Conversations about site speed tend to end at one of two extremes. One side believes rankings will not come until the PageSpeed score hits 100. The other side says Core Web Vitals never changed anything. When you look at what Google has actually said, and at the independent studies published since 2020, both sides turn out to be wrong.
In this article I bring that evidence together in date order, checked against the original sources. Along the way I add short notes from what I have seen on real sites over the years.
What Google has said since 2020
Google announced Core Web Vitals in May 2020. The page experience update reached mobile search in the summer of 2021 and desktop in February and March 2022. Read carefully, the statements since then follow a consistent line:
- May 2020: page experience joins the hundreds of signals used in ranking. A great page experience does not override great content, but when several pages have similar content, page experience becomes much more important for visibility.
- May 2021, Google I/O: Philip Walton from the Chrome team said that improving LCP from 2 seconds to 1 second will not increase rankings, while fixing a very slow page could. Pages also do not need all three metrics to be “good” to benefit.
- August 2021: John Mueller: “It is a ranking factor, and it’s more than a tie-breaker, but it also doesn’t replace relevance.”
- April 2023: Danny Sullivan wrote that there is no single page experience signal and that it was never a separate ranking system. Good Core Web Vitals stats “don’t guarantee good rankings” (Google Search Central, April 2023).
- February 2024: Sullivan said on X that Google does not confirm page experience or Core Web Vitals as a “direct ranking factor”. This is the one statement that breaks the line.
- March 2024: the documentation was clarified: “Core Web Vitals are used by our ranking systems.” Other page experience aspects, such as HTTPS and mobile-friendliness, were described as not directly helping a site rank higher. On the same day, INP replaced FID.
- October 2024: Mueller: Core Web Vitals “are not giant factors in ranking”, and he doubted a site would see a big drop just because of them (Search Engine Roundtable).
- Today: the page experience documentation, last updated in September 2026, still says Core Web Vitals are used in ranking. It also says Google will show the most relevant content even if its page experience is sub-par, and that chasing a perfect score just for SEO “may not be the best use of your time”.
So for six years Google has said the same thing in different words: the signal exists, it is small, and relevance comes first. Google has never published a weight for it, and I do not expect it to.
What the signal actually measures
Much of the debate is about the wrong number. The 0–100 performance score in PageSpeed Insights is not what Google uses for ranking. That score comes from Lighthouse, a lab test: one device, one network, one moment. The ranking signal uses field data from real Chrome users, collected in the Chrome UX Report (CrUX).
This has several practical consequences:
- Thresholds. LCP within 2.5 seconds, INP under 200 milliseconds, CLS under 0.1. A page counts as “good” when it meets these targets at the 75th percentile of page loads, measured separately for mobile and desktop. The data covers the previous 28 days, so a fix takes about a month to show fully in the reports.
- Mobile and desktop are separate. A page can pass on desktop and fail on mobile. A page that scores 95 in a lab test on a fast machine can still fail in the field if your visitors use mid-range phones.
- Evaluation is mostly page-level. Google says its systems “generally evaluate content on a page-specific basis”, with some site-wide assessments. URL groups in Search Console are a reporting convenience. Google has not said that rankings use the group score.
- Most pages have no field data. This is the least discussed point. In a 2022 crawl of 43.7 million pages, Ahrefs found Core Web Vitals data for only about 12% of them. Google does not say what happens in ranking when a page has no CrUX data. What is clear is that, for a large share of the web, this signal has no value to measure.
- INP raised the bar. FID only measured the delay before the first interaction, and almost every site passed it. In the February 2024 CrUX dataset, 96.2% of origins had good FID but only 80.6% had good INP. The share passing all three metrics fell overnight from 48.8% to 45.6% (CrUX release notes, March 2024). The web did not get slower. The measure got stricter.
What the independent studies found
Since 2020, several SEO tool providers and agencies have tried to measure the link between Core Web Vitals and rankings. The most cited ones:
- Screaming Frog (August 2020): 2,500 keywords, about 20,000 URLs. Only 12–13% of results passed; at position 1 the pass rate was 19–20%. Important detail: this study was done before the update.
- Moz (2021): 10,000 keywords, top 20 results. URLs passing all three metrics ranked on average 0.39 positions higher. But the same gap was 0.38 in May, before the update. The clearer change was at the bottom: URLs failing all three metrics dropped about 1.15 positions, against about 0.2 for all URLs with data.
- SISTRIX (September 2021): during the rollout, domains passing Core Web Vitals grew their visibility by 3.7%, against 2.7% for all domains. The author’s own note: correlation does not mean causality.
- Advanced Web Ranking (2022): about 3 million pages. 39% passed, with a slightly higher pass rate in the top positions.
- Whitespark (local search): about 4,200 business landing pages. Better Core Web Vitals did not lead to higher local pack rankings.
