YouTube's New Desktop Layout: Giant Thumbnails, a Comments Sidebar, and What It Changes About the Image You Upload

Somewhere in the last few months a lot of people opened YouTube on a laptop and found the comments where the recommendations used to be, the video player smaller than it was, and thumbnails so large that three of them filled the screen. The reaction was predictable and loud. Almost all of the coverage that followed answered the viewer's question: how do I get the old one back.

The creator's question is different and nobody has answered it. Your thumbnail is a fixed file. The surfaces that display it have been quietly changing shape for a year, and the direction of travel in every single test is the same — more pixels for the picture, fewer for the words. That is a packaging change whether or not any individual experiment survives contact with the complaints.

This is what is actually being tested, with dates and sources, what is confirmed versus what is a screenshot on Reddit, and the part that matters: which of your habits stop working when a thumbnail designed for a feed cell gets shown at four times that size.

What YouTube is actually testing on desktop

Four distinct experiments have been reported, and conflating them is why the advice circulating is so confused. They did not start together and they do not do the same thing.

Experiment	What changes	Reported	Status
Enlarged homepage grid	Home feed cards grow until as few as two or three fit a row, depending on screen size and browser zoom	Android Authority, April 2025; still being written about in guides dated August 2026	Live for some accounts, no opt-out documented
Two-column sidebar	Watch-page recommendations in two columns; titles truncated after a few words	PiunikaWeb, 14 January 2026	Test; YouTube has not commented
Giant sidebar thumbnails	Recommendation thumbnails in the right rail roughly double; title space shrinks	PiunikaWeb, 4 February 2026, updated 13 April 2026 to report wider exposure	Test, reportedly widened
Comments sidebar watch page	Title, description and comments move into a right-hand panel; large recommendations sit directly under a smaller player	Ubergizmo, Android Authority, Beebom and others, around 7 October 2026	Test; a near-identical layout was offered to Premium subscribers in 2024 and withdrawn

Two details in that table deserve more weight than the rest. The first is that the October 2026 watch page is a revival: outlets covering it noted that the same arrangement was tried in 2024, briefly as an opt-in experiment for Premium subscribers, and pulled after the reaction. A company does not resurrect a layout it has abandoned. It resurrects one it believes in and lost an argument about.

The second is the April 2026 update to the sidebar test. PiunikaWeb's original February report was a handful of screenshots; the update said the experiment appeared to have widened, on the strength of further user reports. Widening is the signal worth watching, because that is what YouTube does between an idea and a launch.

Nothing here is announced, and that is the normal state

There is no blog post, no Creator Insider segment and no Help Center page for any of these layouts. That is not evasion; interface experiments are generally not announced to the people in them. Practically, three things follow.

You cannot verify your own exposure by asking someone else. Two creators comparing notes over a call can be looking at two different products. Nor can you opt out: reporting on the enlarged grid found no native control for cards per row, and the guides that promise otherwise are describing browser extensions and filter lists, which are fine for your own viewing and irrelevant to your audience's. The youtube.com/new page is not a master switch — it covers the specific experiments a Premium account has been offered, and a layout you were simply dropped into is usually not among them.

And you cannot tell what fraction of your audience is affected, because YouTube Analytics does not report interface variants. If you want to know what your thumbnails look like in the new layout, the only honest answer is to find a browser profile that has it and look.

The real change: your thumbnail's display size lost its ceiling

For a decade the hard constraint on thumbnail design was smallness. The standard advice — ours included — is built on it: design for the 168-pixel feed cell, three or four words maximum, one subject, contrast that survives the shrink.

That advice is still right. What has changed is that it is no longer the only case you have to clear. Do the arithmetic on a 1,920-pixel-wide window. With four cards to a row, once you take out the guide rail and the gutters, each card lands somewhere in the region of 420 pixels of width. Drop to two cards a row and each one roughly doubles. On a 2,560-pixel display in the enlarged grid, a card can be wider than the 1,280-pixel file you uploaded.

