Transparency checkerboard

The grid is not in the file. Where an image has no pixels, or pixels whose alpha channel is zero, there is nothing at all to draw, and a tool still has to put something in that space so the reader can see the shape of the hole. Every editor answers the same way: two squares of near neutral grey and white, alternating on a small lattice, painted behind the artwork and never part of it. Open the same file in Photoshop, Figma, Sketch, GIMP or a browser’s element inspector and the backing looks the same, which is the unusual thing about this convention. It was never specified anywhere. It simply became the pattern that means empty.

White would have been the obvious choice and it fails for one reason: a white region and an empty region look identical, so the canvas stops being able to answer the question the designer is actually asking. Is that white paint, or is that nothing? A checker cannot be mistaken for content, because almost no real artwork contains a low contrast two tone lattice at that scale, and the eye files it as background rather than as image. The cells are kept small and the two greys kept close together so the pattern recedes instead of competing, and the payoff is that partial alpha becomes legible too: a region at thirty percent opacity shows the grid faintly through it, so you can read roughly how transparent it is before the file is ever placed on a page. Placing it is when the question is settled for real, since the transparent region takes the colour of whatever sits behind it through alpha compositing, and the checker is a stand in for a background nobody has chosen yet.

Because the convention is so well learned, it has moved out of editors and into product interfaces. A colour swatch showing a fully transparent value is drawn as a checker rather than as an empty box. An image upload slot uses it to say the asset it is waiting for may have transparency. A colour picker’s alpha slider runs its gradient over a checker so the low end reads as vanishing rather than as turning white. In CSS the pattern is four lines of conic-gradient with a background-size, so it costs nothing to draw, which is part of why it spread. What it should never do is composite into the output: the grid belongs to the viewer, not to the artwork, and an exported PNG that has the checker baked into it is the classic beginner bug this whole convention exists to prevent.

Not every checkerboard is this term, and the difference is role rather than appearance. A chequerboard floor receding to a horizon in vaporwave, or a warped checker plane in acid graphics, is a decorative motif with a history in op art and racing flags; alternating light and dark full width rows down a marketing page are a layout rhythm device. Those patterns are content. This one is scaffolding, drawn by the software in the places where the file has no colour to offer. The practical consequences follow from that: keep the contrast low enough that the grid never competes with the artwork sitting on it, keep the cells small, and do not reach for a checker as ornament inside a product that also uses it to mean alpha, because a reader who has learned the convention will read the ornament as a hole.