UX and SEO: what moves rankings and what does not

Optimising user experience for SEO means working on the few parts of the experience that Google says it assesses (perceived speed, visual stability, readability on a smartphone, nothing covering the content) and no longer paying for all the rest.

Page one in the UK tells almost the same story throughout: UX and SEO work hand in hand, fix the site and you will climb. On 26 September 2026 I ran those nine pages through Lighthouse, one by one. The two with a perfect performance score of 100 sit seventh and eighth. The page at the top scores 83, the lowest of the group.

What Google says it assesses

Google's documentation page on page experience, updated on 10 December 2025, opens with a sentence that none of the results on page one quotes: "There is no single signal. Our core ranking systems look at a variety of signals that align with overall page experience". There is no single signal and there is no ranking system called page experience. Anyone selling UX work on the promise of moving that system is selling something that has no switch.

The list of what Google suggests checking is short: good Core Web Vitals, HTTPS, content that reads on a smartphone without zooming or horizontal scrolling, few intrusive ads, no disruptive interstitials, a clear distinction between main content and everything around it. Six items, all of them verifiable in an afternoon.

The same document adds the warning that usually stays out of sales presentations: "getting good results in reports like Search Console's Core Web Vitals report or third-party tools doesn't guarantee that your pages will rank at the top of Google Search results".

The thresholds, up to date

There are three metrics and web.dev publishes the thresholds: LCP within 2.5 seconds, INP within 200 milliseconds, CLS below 0.1. INP replaced FID on 12 March 2024. Guides written before then still list FID, and some of them still rank. Among the UK results, the two older pages I checked (UX Design Institute, last modified in November 2023, and Iconic Digital, December 2024) name neither FID nor INP: they talk about speed in general terms. For the detail metric by metric, see our guide to Core Web Vitals; the definition of the starting concept is in our page on user experience.

The test: the nine ranking pages, measured

On 26 September 2026 I took the first nine organic results on Google UK for the query "ux seo" and ran Lighthouse 13.4.0 on each one, desktop configuration, simulated throttling, from a European data centre. These are lab data: for ranking, Google uses field data from the Chrome UX Report, collected from real people's browsers. They are useful for comparing the pages with one another under the same conditions, not for estimating the score Google sees.

Pos.	Domain	Performance	LCP	CLS	Accessibility	Page weight
1	uxdesigninstitute.com	83	1.85 s	0.012	95	2.3 MB
2	semrush.com	96	0.96 s	0.001	75	2.3 MB
3	iconicdigital.co.uk	91	1.73 s	0.004	91	6.1 MB
4	uxpin.com	95	1.24 s	0.001	90	1.1 MB
5	reddit.com	Not measurable: the document request failed
6	uxtigers.com	90	1.93 s	0.000	94	2.0 MB
7	lyssna.com	100	0.65 s	0.011	95	1.7 MB
8	searchengineland.com	100	0.76 s	0.001	93	1.0 MB
9	pointblankdigital.co.uk	Not measurable: the server responded with a captcha challenge

Seven pages out of nine returned a result. Of these, the two with the best performance, both at 100, are in seventh and eighth place. The top result scores 83 and has an LCP of 1.85 seconds, almost three times Lyssna's 0.65 seconds. Page weight ranges from 1.0 MB to 6.1 MB, and the heaviest page sits third: the order of results does not notice.

CLS, the metric most guides warn about, does not separate these pages at all: all seven stay below 0.1, the highest being 0.012. When I ran the same test on Google Italy on 20 September 2026, the page in fifth place, which teaches how to look after UX for SEO, had a CLS of 0.353, more than three times the threshold, and still ranked ahead of four pages that met it. Passing the threshold is not what decides the order.

None of the seven reaches 100 for accessibility: the median is 93, and Semrush, which has one of the most cited articles on the subject, stops at 75, the same score it got in the Italian test. To be fair, I also measured our own page on user experience (the Italian version), which is not on page one either: 97 for performance, 97 for accessibility, CLS 0.004. A score is not a ranking, and that holds when the score is ours too.

Three numbers that have circulated for years and do not hold up

The 53% who leave after three seconds. It is one of the most quoted statistics on this subject. The primary source is a post on Google's blog from 31 January 2017, which attributes the figure to DoubleClick and does not publish the methodology: it does not say how many pages, of what type, in which market, on what connection. It is a measurement of mobile sites from almost ten years ago, and it is still presented as if it described how people browse in 2026.

Bounce rate. It is an analytics metric and does not appear in any Google Search document about ranking. It measures what happens on your site, not what Google observes in its results page. It remains useful for telling whether a page disappoints the people who land on it, which is why we wrote a page on bounce rate, but it is not a ranking lever.

Dwell time. Same story, made worse by the fact that almost nobody actually measures it: what you see in GA4 is engagement time for your session, which knows nothing about whether and when the person went back to the search results. If you want a behavioural signal that Google really shows you, it is the click-through rate in Search Console.