That is the whole story in one sentence: in the small case your image is downscaled by a factor of seven, and in the large case it may be displayed at or beyond its native resolution. Downscaling hides sins. Native display does not hide anything.

The file YouTube serves is not always the file you uploaded

YouTube stores every thumbnail at several sizes and picks one per surface. Community documentation of the thumbnail URL scheme lists the familiar set: a medium variant at 320 × 180, a high variant at 480 × 360, a standard variant at 640 × 480 and a maximum variant at 1280 × 720. The middle two are 4:3 and letterboxed, which is why a scraped "HD thumbnail" sometimes arrives with black bars. The 1280 × 720 variant does not exist for every video.

YouTube does not publish which variant goes to which surface, and the sizes are an implementation detail that can change. The useful inference is narrower and safe: there is a top variant, it is the one a large card needs, and it is generated from what you uploaded. A 1280 × 720 upload is the floor for a layout that may show it at 900 pixels wide, not a comfortable margin.

Three things that only break at large sizes

Each of these is invisible in a feed cell and obvious in a card twice the size. If you have been checking your work by looking at the Studio thumbnail picker, you have never seen any of them.

Upscaled source material. A 720-pixel-wide game capture, a screenshot off a phone, a stock image you enlarged, a frame grab from a 1080p export that was itself a 720p upload — all of them look acceptable at 420 pixels and soft at 900. Upscaling is additive: YouTube's own re-encode sits on top of yours. If the subject came out of a video file, pull it at full quality from the source file rather than screenshotting the player.

Compression artefacts. The 2 MB limit pushes people to save JPEGs at quality settings that are fine when the result is shown small. Blocking around high-contrast edges — which is to say, around text and cut-out subjects — scales up with the card. Save at a higher quality and, if the file is then over the limit, bring the file size down by resizing or compressing deliberately rather than by dragging the quality slider until it fits.

Cheap composites. A cut-out with a halo, a drop shadow that falls the wrong way, a face lit from the left against a background lit from the right, an AI-generated hand holding something it cannot hold. At small sizes the eye reads the silhouette and ignores the seams. At large sizes the seams are the first thing it finds, and they read as low effort, which is the one impression no packaging can afford.

The 168-pixel problem has not gone away

It would be a mistake to read any of this as permission to design big. The enlarged layouts are desktop tests, and the majority of viewing is not on desktop at all. Mobile has its own history of thumbnail experiments, including the variable display sizes reported in the mobile home feed, and a 2025 test that placed the title and channel name on top of the image in a semi-transparent bar. On a television the picture is enormous and the viewer is three metres away, which is a third set of constraints again.

So the brief has not been replaced, it has been widened. One file, one 16:9 frame, and a range of display widths that now runs from under 200 pixels to over 900. The design that survives that range is not the one tuned to either end. It is the one with a strong silhouette, a high-contrast focal point, and type that is large enough to read small without being so large that it looks shouted at size.

Titles are the casualty in every version of the test

This is the part with a direct, immediate consequence. Every enlarged layout reported takes its pixels from the same place. PiunikaWeb's January report described sidebar titles cut off after a few letters; the February report quoted a user who found room for only the first two or three words. Coverage of the October watch page noted reduced title visibility among the complaints.

A title truncated to three words is not a title. It is a label. Whatever work your title was doing — the specificity, the number, the stake, the payoff — happens in the words that are no longer there.

Two habits follow, and both are good practice regardless of which layout wins. Front-load. The distinguishing words belong in the first three or four, which is also how titles behave in mobile search and in a notification. And stop splitting the promise across the title and the image in a way that needs both halves to make sense. A thumbnail whose text completes a sentence begun by the title fails whenever the title is clipped, and it was always failing on surfaces where the title sits further away than you think. Our guide to writing titles goes through the front-loading mechanics in detail.

