limited render budget in a visitor's browser. On a static site, each addition is just another finished page, no heavier to load than the one before it. The foundation is the part of this decision that gets more expensive to change the longer you wait, because there is more content sitting on top of it every month you publish.

Questions agencies ask