The methods differ, but the conclusion is the same: pages in the top positions pass a little more often, and the gap is small. For me the most telling number is Moz’s. When Google switched the signal on, the gap went from 0.38 to 0.39. It barely moved. Screaming Frog shows the same thing from another angle: the top results were already faster before the update. Most of the link between speed and rankings seems to come not from the signal, but from the fact that well-run sites tend to be fast anyway.
Why nobody can prove cause and effect
None of these studies can say “we changed only speed, and rankings moved this much”. There are four technical reasons:
- Confounding factors. A team that invests in speed usually invests in content, links and technical SEO too. As Patrick Stox of Ahrefs put it, these correlations are not that different from saying that sites working on their SEO rank better.
- Missing data. The signal is based on field data, but most pages do not have any. Studies either look only at popular pages that have data, which skews the sample, or use lab scores, which measure the wrong thing.
- Overlapping updates. The 2021 page experience rollout ran alongside core updates. Changes in visibility cannot be tied to one cause.
- A non-linear effect. Most studies split pages into “pass” and “fail”. As the next section shows, the effect sits at a threshold, and that split makes it harder to see.
There is no public controlled experiment that keeps content and links fixed, changes only Core Web Vitals and measures rankings. That gap is part of the evidence. If the effect were large, someone would have shown it by now.
From the field: I made similar Core Web Vitals improvements on two news platforms. One showed a clear result within three months (more on that below). On the other, neither engagement nor rankings changed. That audience simply had no expectation of speed; the content and the way the site was built did not need it. Same fix, same type of site, two different outcomes. Perhaps the “small effect” that correlation studies see on average is the sum of this: meaningful on some sites, zero on most.
The threshold effect: a muddy windscreen versus a slightly dirty one
This is where Google’s statements and field experience meet most clearly. Mueller said you will start to see positive effects once you get out of the “poor” area. Walton said making an already good page faster will not increase rankings. The effect is not linear. It sits at a threshold.
I think of it like this. On a road with a 20 km/h speed limit, you cannot drive if your windscreen is covered in mud. If it is only slightly dirty, you can still drive comfortably. Polishing the glass will not make you go faster, because the road sets the speed, not the glass. A “poor” page is the muddy windscreen. Going from “good” to a score of 100 is polishing the glass.
From the field: When I fixed sites that were in a really bad state, I almost always saw traffic and rankings go up. But every time, the gain came through user behaviour. On the news platform with a loyal readership, after three months session duration and returning users went up, the share of indexed pages rose, and rankings followed. After a certain point, the site settled at its natural ceiling. Speed did not raise the ceiling. It removed what was holding the site back. The ceiling is set by the content and how the site is built.
The practical rule: first find the page groups that are “poor” and bring them to “good”. A CLS or LCP problem at template level affects hundreds of pages at once. Once you are in “good”, chasing milliseconds is usually poor value for SEO.
Where speed actually pays: revenue and crawling
A small ranking effect does not make speed worthless. Its return shows up in two other places, and there it is much easier to measure.
Conversions. The cleanest public test that isolates speed is Vodafone’s 2021 A/B test. Two versions of the same landing page were served to paid traffic split in half. The version with a 31% better LCP brought 8% more sales. Because the traffic was paid, rankings had nothing to do with the result. A Google-commissioned Deloitte study of 37 brands found that a 0.1-second improvement on mobile lifted conversions by 8.4% for retail sites and 10.1% for travel sites.
From the field: On an e-commerce site, rankings did not move after the Core Web Vitals fixes. Sales went up by about 20%. When you explain this to management, talking about revenue is far more convincing than talking about rankings.
Crawling. This part is about server response time rather than Core Web Vitals, but the two are usually handled as one piece of work. Google’s crawl budget documentation is clear: if a site slows down or returns server errors, “the limit goes down and Google crawls less”. If response times stay stable or improve, the limit goes up. On large sites, this directly affects how quickly new and updated pages get indexed. The rise in indexed pages I saw on the news platform fits this side of the story.
Myths that keep circulating
- “The PageSpeed score is a ranking factor.” It is not. The signal is field data; the Lighthouse score is a diagnostic tool.
- “Core Web Vitals are only a tie-breaker.” According to Google, a little more than that, but they do not replace relevance.
- “All three metrics must be good.” Google said in 2021 that this is not required.
- “Pass and you rise, fail and you drop.” Google’s documentation says good scores do not guarantee top rankings, and Mueller doubts a big drop from Core Web Vitals alone.
- “The faster, the better.” Beyond “good”, the ranking return is close to zero.
- “The Search Console group score applies to every page.” Groups are for reporting; ranking is generally page-level.
Use the score as an alarm, not a target
Despite all this evidence, I like that Core Web Vitals scare people. In SEO, metrics and scores are motivation as much as measurement. A red Core Web Vitals report can get a template fix done in a week after it has been postponed for years. The problem is not the score itself, but how you look at it. Treat it as a target and you will waste time chasing 100. Treat it as an alarm and you will find the “poor” pages, fix them, and spend the rest of your effort on content and users.
Core Web Vitals are an important part of a pocket watch. The watch will not work without it, but that part alone does not make it run.

Top comments (0)