Why AI art specifically runs into trouble
Midjourney, DALL-E and similar tools default to square or near-square canvases, at whatever resolution that tool ships with. Almost no print product is square. A t-shirt front, a mug wrap, a poster and a phone case each need a different aspect ratio and a different exact pixel size, so the very first mismatch happens before you've done anything wrong — the source image just wasn't built for the product you're putting it on.
On top of that, AI-generated images often carry two things that specifically cause POD rejections: soft, slightly inconsistent edges (which cause a visible white or gray halo when a background is removed), and a fixed resolution that was fine for a screen preview but falls short once stretched across a 12×16 inch print area.
What each product actually needs
Exact requirements vary by platform and by product, but as a general guide:
- Apparel (DTG print): a transparent PNG, sized to the platform's exact print-area pixel dimensions at 300 DPI — commonly in the range of 4500×5400px for a full front print, though this varies by product and platform.
- Mugs, posters, phone cases: a full-bleed JPG or PNG (no transparency needed), sized to the product's exact print template, since the image needs to reach every edge with no white border.
- All products: no compression artifacts, no leftover watermark from the AI tool, and edges clean enough that a background removal tool (or a person) doesn't leave a visible fringe.
Always check the exact current spec sheet on the platform you're using (Printify and Printful publish per-product templates), since these numbers change as new products are added and some vary by supplier.
The commercial-use question, separately
This is a different issue from file specs, and worth checking before you invest time preparing a file: the AI tool that generated the art (Midjourney, DALL-E, etc.) has its own commercial-use terms, and the platform you sell through (Etsy, Printify, Printful) has its own AI-content disclosure policy. Both can change, so check both directly rather than assuming last year's rule still applies. We can prepare a file to be print-ready; we can't tell you whether a specific piece of AI art is clear to sell, since that depends on terms outside our control.
How to prepare the file yourself, roughly
- Get the platform's exact template dimensions for the specific product before you do anything else — not a generic "print size," the platform's own pixel template.
- If the AI image's aspect ratio doesn't match, you'll need to either regenerate at a closer ratio, extend the canvas (AI outpainting or a manual extension), or accept a crop — our free Aspect Ratio Cropper will crop and export to the exact target pixel size without distorting the art, once you know which product you're sizing for.
- Remove the background cleanly if the product needs transparency, and zoom in at 100% to check the edge for a leftover halo or fringe.
- Upscale properly if the resolution falls short of the template's requirement, and check for new artifacts introduced by the upscaler at 100% zoom, not just at thumbnail size.
- Export in the exact format and color mode the platform's spec sheet asks for.
If a step above turns into more troubleshooting than you'd like, this is exactly the kind of file our AI Art → Print-on-Demand Ready service handles.
Fixing a Printify or Printful "publishing error"
If the platform itself is throwing an upload or publishing error rather than just warning about quality, it's almost always one of a handful of causes:
- File resolution below the product's minimum — the platform checks pixel dimensions against its own template the moment you upload, before you even see a print preview.
- Wrong file type — most POD platforms expect PNG or JPG for print files; a PDF, WEBP or an AI/EPS vector file dropped into an image uploader will usually fail outright.
- Color mode mismatch — a CMYK or indexed-color PNG can fail to upload on platforms that expect RGB for the upload step (print-side conversion happens later, on their end).
- File size over the platform's limit — a very high-resolution PNG with transparency can be larger than expected; re-exporting with standard compression usually fixes this without losing quality.
These are platform upload-pipeline issues, separate from whether the design itself will print well — a file can pass the upload step and still look soft or get a halo around a removed background, which is what the checks earlier in this guide catch.