Fewer cards on screen changes what "standing out" means

A thumbnail is never judged alone. It is judged against whatever surrounds it, which is why the most reliable colour advice is relative rather than absolute — be the one warm frame in a column of cold ones, as the colour guide argues.

That tactic depends on neighbours. In a twelve-card view, differentiating against eleven competitors is a real lever. In a four-card view it is close to a coin flip: you cannot plan to contrast with a set you cannot see, and with three neighbours the chance that your deliberate orange sits next to three other oranges is no longer small.

What gains weight instead is intrinsic clarity. A frame a viewer can parse in a single fixation without comparing it to anything — one subject, one emotion, one legible idea — does not depend on the grid. The practical instruction is unglamorous: keep the contrast work, because it still helps on the dense surfaces, and stop treating cleverness-relative-to-neighbours as a strategy that will be there when you need it.

What the comments panel moves besides comments

The October watch page rearranges more than the discussion. In the reported version the title and description move into the right-hand panel, the player is smaller, and large recommendation thumbnails sit immediately under the video. Three knock-on effects are worth planning for.

The recommendation strip is now in the most valuable rectangle on the page — directly below what the viewer is watching. Reporting noted that viewers found accidental clicks easier with big cards there. Easier clicks on recommendations under your video is a mixed blessing: it is traffic to whoever is in that strip, and sometimes that is you.

Description links move. If your sponsor link, affiliate disclosure or next-video link depends on being the first thing under the player, it is now in a side panel the viewer may not have scrolled. That makes the in-video routes more important rather than less — chapters, the pinned comment and end screens, which occupy the player itself and do not care how the page around it is arranged.

And comments are permanently in view rather than below the fold. Audiences of spoiler-sensitive content noticed this first. If you have ever relied on the comments being a scroll away during the first hours of a release, that assumption is gone for part of your audience.

Your A/B tests are still valid; your CTR trend is not

A layout experiment running across part of your audience sounds like it should contaminate thumbnail testing. For Test & Compare it mostly does not, and the reason is worth understanding. The test allocates viewers to your variants as they arrive, so whatever share of your audience sits in an enlarged layout should land across all of your options rather than in one of them. Each variant faces roughly the same mix of interfaces, and the comparison stays internally valid — which is exactly why a controlled test beats a before-and-after.

What is not safe is reading your absolute click-through rate across weeks. If part of your audience moved into a layout with fewer, bigger cards, their impressions and their clicking behaviour both changed for reasons that have nothing to do with your work. A 0.4-point shift in a monthly CTR graph was always mostly noise; during a period of interface experiments it is noise with an extra source. A click-through rate moves on its own for half a dozen reasons before your packaging is implicated at all.

Five changes that are safe either way

The temptation with an unannounced test is to do nothing or to rebuild everything. Neither is right. These five are worth doing because they improve your thumbnails in the current layout and in every reported variant of the new one.

1. Stop uploading anything below 1280 × 720. It has always been the documented minimum and it is now also the ceiling a large card can display. Anything upscaled to reach it is borrowing against the large case.
2. Export at a quality setting you have actually checked at full size. Open the file at 100% before you upload. If text edges are blocky there, they will be blocky in a 900-pixel card.
3. Front-load the title and make the thumbnail self-sufficient. Assume the viewer sees three words and the picture. If the picture needs the other nine, redesign one of them.
4. Hold the text down. Three or four words, not six. The temptation in a bigger card is to use the space; the result is a frame that reads as an advertisement at size and as a wall of letters when it shrinks.
5. Check both ends of the range before publishing. Not the Studio picker, which shows you one size — the two that actually exist.

A two-size check that takes ten minutes

Open your candidate thumbnail in a browser tab and zoom the page to 25%. That approximates the dense feed cell: does the subject still read, is the text legible, can you tell what the video is about? Then view it at 100% on the largest screen you own and look specifically for the three large-size failures — soft upscaling, artefacts around high-contrast edges, composite seams. Most thumbnails that fail one of these pass the other, which is the point of doing both.

