1. Know your customers
Direct relationships are your first defense: build email lists, own your own site, create member access, share private previews, and consider more local opportunities. If you don’t know who’s viewing your work, you don’t control it. Access should feel personal: “Request to view”, “Collector preview”, “Private release”, and protected sharing.
2. Own your primary archive, and don't make it public
Your full body of work should live somewhere you control: your website, gated series and past work, Social platforms and marketplaces are satellites, not homes. Keep your archive/high resolution images private separate archive from promotion. Public-facing images should be partial, contextual, and intentional. Your archive is not your feed.
3. Fragment public exposure generally by rotating and removing
In public spaces show 2–3 images max never full series and never high resolution
Public is for discovery, not consumption. Never have a full archive of work online, especially in chronological order. Friction protects value and blocks automation. Avoid scrollable galleries for complete series. PDFs invite slow interaction and more intentional viewing. Avoid scrollable galleries for complete series. PDFs show sequencing and are harder to scrape and easier to gate.
Think portfolios, not feeds
4. Manage circulation and access
Require something for the user for indepth views, larger or more images. Do a softwall that will gate access by default. This allows you understand who is viewing your art and make a connection. Require minimal sign up: email and consent for emails. Allow light browsing, then pause “View more” or “Request access”
5. Use PDFs for full bodies of work
Never have a full archive of work online, especially in chronological order. Friction protects value and blocks automation. Avoid scrollable galleries for complete series. PDFs invite slow interaction and more intentional viewing. Avoid scrollable galleries for complete series. PDFs show sequencing and are harder to scrape and easier to gate
Think portfolios, not feeds.
6. Ambiguity is a defense. Avoid clean, trainable images that are high-resolution
Do not post: flat scans, perfect crops, straight-on documentation. Instead: angled shots, partial views, visible edges, layered backgrounds. Ambiguity is a defense. Strip unnecessary metadata, even data like "painting" makes it clear it is art. Remove excess metadata from public files. Modern image-training pipelines strongly prefer:
flat, front-facing images, high resolution, clean edges, minimal background noise, consistent framing. Those images are easier to: scrape, label, normalize, and learn from.
7. Embed lived context, Context resists extraction and preserves meaning.
Always show work within space with furniture, walls, hands, shadows, natural light. Models can still learn, but the signal-to-noise ratio drops. More importantly, meaning is preserved for humans while value drops for extractors.
Context doesn’t block AI. It de-optimizes extraction. Angled shots, partial views, reflections, edges, and complex backgrounds reduce training efficiency and increase preprocessing cost. They do not make training impossible, but they raise friction enough that mass systems skip them when better data exists.
8. Favor video over stills
Training systems want the subject isolated. Context does the opposite. Video is harder to scrape and less useful for training (for now). Think slow pans, walk-throughs, handheld movement, any kind of movement protects nuance. Models can still learn, but the signal-to-noise ratio drops. More importantly, meaning is preserved for humans while value drops for extractors. Still images are cheaper to scrape, easier to store, easier to label, easier to reuse for training. Video adds temporal complexity, motion blur, changing lighting, compression artifacts, and higher storage cost.
Video + Context doesn’t block AI. It de-optimizes extraction (at least for now)
9. Avoid bulk uploads and anything done consistently
Do not upload large batches publicly. AI loves volume. Humans prefer pacing. Break consistency lighting, framing, enviornments, and scale. Consistency helps algorithms. Variation protects artists. Consciously introduce visual noise, foreground elements, reflections, etc. Natural complexity reduces extraction and training value.
10. Delay publication
Do not post work immediately after completion. protect first circulation, preserve surprise time is a form of scarcity. Art thrives on context, ambiguity, pacing, and relationships.
The takeaway is this...
There are no magic shields. Tactics like angled shots, context, or video are not locks, they are friction. AI systems optimize for the cheapest, cleanest data at the largest scale. Artists don’t win by trying to block everything, but by lowering training value, increasing preprocessing cost, and preserving human meaning. When work becomes harder to isolate, harder to normalize, and harder to batch, it gets skipped in favor of easier sources. The strongest protection doesn’t come from any single visual trick, but from layering strategies: gating combined with low resolution, context paired with fragmentation, delayed release supported by real relationships. Visual tactics alone are weak. Access control and intentional circulation design are where real protection now lives.