Very few. A trend is worth your money only when it solves a real problem for your visitors or removes work from your team. Most trends do neither. They change how a site looks for about eight months and then date it badly.
We get asked about this constantly, usually because a founder saw a competitor’s new site and wondered whether theirs now looks old. Sometimes the answer is yes. More often the competitor spent money making a slower site that photographs well.
This is our honest view on what is worth following in 2026 and what is not, including the filter we use to decide. You will not agree with all of it, and that is fine. A list of trends with no opinion in it is useless to you.
We use three tests. Does it make the site measurably better for a visitor? Does it survive a browser or platform change? Would we still want it on the site in three years? A trend has to pass at least two before we recommend it to a client.
The three year test is the one that kills most candidates. Anything driven purely by novelty fails it immediately, because novelty has a shelf life by definition. A layout idea that looked fresh in January reads as dated by the following spring, and rebuilding a site is not a cheap way to stay current.
The browser test is quieter but it matters more. Techniques built on a browser feature that vendors are actively converging on are safe bets. Techniques built on a hack that happens to work today are not, and we have inherited enough sites held together by those hacks to be cautious.
Yes, with restraint, and this is the strongest technical trend on the list. CSS scroll-driven animations let you tie an animation to scroll position rather than to a timer, using properties like animation-timeline, scroll-timeline, and view-timeline. That moves work out of JavaScript and into the browser.
The signal that this is durable is Interop 2026. The joint browser vendor effort, announced in February 2026, includes scroll-driven animations among its focus areas alongside view transitions. WebKit’s announcement describes twenty focus areas in total, fifteen of them brand new and five carried over from Interop 2025.
When browser makers publicly agree to converge on a feature, that is about as good a durability signal as the web offers. It is also the reason we now treat the View Transitions API as a real option rather than a demo, since it sits in the same programme.
The restraint part still applies. A scroll animation that reveals content as it enters view is useful. A scroll animation that hijacks the scrollbar and takes the page hostage is not, and the evidence on that is fairly settled. Our piece on whether website animations help or hurt conversions covers where the line sits.
Only where the image is genuinely decorative. AI imagery has become cheap and fast, and that is exactly the problem. When every site in a category uses the same generated aesthetic, the aesthetic stops signalling quality and starts signalling that nobody spent money.
Our position is that generated images are fine for abstract backgrounds and texture, and poor for anything a visitor might read as evidence. Team photos, product shots, and case study imagery should be real, because a visitor who spots a generated face on your about page will discount everything else on the site.
This is a trend where the correct move is partial adoption rather than a yes or no. We wrote a fuller argument in our piece on whether you should use AI-generated images on your website, and the position has not changed.
A baseline, and treating it as a trend is why the web is still failing at it. The WebAIM Million report published in 2026 analysed one million home pages and found an average of 56.1 detected errors per page. That is not an edge case, that is the normal state of the web.
The specific failures are worse reading. Low contrast text appeared on 83.9 percent of home pages. Missing alternative text for images appeared on 53.1 percent. Missing form input labels appeared on 51 percent, and empty links on 46.3 percent.
The line from that report that should embarrass the whole industry is this: 96 percent of all errors detected fall into six categories, and those most common errors have been the same for the last seven years. We are not failing at something hard. We are failing at the same six easy things, repeatedly.
There is a design trend hiding in here that we do think is worth following. Interop 2026 includes a new CSS function called contrast-color, which selects a colour with guaranteed contrast. Anything that makes the most common failure on the web harder to commit is worth paying attention to.
Both are fine, and neither is a strategy. A bento grid is a sensible way to show several unequal things at once, which is a real problem on a homepage. Oversized type creates hierarchy fast. Used well they help, and used as a costume they just make your site look like everyone else’s.
The failure mode with bento grids is filling the boxes with nothing. We regularly see six panel layouts where four panels contain a stock icon and three words. The layout implied there was substance behind it and then there was not, which is worse than a plain section that delivers.
Oversized type has a measurable cost, which is the honest argument against overdoing it. Huge display text is usually a web font, and web fonts are one of the causes Google names for layout shift. If the giant headline is what pushes your Cumulative Layout Shift past 0.1, the trend just cost you a ranking signal.
Not redesign, but restructure. AI answer engines read structure, not aesthetics, so nothing about your visual design changes how you get cited. What changes is whether your content answers questions directly and whether the text is in the HTML at all.
The pressure behind this is real. Pew Research Center found in July 2025 that Google users clicked a search result 8 percent of the time when an AI summary was present, against 15 percent when it was not. Fewer clicks means the pages that do get visited have to work harder.
The design implication is modest and specific. Lead each section with a direct answer, keep headings phrased as questions people actually ask, and do not bury text inside images or client-side rendered components. None of that requires a new visual direction, which is exactly why we push back when this gets sold as a reason to redesign.
Custom cursors, scroll hijacking, autoplaying video heroes, and preloader animations on a site that loads in under a second. Each one adds weight and friction in exchange for a moment of delight the visitor did not ask for.
Preloaders are the one we argue about most, because clients often like them. A loading animation on a fast site is a delay you built on purpose. If the page is ready in 600 milliseconds and the animation runs for 1.5 seconds, you have made the site slower to look like you care about craft.
We are equally unmoved by the periodic return of anti-design and deliberate ugliness. It works for a handful of brands with enormous existing recognition and it works for nobody else. If your visitor does not already know who you are, confusing them is not a strategy.
Ask who benefits. If a technique makes the site faster, clearer, or cheaper to maintain, it will outlast the fashion cycle because it is solving something. If the only beneficiary is the design team’s portfolio, it is a fad wearing a trend’s clothes.
The second test is what happens when it fails. Scroll-driven animation degrades to a static page, which is fine. A layout that only makes sense with a custom cursor breaks badly for anyone on a touch screen. Techniques that fail gracefully are safe to adopt early. Techniques that fail loudly are not.
Across 150+ projects, the sites that have aged best were not the most fashionable ones at launch. They were the ones built on clear hierarchy, fast loading, and honest content. That is a boring conclusion and we have stopped apologising for it.
Fix contrast and labels before you chase anything new, because those are the failures affecting most of the web and yours is probably among them. Then adopt scroll-driven animations and view transitions carefully, since browser vendors have publicly committed to both. Ignore the rest until next year proves it lasted.
If your site genuinely looks dated, the cause is usually spacing, typography, and hierarchy rather than a missing trend. Those are cheaper to fix than a redesign, and they are what a visitor actually reacts to in the first second.
If you want an honest opinion on whether your site needs a refresh or just needs tightening up, we’re happy to look. Reach out at phoenix.studio and we’ll tell you plainly which it is, including when the answer is that you should leave it alone.
Tell us where you want to go. We'll tell you how we'd get you there.