Canva spent real money on a pigeon named Picci1 — a stop-motion puppet, animated one frame at a time, rebranding an Italian trattoria in a piazza. The pitch: design is human, imperfect, joyful. The catch: the ad runs inside a product that is actively shipping AI generation to millions of users.
Call it what it is, but don't call it hypocrisy. Hypocrisy is a moral failure. This is a tell.
When a company automating handcraft chooses handcraft as its marketing, "craft" has already completed a quiet migration — from a value you practice to a position you occupy. The pigeon isn't a lie about Canva's product. It's an accurate signal about what craft has become: available for licensing.
The pigeon isn't a lie about Canva's product. It's an accurate signal about what craft has become: available for licensing.
One commentator noted that, per that piece, the "recent backlash against AI-linked brands using handcrafted art in their branding, as seen with GitHub Copilot," reportedly highlights the potential pitfalls of this approach. The move is legible enough to have a precedent. That's the problem. A move with a precedent is a genre.
Here's the thesis, up front: the craft counter-movement against AI sameness is already eating itself. The moment visible human labor becomes a differentiator, it becomes a look — and a look is exactly what AI averages best.
Watch how fast the current commoditizes itself.
WGSN has already forecast imperfection as a sellable trend, naming the "Renaissance of Real" for A/W 26/27 — worn finishes, stained cuffs, intentional abrasions at Prada, raw-edge tailoring with the construction lines "highlighted rather than concealed." Hannah Watkins of WGSN told Fashionista that "design is increasingly signalling its human origin."
Read that verb again: signalling. Not being human. Signalling human origin. That is a forecast, sold to brands, telling them which markers of humanity will move product two seasons out.
The anti-design press says the same thing in a warmer voice. "Perfection catches the eye, but imperfection catches the heart," offers one managing director, while the same piece admits, in unattributed editorial text, that the real skill is knowing "when to showcase beautiful craftsmanship and when to let their unique personality shine through." Imperfection as a scheduling decision. A slot in the media plan.
And the moment the market wants "handmade" cues at scale, the analogue infrastructure to actually produce them isn't there. So designers fake it.
Elizabeth Goodspeed named the mechanism: most designers don't have the time, tools, or support for fully analogue work, so they simulate handmade cues digitally. She compared it to penciling in a beauty mark — an intentional imperfection designed to signal authenticity, not the result of any actual process.
This is the counter-movement's autoimmune disease. The whole point was that a machine can't fake the hand. But a designer can — with a rough-texture overlay, a sketchy-florals filter, a #WorkInProgress print — and once designers are simulating the hand digitally, the loop closes. There is nothing left for the AI to fail to replicate, because the human already replicated it themselves.
Fred McHale draws the line that actually matters, and it isn't between rough and smooth. He argues AI's sameness isn't a sign that it's "too polished." It's a statistical tendency — models converge on overlapping data, average out the edges, land on narrow defaults — multiplied by sheer volume.
Which means the fix was never to make things rougher. Rougher is just a different default.
His sharpest sentence: slapping a rough texture on a design to signal "not AI" and generating a hyper-polished image without thinking are "two sides of the same coin." The aesthetic differs. The thoughtlessness is identical.
That's the whole game. Slop and rough-for-signal are both what you get when you skip the creative decision — one skips it toward polish, the other toward grime. Both are averageable, because both are outputs of a process with the thinking removed.
So the scarce good was never roughness, and it was never the handmade look. Craft-washing strips out the deliberate decision just as thoroughly as a lazy prompt does. It just leaves a different residue.
This extends a point worth stating directly: the durable differentiator in an age of generative abundance is judgment — the creative decision itself. Visible human labor looked, for a hot minute, like the next scarce good. It isn't, if the labor is only penciled in. A market that can average judgment out of a smooth image can average it out of a rough one too.
But some cases don't average. And the reason they don't is instructive.
Fernando Laposse insists "the community is the client." Not a phrase — a structure. He refuses to adapt existing crafts for a global market, because that's the pattern where designers "enter communities, exploit their skills, and leave." Instead he builds new systems rooted in agriculture — corn husks into Totomoxtle, agave fiber harvested by pruning rather than killing the plant — so the benefit stays local.
You cannot pencil that in. There's no texture overlay for "the farmers grow heritage corn again." The labor is structural, not cosmetic, and structure resists being reduced to a look because the look was never the point.
Priyanka Narula's Wicker Story runs the same logic from the other direction. She translates parametric algorithms into instructions a bamboo weaver can execute by hand — and, crucially, the algorithm bends to the weaver's process, not the reverse. The machine is a notebook. The human is the visionary. That hierarchy is doing the work.
The shokunin makes the point older than any trend report. The ideogram itself puts an ear — 耳, mimi — beside the person, because the first prerequisite of the craftsperson is the capacity to listen: to the tools, to the material, to the community served. Craft as attunement, not surface.
None of these can become "a look" without becoming something else entirely. A Kaikado tea caddy built from copper weathered on a shrine roof for 100 years has a chief feature — it takes a century to mature — that no rough-texture filter and no A/W 26/27 print can simulate, because the feature is time, borne by an actual process.
Compare that to the pigeon.
Picci1 is genuinely handmade. Real puppetry, real frames, real labor. And it still functions as a look — because its job in the campaign is to signal human warmth for a product selling automation. The labor is real; the position is decorative. That's the trap in its purest form: even authentic craft becomes craft-washing the instant its structural role is to market its own opposite.
Which gives us the actual dividing line, and it's not medium, and it's not effort, and it's not polish.
Visible labor scales as a counter-position only where the labor is real and structural — where it changes who gets paid, what grows in the field, whose process the algorithm serves. Where the labor is a texture penciled onto the surface to read as "not AI," the market does exactly what markets do. It notices the differentiator, names it a trend, forecasts it for A/W 26/27, and averages it into next season's default.
Roughness is not a counter-position. It's a style, and styles are the most averageable thing there is.
Craft is a counter-position only when it's a process — one with real decisions and real consequences that a model can't reverse-engineer from the output, because the output was never where the meaning lived.
Everything else is a beauty mark. And a beauty mark, penciled in, is just prompt engineering with better lighting.