By JABstractPublished October 15, 2026
How we tested
Every site was measured the same way. Lighthouse 13, the test built into Google Chrome, run three times on each page, phone and desktop, with the middle result kept. The WordPress version was tested on the live domain before we touched it, and the rebuilt version on the same domain after cutover. One exception: the landscaping site’s WordPress version was tested at its own server rather than through its CDN cache, which is noted on the results page along with every test number for every page.
Three numbers matter most to a customer on a phone: how many seconds until the main part of the page shows (Lighthouse calls this Largest Contentful Paint), Google’s 0 to 100 speed score, and how much the phone has to download. Those are the three in the table. The rebuilt numbers were re-tested on October 14, 2026, after the fixes described below.
The home page on a phone, six sites
| Site | Shows up in | Speed score | Page size |
|---|---|---|---|
| JABstract (our own site) | 21.1 s → 1.9 s | 54 → 99 | 29.8 MB → 131 KB |
| Budget Roofing Service, Winter Haven | 30.8 s → 3.1 s | 55 → 92 | 8.7 MB → 1.1 MB |
| Slumberr, kids party rentals | 28.1 s → 3.7 s | 58 → 86 | 8.3 MB → 371 KB |
| Preferred Dumpster, Winter Haven | 13.7 s → 3.4 s | 51 → 88 | 4.0 MB → 405 KB |
| A Central Florida landscaping company | 11.3 s → 3.3 s | 70 → 91 | 2.5 MB → 422 KB |
| J&A Stone Designs, Auburndale and Lakeland | 9.4 s → 2.9 s | 62 → 90 | 3.1 MB → 705 KB |
Sorted by how much faster each home page shows up, the six sites run from 3.2 times faster (the stone showroom) to 11.1 times faster (ours). The middle of that range is 5.8 times, which is the honest headline. A mean would be flattered by our own site and we don’t use it.
What moved most, and why
The biggest change on every site is page weight. A WordPress page built with a page builder loads the builder’s code, every plugin’s code, a slider, a chat widget and full-size photos before it shows you a phone number. Our own old home page downloaded 29.8 MB. The rebuilt version downloads 131 KB, because the page is built ahead of time into plain HTML and the images are resized to the space they fill.
Speed score follows weight. All six sites now score 86 or better on a phone, and four score 90 or better, which Google shows as green. On desktop all six score 99 or 100.
Readiness checks moved too. We run nine checks on every site, covering search crawlers, analytics, security headers, business schema and Lighthouse’s Agentic Browsing audits. The landscaping site passed one of nine before and nine of nine after. Our own site and the stone showroom site went from five to nine. The other three went from six to eight. All six pass every Agentic Browsing audit Lighthouse runs on them.
What we got wrong at first, and how we fixed it
Our first results had two weak rows. The landscaping site showed up in 8.5 seconds with a speed score of 72, and the roofing site in 9.2 seconds, also scoring 72. Re-testing every site this week showed something else: two sites that had launched fast, the dumpster site and the stone showroom, had slipped into the 70s.
The cause was the same everywhere. Google’s tracking scripts, 150 to 350 KB of them, were loading before the page had drawn anything. Sites that held those scripts back until the visitor tapped, scrolled or had been on the page for four seconds were scoring 90 to 96; sites that loaded them up front were scoring in the 70s. We moved every site to the first pattern, deferred one other script that was holding up the first paint, and stopped one site’s menu from downloading 2.3 MB of thumbnail photos nobody had asked to see. Tracking still records every visit and every lead; it just waits its turn.
After the fix, the landscaping site went from 72 to 91 and the roofing site from 72 to 92. We publish this because a results page with only wins is not evidence of anything, and because it is the most useful thing in this post if you run your own site: anything that loads before your page is drawn costs you customers on a weak signal.
What this means if your site is on WordPress
You can run the same test on your own site today. Open PageSpeed Insights, paste your address, and look at the phone score and the Largest Contentful Paint time. If your home page takes more than four seconds to show its main content, you are losing the customers on the weakest signal, and that is most of them on a job site.
Speed is also the first thing an AI assistant notices. ChatGPT, Gemini and Google’s AI Mode read websites to recommend a business, and a page that is slow, buried in scripts and short on plain text is easy to skip. Our guide What Is an Agentic-Ready Website? explains the three jobs a site has to do for software, and how to check yours in ten minutes.
If you would rather have it done, the agentic-ready rebuild is the same work we did on these six sites: your existing pages rebuilt on our platform, an interactive mock-up you review before anything goes live, and your own before-and-after report. Your place in the build queue is public, so you can see when your build starts.
The fine print on the numbers
- Lab tests, not field data. Real visitors on a weak signal will see longer times on every version of every site.
- Median of three runs, same pages, same device settings. Mobile uses Lighthouse’s simulated slow 4G connection. Before numbers were taken before each rebuild; the rebuilt numbers were re-taken on October 14, 2026.
- Sites are named with the owners’ agreement; they are the same sites listed on our results page.
- Nothing here is a ranking claim. A faster, readable site removes technical reasons to be skipped. It does not promise a position in Google or in an assistant’s answer.