How to Prevent Your Art from Being Used in AI
People who type how to prevent your art from being used in AI usually mean three different things at once. Someone scrapes posts for training. Someone prompts for a look that feels like yours. Someone already has your PNG and runs an AI edit or wipe. One tool does not cover all three. A short stack does more.
You cannot lock the open web forever. You can make training scrapes less useful, make style learning noisier, and make a clean anonymous reuse of your posted file more expensive. Glaze and Nightshade sit on the training side. Jingwei sits on the posted-file side. Use them as separate steps when both risks matter.
Name the risk before you pick a tool
Training scrapes pull public posts into model data after you upload. Glaze and Nightshade were built for that lane. They run as local apps, often need a strong GPU, take longer per image, and leave a more visible perturbation.
Style mimicry is close to that lane. Someone prompts for your palette, line weight, or name. Glaze is the tool most illustrators hear about first for that job. Jingwei does not replace it.
File reuse is different. Someone downloads what you posted, then inpaints, fuses, or erases a mark. Corner logos and soft overlays are easy to wipe. Jingwei's visible layers change pixels so a forced fill tends to leave smears and warps. A JW claim can still be readable on Verify when the signal survives.
The broader stack is in How to protect your art from AI. Style-only detail is in Protect art from AI style mimicry. Day-to-day watermark habits are in How to watermark your art.
What actually slows AI use of your work
1. Keep the unmarked master offline. Post a sized export, not the largest source.
2. If training scrapes worry you, run Glaze or Nightshade on the export you are willing to show, using that project's own settings.
3. If edits, reposts, and watermark removal worry you, run that public export through Jingwei Protect. Quick Credit for a light everyday credit and JW claim. Manual when you expect a targeted wipe or face edit.
4. Download PNG when you care about invisible layers. Verify once so you know what a readable claim looks like for that file.
5. Save the post URL, the time, and a screenshot of the live post.
Step 2 and step 3 are different exports. This site does not merge them into one button, and it does not promise that one step preserves the other's effect.
What prevention cannot promise
No stack can promise that every future model or eraser will fail. Determined redraw can still damage marks. Heavy compression and full-image repaint can weaken invisible signals. Screenshots sometimes still verify, and sometimes do not.
A JW claim is evidence and stated intent. It is not independent legal proof of ownership. Copyright rules depend on where you live. Registration and legal advice sit outside any watermark tool.
robots.txt and a site notice are polite signals. They do not stop hostile scraping by themselves.
Can I stop AI from training on my art completely?
No public posting setup can guarantee that. Glaze and Nightshade make style learning and scrape usefulness harder for many current pipelines. Keep originals offline and post only protected exports.
Is a watermark enough to prevent AI use?
A corner logo usually is not. A stronger visible mark raises wipe cost on the file you posted. It does not replace anti-training tools when scrapes are the main fear. Pair the jobs when you have both risks.
Where Jingwei fits in this stack
Jingwei runs in the browser. Everyday Protect needs no account. Processing is classical image processing, ships no deep-learning model files, takes about 30 seconds, and deletes the upload after the job. Use it on the file you are about to publish when someone holding that file is the worry.
Protect a public export on Jingwei