Every watermark damages the photo it is protecting. There is no version of this where you add a mark and lose nothing. So the useful question is never "should I watermark", it is "how much damage am I buying, where am I putting it, and what do I actually get back".
Most people get this wrong in one of two directions. Either a tiny pale logo in the bottom right corner, which costs almost nothing visually and protects almost nothing in practice, or a heavy opaque bar across the middle, which does protect the image, mostly by making it not worth looking at. The craft lives in between those two, and it is mostly about opacity, placement and knowing what a watermark is realistically for.
What a watermark actually stops
A visible watermark stops casual reuse. That is the whole job, and it is a bigger job than it sounds.
Casual reuse is someone right-clicking your product shot for their own listing, a content farm scraping your gallery, a Facebook page reposting your photo with no credit, a slide deck that needed a picture. These people are not attacking you. They are taking the path of least resistance, and a watermark makes your image a slightly worse option than the next one down the search results. That is enough. Most image theft is opportunistic, and opportunism is defeated by mild inconvenience.
What a watermark does not do is stop anyone who has decided they want your specific image. Content-aware fill, inpainting and generative removal tools have made corner marks trivially removable, and they are now built into consumer photo editors. Someone motivated can clear a small mark in seconds and a large one in a couple of minutes. If your business plan depends on nobody being able to remove your watermark, you do not have a business plan.
The second job, and the one people underrate, is attribution. A mark carries your name with the image wherever it travels. When your photo ends up on a Pinterest board with the link stripped, the watermark is the only surviving thread back to you. That is worth something even when the image is being used without permission, and sometimes it is worth more than the licence fee you were not going to collect anyway.
Opacity is the setting that decides everything
Opacity does more work than size, font or color combined, and it is the setting people leave at the default.
At full opacity, a watermark is not a watermark. It is a caption painted over your photo. It reads as amateur, it destroys whatever is underneath it, and it does not deter any more effectively than a semi-transparent mark, because the removal effort is the same either way.
Somewhere in the middle band, roughly a third to a half opacity for a white mark on a busy photo, you get something that stays readable at thumbnail size while letting the image show through. Below that it starts to disappear against bright areas and stops functioning at small sizes, which matters because thumbnails are where most people first see your work.
The right number depends entirely on what is underneath. A pale watermark on a bright sky vanishes. The same mark on a dark forest reads clearly. If you watermark a batch of photos with one setting, you are guaranteed to have some that are illegible and some that are too heavy. Batch consistency and per-image correctness are in direct conflict, and for most people batch consistency wins because the alternative is opening every file.
One practical way out: put the mark where the tonal range is predictable. If you always place it over the same corner of a studio product shot on a white background, a mid-grey watermark behaves the same way every time.
Why corner watermarks get cropped out
The bottom right corner is where almost everyone puts a watermark, and it is the single easiest place to remove without any editing skill at all. You do not need inpainting. You need the crop tool.
Crop off the bottom fifteen percent of the frame and the mark is gone, along with very little that matters in most compositions, because photographers habitually leave the edges of the frame quieter than the middle. Every social platform hands the user a crop tool for free. Every phone gallery has one.
Worse, platforms crop for you. A 3:2 landscape photo shown as a square Instagram grid thumbnail loses the left and right edges. A 16:9 image shown in a 1.91:1 link preview loses the top and bottom. Your carefully placed corner mark can be outside the visible area in exactly the contexts where you most want to be credited. That interaction between watermark placement and platform cropping is worth thinking about alongside the target dimensions themselves, which is what the image size cheat sheet covers.
If a corner is still where you want the mark, move it inward. Placing it a visible distance in from the edge, overlapping the subject rather than sitting politely beside it, means a crop that removes the watermark also removes part of the photo. That is the entire trick. Attach the mark to something the thief wants to keep.
Center placement, and why professionals use it more than you would think
Stock libraries and preview galleries put the mark straight through the middle at low opacity, often large and diagonal. It looks crude next to a tasteful corner logo, and it is dramatically more effective.
The reason is simple. The middle of the frame is where the subject is. You cannot crop it out, you cannot clone it out without repainting the subject itself, and the mark is visible even in a small thumbnail. The cost is that the image is no longer pleasant to look at, which for a stock preview is the point. The preview is meant to sell the clean version, not substitute for it.
Use this when the watermarked file is a sample and a clean file exists behind a payment or a licence. Do not use it for photos you are publishing as your finished work, because you are publishing something you have deliberately spoiled.
Tiled watermarks: when the ugliness is worth it
Tiling repeats the mark across the whole frame in a grid or diagonal pattern. It is the most robust option available to you short of not publishing the image.
