2026: the year art got tired of being flat
Something shifted in the art market this year, and it's not subtle.
Gallery owners are reporting that collectors are asking for texture they can feel. Interior designers are specifying "no flat digital prints." Art fairs are seeing a surge in impasto, encaustic, mixed-media — anything with physical surface presence.
The industry is calling it the "Anti-Algorithm" trend: a broad consumer rejection of the smooth, perfect, frictionless aesthetic that defined the first wave of AI-generated imagery.
The sentiment, as one art market forecast put it: we're done with the fake, the fast, and the flat.
What does "Anti-Algorithm" actually mean?
It's not anti-AI. That's the crucial distinction.
The backlash isn't against AI as a tool — it's against the default output of AI as a tool. The smoothness. The uncanny perfection. The way every surface looks like it was rendered rather than made. (I've been calling this the Wet Problem — everything looks slightly misted.)
What people are reaching for is:
Evidence of process. Brushstrokes, tool marks, drips, irregularities — anything that suggests a sequence of physical decisions.
Material honesty. Surfaces that look like they're made of something specific, not "generic digital medium."
Tactile invitation. The kind of image that makes you want to touch it.
In other words: people want materiality. They want art that implies a physical existence, even when they're looking at it on a screen. (For the deeper argument on why AI images lack this, see Why AI Doesn't Know the Weight of Paint.)
The opportunity for AI artists
The Anti-Algorithm trend doesn't mean AI art is out. It means generic AI art is out. The bar has moved from "can AI make a beautiful image?" (yes, solved) to "can AI make an image that feels real — that carries the weight and texture of a physical object?"
This is a solvable problem. And LoRAs are the key. The challenge breaks down into three layers:
Layer 1: Surface appearance (partially solved)
We already have good LoRAs for brushstroke styles, canvas overlays, and painterly effects. ClassipeintXL, OilpaintZ, Impasto LoRAs — excellent tools for visual texture. But appearance alone isn't enough. A Photoshop canvas texture filter also gives you appearance. The Anti-Algorithm audience can tell the difference.
Layer 2: Material behavior (emerging)
This is the frontier. Not "what does oil painting look like?" but "how does oil paint behave on linen canvas?" I explored this distinction fully in Style LoRA vs Texture LoRA — the key insight is that material behavior LoRAs train on the physics of specific materials rather than the aesthetics of specific styles. Surfaces get logic. Cracks follow stress patterns. Gold doesn't just glow — it fractures along fault lines.
Layer 3: Scale consistency (the open frontier)
The hardest challenge. A truly convincing physical texture should work at macro, mid-range, and full-frame scales. This is exactly what the 3-Distance Method is designed to address — training datasets that capture material at three scales so the model learns consistent behavior from close-up grain to environmental context. Most current LoRAs work at one, maybe two distances.
Commercial implications
Here's the part that matters for anyone generating commercially:
The market for "AI art prints" is saturated. You can buy AI-generated wall art for $15 on Etsy. The margins are collapsing because the output all looks the same — smooth, stylized, interchangeable.
But the market for textured, material-specific AI art — output that looks like it could be a photograph of a physical artwork — is wide open. Because almost nobody is generating it yet.
Think about it from the buyer's perspective:
Generic AI art print: "That's a nice image." → $15
AI-generated image with convincing gold leaf texture, visible material behavior, print-resolution detail: "Wait, is that a photograph of an actual painting?" → Very different price point.
The [Commercial Use OK] label on models suddenly matters a lot more when the output quality crosses the threshold from "digital print" to "convincing reproduction."
The bigger picture
The Anti-Algorithm trend is actually great news for AI artists. It raises the bar, which means it clears out the low-effort noise and creates space for people who care about craft.
If you're already thinking about texture, materiality, and physical presence in your generations, you're ahead of the curve. The market is coming to you.
The future of AI art isn't smoother, more photorealistic, more perfect output. It's output that carries the evidence of material reality — even when it was never physically made.
Physical texture is the new frontier. The algorithm is learning to be anti-algorithm.
Start here:
SHIFUKU Gold Leaf v1 — Physical Texture LoRA for SDXL (Commercial Use OK)
SHIFUKU Gold Leaf v2 — Kiribaku / Sunago / Noge — 3 gold leaf techniques
SHIFUKU Kintsugi — Physical Texture LoRA for SDXL (Commercial Use OK)
SHIFUKU Hamon Steel (Beta) — Japanese sword textures
Read next:
Why Does Every AI Image Look Like Plastic? — the user-facing texture problem
"This Is Not a Style LoRA" — Soul = Materiality — why "soul" lives in the surface
Style LoRA vs Texture LoRA — They Solve Different Problems — the technical core
Training Texture LoRAs from Real Materials: A 3-Distance Method — if you want to build your own
Tags: art trends 2026 anti-algorithm texture matière commercial use physical materials LoRA SDXL AI art market musing