Hey AI, you can show me in search results, but don't steal my work

If you put original images or video on a Bernard site, our planned protection will give you a stronger position when somebody uses your work without permission.

It will give you five practical things:

• your confirmed name and rights attached to the public file;
• a public statement of what other people may do with it;
• a private original, fingerprint and dated history kept by Bernard;
• a barrier against training crawlers that identify themselves; and
• an evidence pack if you find a copy elsewhere.

This is what intellectual property protection can usefully mean on a website. It makes your claim clear before a dispute, puts some collectors outside the gate, and keeps the records you may need afterwards.

It cannot make copying impossible. A person can take a screenshot. A crawler can lie about its identity. Metadata can be stripped from a later copy. A website cannot decide a legal dispute. Bernard will preserve the facts. A platform or court decides what they prove.

What is live today

Updated November 2026. Some of this is now working on every Bernard site.

• Private details come out. When you upload a photo, Bernard removes the location, the camera's serial number and the camera owner's name from the copy your site shows. New photos brought over from an old site get the same treatment.
• Your name goes in, once you confirm the photo is yours. For a photo you have marked as yours, Bernard writes your name and copyright line into the copies it stores from then on, in the standard IPTC fields. A photo whose rights are unknown gets nothing. Copies published before you marked a photo are brought up to date in a later step.
• "Not for AI training" goes in when you ask. While your site's Protect images and video switch is on, those photos also carry the IPTC data-mining field set to prohibited for AI and machine-learning training.
• The original is kept. For each new photo, Bernard keeps the untouched original privately, with a fingerprint of its bytes and the date it arrived.
• Your choice is recorded. Each site has two switches, one for images and video and one for page words, and Bernard follows them as each step arrives.

Everything else on this page is still being built: the rights page, the instructions to training crawlers, refusing the crawlers that identify themselves, the evidence pack, the search for copies and the rights fields in video.

Your work leaves with your name on it

The public copy of a protected image will carry the confirmed creator, copyright notice, credit, usage terms and a link to its rights page. These are standard IPTC fields that photo software and search engines can read. Video has an equivalent set of fields.

The rights page will say whether the work is available to licence and how to ask. It will also state the owner's choice about AI training. Google says the same embedded image information can make an image eligible for its Licensable badge. Google still decides whether to show it.

Bernard will write ownership into a file only after the owner confirms it. Putting a stock photograph on your site does not make it yours.

Bernard keeps the original and the history

Bernard will keep the uploaded original private. The website gets a clean public copy with location and device details removed. That copy gains the confirmed rights information instead.

Alongside the file, Bernard will record a fingerprint of its bytes, when it arrived, who the rights record names and each later version. Replacing an image or video will add to that history instead of rewriting it.

If you find a copy elsewhere, you will be able to download those records. They can support a report to a marketplace, a discussion with a solicitor or a claim. Bernard supplies the record. A platform or court decides whether it proves authorship or infringement.

Images will also get periodic searches for likely public copies. General reverse search for video is not ready. The first video version will search its poster image and let the owner record a sighting, with each result labelled for what it is.

Search can stay open while training stays closed

Artists need search engines and useful answer services to find their work. They do not have to make the same choice for AI training.

AI companies already identify different crawlers for different jobs. Anthropic, for example, names separate crawlers for training, search and a user's request. Cloudflare uses the broad categories Search, Agent and Training.

Bernard's intended default for a maker is to welcome search and live answers, then refuse declared training crawlers. The site will publish the same choice in machine-readable instructions. Cloudflare's Content Signals express it as search=yes, ai-input=yes, ai-train=no.

The owner will choose separately for visual work and for the words on the site. A photographer may keep her shop visible everywhere while reserving her photographs. A teacher who publishes tutorials may choose differently.

The instructions alone are voluntary. Bernard will also refuse requests from training crawlers that identify themselves. Its report will distinguish verified crawlers from visitors that merely claim a familiar name, and count images, video and pages separately.

Why we are building it

We saw a maker on Threads with about 30,000 followers. He earns part of his living through Etsy and said he would leave if there were a credible alternative.

That maker needs the internet to find buyers. The same open door lets somebody copy a lino cut onto a poster or collect thousands of images for a training set. He has already done the creative work. His website should help people find it while keeping his terms clear.

Right-click blocking would inconvenience his customers without stopping a copy. A visible watermark would spoil the work he is trying to show. Glaze, Nightshade and commercial watermarking suit particular artists and files. They are outside Bernard's basic protection.

Bernard's promise will stay narrower: name the confirmed maker, publish her, their or his terms, clean the public file, refuse declared training crawlers and keep the receipts when prevention fails. Images and hosted video will receive the same core protection, with the limits of each medium stated clearly.

This is being built now. It is not live yet. We will update this page as each part passes its file tests, legal review and real-world checks.

Bernard is an AI-assisted website and marketing platform for small businesses. He builds a new site or moves an existing one, keeps it fast and readable to search and AI tools, and looks after it. Once measuring is on, he reads your numbers each week and suggests what to do next, with a button that does it. On the Nautilus Medium and Nautilus Large plans he also plans your marketing every six weeks and drafts your social posts, emails and pages - but only if you want him to. Everything lands on a draft you approve. One flat price per site.

---

See what Bernard would do with your business. Already have a website? Paste its address and he measures it up, free, and shows you the new place before you decide anything. Starting from nothing? Tell him what you are building.