A tiled mark cannot be cropped off, because there is no region of the image without one. Removing it means repainting the entire photo rather than one patch, and while generative tools have got good at small removals, reconstructing a whole frame under a repeating pattern is a different order of work. The mark also survives resizing and re-encoding, since some instance of it is always large enough to read.
The cost is real. A tiled watermark is on top of every part of the picture, including the subject's face, and there is no opacity low enough to make that invisible while still leaving the pattern legible. Tiling also adds file size, because you have introduced high-contrast edges across the entire frame and JPEG compression spends bits on edges.
Tiling earns its place for proofs sent to a client before payment, event photo galleries where people are choosing which shots to buy, and any image posted somewhere you expect scraping. It is the wrong choice for a portfolio piece or a social post, where the image has to work as an image.
Size, color and staying legible at thumbnail scale
Size the watermark relative to the image, not in pixels. A mark that is comfortable on a 4000 pixel export is invisible on the 600 pixel version the platform actually serves. Setting size as a percentage of the image width means one setting works across every export size, which is why the watermark tool takes size as a percent rather than a font size in points.
On color, the usual answer is white or black, whichever contrasts with the area the mark sits on. A brand color is tempting and usually a mistake, because brand colors are chosen to work against a controlled background and your photos are not a controlled background. Mid-grey is the compromise that fails least often, since it stays visible against both bright and dark regions instead of disappearing into one of them.
Keep the text short. A studio name is fine. A name plus a website plus a copyright symbol plus a year is four things competing for the same few pixels, and at thumbnail size it becomes a grey smear that communicates nothing. If people can find you from your name alone, use your name alone.
Test at the size people will see. Export it, look at it small, and if you cannot read the mark at the size the platform displays, either the mark is too small or too transparent, and it is currently doing neither of its two jobs.
The watermark is not your only credit, and it is not the most durable one
Embedded metadata does the attribution job without touching a single pixel. The IPTC fields in a JPEG carry creator name, copyright notice, contact details and usage terms, and Google Images reads those fields to display image credit information in search results. That is credit attached to your file at zero visual cost.
The catch, and it is a large one, is durability. Many platforms strip or rewrite metadata on upload, screenshots discard it entirely, and anyone can clear it with a free tool in one click. Metadata is credit for people acting in good faith. A watermark is credit that survives people who are not. You want both, and you should never treat filling in IPTC fields as a reason to skip the visible mark, or vice versa. If you have never looked at what your files are actually carrying, the guide to EXIF data covers what is in there and how to read it, and the metadata viewer will show you the fields on a file you already have.
Content Credentials, the cryptographically signed provenance standard from the C2PA group, is the more interesting long-term answer, since it binds authorship to the file in a way that is tamper evident rather than merely present. Support is still uneven across cameras, editors and platforms, so treat it as something to watch rather than something to rely on this year.
A workflow that keeps the damage small
The order of operations matters more than any individual setting.
- Finish the photo first. Crop, color and retouch before the mark goes on. Watermarking early means every later crop moves or removes it.
- Keep a clean master. Watermark an export, never your original. This sounds obvious and is the mistake people make once and never again.
- Fill in the metadata on the master so the credit exists in the file regardless of what happens to the pixels.
- Choose placement by destination. Client proof gets tiled. Public portfolio gets a single inset mark over a low-detail part of the subject. Stock preview gets a center mark.
- Export at the size you will publish and place the watermark on that export, so the mark is sized for the actual viewing size rather than scaled down with everything else.
- Look at it small. If it is unreadable at thumbnail size, adjust before you publish, not after.
Anything that uploads your files to a server to burn a mark into them is worth a second thought, particularly for client work under an NDA or images of people who have not consented to a third party holding a copy. Watermarking is a pixel operation, and a browser is perfectly capable of it, which is why the tool here does the whole thing locally and never sends the file anywhere.
When not to watermark at all
Plenty of images should go out clean.
Photos on your own website, where the surrounding page already establishes authorship, gain nothing from a mark and lose a little polish. Product photos for a shop listing should be clean, because the shopper is trying to see the product and a marketplace may reject listing images with promotional overlays anyway. Images intended to spread, such as press photos, memes and anything you are actively hoping gets reposted, should not carry friction you deliberately added.
There is also the case where the photo simply cannot take it. A minimalist composition with large areas of flat tone has nowhere to hide a mark. A tight portrait has nothing but face. For these, publish at a size that is not useful to steal, which for most casual reuse means somewhere around a thousand pixels on the long edge, and put your effort into the metadata instead.
The overall calculation is unglamorous. A watermark buys you deterrence against lazy reuse and a durable credit line, and it costs you some of the image. When the thing you are protecting is worth more than the pixels you are spending, mark it. When it is not, do not, and spend the effort on publishing small and keeping your metadata intact.
