6 Ways to Protect Your Custom Website Images in 2026

In 2026, publishing images you’ve created on your website means accepting the reality that those custom images are machine-readable by image-hungry bots searching for copyright infringement. 

“But,” you say, “my custom images are safe. I created them. I own them. There’s no risk to me.” We wish that were true, but as some well-publicized cases (covered below) have shown, you can still run into problems. This article highlights several strategies to help protect your images, along with one way to track whether bots have scraped your site. 

AI systems, bots and automated crawlers constantly scan public web pages. For photographers, designers, artists, site owners and content teams, this creates a distinct risk: 

The potential for unauthorized AI training on your original work, which (almost unbelievably) could result in you receiving a legal demand letter. 

This may sound ridiculous or even impossible when you consider that you (or someone you hired) produced your original images. You’d think so, but consider the following real-life scenario. 

Later, at some future time, when your site is crawled by an automated enforcement bot (which today seems to be more a question of when, not if), the bot discovers that you have images on your site that match images in a stock library. 

Potential Result: One fine day, you receive a letter of demand for copyright infringement on your own image. 

You can read more about the parallel situation of the Highsmith v. Getty Images case in The Impact of Automated Copyright Enforcement on Small Business. 

Protecting your custom image portfolio today requires more than merely blocking scrapers. It requires a layered protection strategy and continuous record-keeping so that you can avoid being targeted in a copyright claim (or at least give it your best shot). 

Here are some suggestions for a tactical approach: 

When publishing original, self-created images, basic corner watermarks are no longer enough. Modern editing tools can remove simple overlays quickly. There are still some effective ways to do watermarking. Let’s look at two. 

Visible watermarks are one option. They are most effective in high-risk contexts, such as stock previews, proof galleries or downloadable media, rather than across your primary site imagery. 

Yes, visible watermarks are seen by visitors. And if overused, they can negatively affect user experience. For that reason, many website owners prefer to apply them selectively: 

• Semi-transparent overlays intersecting high-detail areas. 
• Integrated text layered into textured regions. 
• Repeating but subtle patterns across an image. 

For brand-sensitive pages (such as homepages or sales pages), creators often rely instead on lower-resolution display files combined with invisible watermarking and monitoring tools. 

Visible marks won’t prevent scraping, but for the bots, they reduce the usability of scraped images and signal clear ownership. They’re also the cheapest form of watermarking and are often free. 

Steganographic watermarking is a method of embedding hidden information directly inside a digital image in a way that is invisible to the human eye but detectable by specialized software. The word steganography comes from Greek roots meaning “covered writing.” It refers to hiding information inside something else so that its existence isn’t obvious. 

Unlike visible watermarks (logos, text overlays), steganographic watermarks do not alter how the image appears. The image looks completely normal, but it quietly contains embedded data within its pixel structure. The only downside is that steganographic watermarking costs more. But when you’re aiming for Jason Bourne-level stealth, that’s to be expected. 

These invisible marks generally survive resizing, compression and minor edits. They don’t stop the scraping. They do strengthen authorship and authentication if disputes arise. And arise they do. 

In the AI era, attribution evidence matters. 

So-called “poisoning” tools such as Glaze and Nightshade, developed by researchers at the University of Chicago, introduce subtle, often imperceptible, modifications that make it harder for AI systems to accurately learn from or replicate your artistic style. 

Poisoning tools degrade or distort what the bots learn. Instead of blocking access, it attempts to corrupt the training signal, causing the AI model to misinterpret styles, features or labels derived from the poisoned data. 

These tools are not foolproof, and their effectiveness depends on how datasets are processed. However, they can make life a lot harder for the bots. 

Best practice: Apply them to public-facing original images, not your archival master files.