Sharpening

Because it has nothing to work from. Blur, whether from missed focus or movement, spreads the information from each point across neighbouring pixels and mixes it irreversibly with theirs. Sharpening only looks at the differences that survived, so it amplifies a smeared edge into a bolder smeared edge. The photo gets more contrast and no more truth.

Machine learning deblurring and upscaling go further, but the honest description is that they generate plausible detail consistent with their training rather than restoring the original. That is fine for texture on a jumper and risky for anything a buyer will read as fact, such as a serial number, a care label, a hallmark or the fine print on packaging.

Last, and for the size the image will actually be displayed at. Sharpening at full resolution and then resizing down discards most of what you applied, and the resampling can leave crunchy edges behind. Resize first, sharpen second, export third.

Professionals often split it into two stages: a light capture sharpening pass to compensate for the sensor's own softening, then an output pass tuned for the final destination. A web thumbnail wants more relative sharpening than a large print, because downsampling softens.

If you are preparing a run of listing photos, applying the same output settings across the set matters more than tuning any single image. Photomop's batch editing handles that kind of uniform pass, which is what keeps a gallery looking like one shoot rather than several.

Three things, in rising order of annoyance. First, halos: a pale outline that is invisible on a busy background and obvious against the pure white backdrop most marketplaces expect, where it reads as a badly done cut-out even when the cut-out was fine. Second, amplified faults, because sharpening does not know the difference between real detail and noise, JPEG blocking or coloured fringing from the lens, and it strengthens all of them equally. Third, file size, since crisp high-contrast edges and boosted grain both compress worse than the original.

There is also a credibility cost. Buyers cannot name what is wrong with an over-processed photo, but they register it as retouched, and a listing that looks manipulated invites more questions than one that looks plain.