The work that moves something, in order of return

1. Match the page to the intent

This is the user experience work with the highest return, and it does not go through a design file. A page that answers the wrong question cannot be saved by any score. In the test above, the top result does not win on a Lighthouse number: it opens by defining both acronyms, SEO and UX, before saying what to do. The People Also Ask box on the same results page asks whether UI/UX is a well-paid job and whether SEO is dead because of AI: for this query, many of the people searching are still working out what the two terms mean together. The method for checking this is in our guide to search intent.

2. Mobile is not a layout, it is the content

Google indexes with its mobile crawler and its documentation is explicit: "If it's your intention that the mobile page should have less content than the desktop page, you can expect some traffic loss". Text cut from the mobile version to tidy it up is not an aesthetic compromise: it is content that stops existing. The same document makes clear that putting that text inside an accordion or tabs is fine: what matters is that it is in the code.

3. LCP: look at field data, not the lab

The number Google uses is the 75th percentile over 28 days of real visits, the one in the Chrome UX Report, not the score a tool gives you on a single run. If the 75th percentile is above 2.5 seconds, the fix with the best return is almost always the main image: format, declared dimensions, loading priority. The rest of the method is in our guides to page speed and image optimisation.

4. CLS: the banner pays the bill

In the Italian test, the two pages with the worst CLS, 0.353 and 0.118, had consent management and push notification scripts among their third parties. In the UK sample, the two pages with the highest CLS, 0.012 and 0.011, both load a consent management platform (Cookiebot and CookieYes). The values are small, so this is not a general rule: it is what I saw in these samples, and it matches what we find in audits. Layout shift almost always comes from the cookie banner, the announcement bar and the font that loads late, not from the design.

5. Nothing covering the first screen

Intrusive interstitials are one of the six items on Google's list, and in the UK they are the one there is least argument about, because PECR requires consent for non-essential cookies, so some level of coverage is unavoidable. The difference lies in what happens next: a banner that stays, that comes back on every page or that covers half the screen on a 6-inch phone is a problem that shows up in the data before it shows up in opinions.

6. Accessibility, which comes with legal duties

In the UK, the Equality Act 2010 requires service providers to make reasonable adjustments for disabled people, online as well as offline, and the duty is anticipatory. Public sector bodies have had their own regulations since 2018, with WCAG as the reference. And a UK business selling covered services to consumers in the EU, e-commerce included, has had to comply with the European Accessibility Act since 28 June 2025. Many of the checks involved (contrast, field labels, heading hierarchy, text alternatives) are the same ones that raise the web accessibility score and make the page readable to a crawler. If you sell online, our page on e-commerce SEO starts here.

The constraint is not the design, it is who is on the other side

The Lloyds Consumer Digital Index 2025 puts the share of UK adults who are online at 95%. Being online is not the same as being at ease, though. The Essential Digital Skills 2025 survey, carried out by Ipsos for Lloyds on 4,175 adults and published on 28 October 2025, finds that 85% of UK adults have foundation-level digital skills: roughly one in seven does not. People with lower digital ability are, in the report's words, overrepresented among older adults.

These are the numbers that change the order of priorities compared with a checklist written for designers. For one adult in seven, body text at 14 pixels, a hamburger menu with three levels and a form with nine required fields are not aesthetic choices, they are filters. And filters show up in the completion rate long before they show up in rankings.

The same page is also read by an agent

Lighthouse 13 has added a category called agentic browsing, which assesses how readable a page is for an automated agent. Google describes it as under development and subject to change, so it is not a score to chase, but the values in my sample are instructive: 1.00 for the top result (UX Design Institute), UXPin, Lyssna and Search Engine Land, 0.67 for Semrush, Iconic Digital and UX Tigers.

It is the same game being played in AI Overviews, which open the Google UK results page for this very query, and in zero-click searches: a page that exposes its content in readable HTML, with real headings and data in tables rather than inside an image, is understood by both Google and a language model. A page that builds everything in JavaScript after loading is not. This is where user experience and technical SEO stop being two separate departments.

How we measure whether it worked

On a real project the check is not the score, it is the queries that start bringing traffic. At Macropix, a Milan-based manufacturer of custom LED screens, none of the nine keywords that now bring in the most qualified enquiries contains the company name: the biggest is "ledwall", with 4,400 searches a month in Italy, and the others describe products and applications. The path, the constraints and the numbers are in the case study on SEO for manufacturing, and you can find other projects among our published work.

When instead you need to know which fixes are worth the budget on a specific site, that is the work we do in our SEO projects: we start from the Chrome UX Report and the queries in Search Console, not from a redesign.

The only number with no exceptions

Across the seven pages I was able to measure, performance, LCP and weight go in every direction and do not explain the order of results. One number behaves differently: accessibility does not reach 100 anywhere, not even on the two pages that score 100 for performance with an LCP under 0.8 seconds. It is also the only one of the six checks listed by Google with a legal duty behind it: the Equality Act in the UK and, for anyone selling covered services into the EU, the European Accessibility Act since 28 June 2025. The first user experience fix with a guaranteed return is not the one that raises a score: it is the one that has to be done anyway, and that happens to make the page readable to a crawler as well.

Frequently asked questions