A look back, written in 2026. In January 2018 Google announced that page speed would count towards mobile rankings from July. For the next six months, every client call seemed to start with a screenshot of a red PageSpeed score.
What was actually announced
Less than the panic suggested. Google said the Speed Update would only affect pages delivering the slowest experience to users, and only a small percentage of queries. A relevant page that was a bit slow would still rank. It was a tie-breaker at the bottom end, not a new league table.
The same year, mobile-first indexing started rolling out, and that mattered more. If the phone version of a site was thinner than the desktop one — fewer words, missing pages, hidden navigation — that was the version Google would increasingly judge. Plenty of sites had exactly that problem and nobody was screenshotting it.
The score became the goal
The trouble with a single number out of a hundred is that people start optimising the number. I saw sites stripped of useful images, fonts and tracking just to turn the circle green, while the thing that actually annoyed visitors — a slider that pushed the content down while it loaded, a pop-up in the first second — stayed where it was.
A lab score is a test run on one simulated device. It is a useful diagnostic and a poor target. The question is what real visitors on real phones experience, and in 2018 very few small sites were measuring that at all.
What was worth fixing
Images, nearly always first. Photographs uploaded straight from a camera, three thousand pixels wide, displayed in a box three hundred wide. Resizing and compressing them was the cheapest, largest win on almost every WordPress site I looked at that year.
Then plugins. A small business site with forty plugins loading scripts on every page, most of them for features used on one page or none. Removing the unused ones did more than any caching layer.
Then hosting. Cheap shared hosting that took a second before it sent the first byte. No amount of front-end work gets that second back.
What was not worth it
Rebuilding a working site in a new framework because it would be faster. Chasing the last ten points of a lab score. And, for most small firms, the separate stripped-down mobile pages that some were being sold that year as the answer — a second version of the site to keep in step with the first.
What stayed with me
Speed is a user problem first and a ranking problem a distant second. A page that loads quickly on a mid-range phone on a weak signal keeps people who would otherwise have gone back to the results. That was true before the announcement and it was true after it.
The useful thing the 2018 panic did was get clients to look at their sites on a phone. Most of the fixes followed from that, and none of them needed a green circle to justify them.