If you want it side by side in real feed furniture rather than by eye, the thumbnail preview renders a candidate at the sizes the interface actually uses, including the small ones that are easy to forget. Ten minutes once per upload is cheaper than finding out from a comment.

Frequently asked questions

Is YouTube's new desktop layout permanent?

Unknown, and nobody outside YouTube can say. The October 2026 watch page has not been confirmed as a rollout or described as a test by YouTube itself, and a very similar layout was withdrawn in 2024 after criticism. The enlarged homepage grid has persisted through more than a year of complaints without becoming universal, which is what a long-running experiment looks like.

Can I turn off the new YouTube layout?

There is no documented setting. Reporting on the enlarged grid found no native control over cards per row. Signing out, clearing site data or using a different browser profile sometimes lands you in a different group, and browser extensions can restyle the page for you, but neither changes what your audience sees — which is the only version that matters for packaging decisions.

Should I upload bigger thumbnails now, like 4K?

No. YouTube's documented maximum file size is 2 MB and the upload target is 1280 × 720; a much larger file either gets rejected or gets compressed harder to fit, which is the opposite of what you want. Upload a clean 1280 × 720 built from genuinely sharp source material rather than a larger file built from soft material.

Does the new layout change my click-through rate?

It changes the conditions under which CTR is measured — fewer cards per screen, more pixels per card, less title — for whichever part of your audience is in the experiment. That makes month-over-month CTR comparisons less reliable than usual. It does not invalidate a controlled thumbnail test, where both variants are shown across the same mix of interfaces.

Do I need to redesign my old thumbnails?

Not as a project. The videos worth revisiting are the ones still earning impressions, and the trigger for revisiting them is weak packaging rather than a layout test. If you are refreshing an old upload anyway, the large-size checks above are the ones to add.

Why do my thumbnails look different from my friend's on the same video?

Because you may be in different experiments, and because YouTube shows the same thumbnail at different sizes on different surfaces. On mobile the picture may also be cropped rather than resized, which is a separate mechanism from resizing.

Sources

Checked in November 2026. Every layout described here is an unannounced experiment reported by third-party outlets from user screenshots; YouTube has not documented any of them, and exposure varies by account. Treat dates as the date of reporting rather than the date of the change.

• Android Authority — YouTube is testing giant homepage thumbnails
• Android Authority — YouTube on desktop now has a comments sidebar
• Ubergizmo — YouTube tests desktop redesign, comments moved to right panel
• PiunikaWeb — YouTube sidebar thumbnails get bigger in UI test (updated April 2026)
• PiunikaWeb — two-column sidebar test, titles cut off
• Beebom — YouTube tests reading comments while watching on desktop
• TechIssuesToday — mobile test placing title and channel name on the thumbnail
• Community documentation of YouTube thumbnail URL variants and sizes
• YouTube Help — custom thumbnails (1280 × 720, 2 MB, 16:9)
• YouTube Help — A/B test titles and thumbnails

The bottom line

The layout tests are not a crisis and they are not nothing. They are evidence of a direction YouTube has now tried four separate ways: give the picture more room and the words less. Whether the October watch page ships in that exact form matters far less than the fact that it is the fourth attempt at the same idea.

So the defensible response is not to rebuild a channel's packaging around a screenshot. It is to stop designing for one display size. A thumbnail built from sharp source material, exported honestly, with a silhouette that reads at 25% and seams that survive at 100%, and a title that says what it needs to say in its first four words — that image is correct in the layout you have, in the layout you may be moved into, and on the television where it is three feet wide.

The awkward part is that checking both ends of the range means producing more than one candidate, and producing more than one candidate is where most packaging workflows quietly stop. Thumblore generates variants from a prompt and a reference image, which makes "make a version that holds up big and a version that holds up small, then keep the one that does both" a ten-minute job rather than an evening. If you want the part the specs do not cover — where the subject goes and why — the composition guide is the next stop.