The bot and Monet
This week in AI & marketing
Every major technology follows the same arc. It shows up as a tool, useful but optional, something you adopt or ignore. Then, at some point, it stops being a thing you use and becomes part of the air everyone breathes. It shapes how people talk, what they trust, and who they signal allegiance to.
Social media made that jump fifteen years ago. It started as a way to share photos and ended up rewiring elections, identity, and the news. AI is making the same jump right now, and this week made it obvious.
The clearest signal was the Monet story recently.
Someone posted a real Monet painting on X, labeled it AI-generated, and thousands of people lined up to explain why the “synthetic” brushwork and “soulless” composition gave it away. It was a real Monet. One part of this story is about how culturally uneducated people are these days, so that they can’t name a very famous artwork. But that’s just half of the story.
Research now backs up what that moment showed: people downgrade the exact same work the second they believe a machine touched it, even when they preferred it moments earlier. That is not an aesthetic judgment. It is a cultural one. Rejecting AI has become a way to signal taste, identity, and your side.
Marketers who pride themselves on trying the newest tools first now publicly claim they don’t use AI for copywriting. Even when they do. Most of them do. But “doing it manually” is the new European vacation, the new vintage shopping, or whatever the current cultural hobby of the upper-middle class is these days.
That is what a technology looks like when it crosses from tool to culture. The arguments stop being about whether it works and start being about what using it says about you.
— Torsten and Peter
PS: We, of course, are happy to admit that we use AI wherever we can. Not just because we don’t give a flying f about cultural norms (we don’t, never were), but because it just works, regardless of what humans think about it. And while craftsmanship has its beauty, it also has its limits.
1. The Monet Trap and Why the AI Backlash Is More About Culture Than Quality
A few days ago, someone posted a real Claude Monet painting on X, labeled it as “AI-generated,” and asked people to critique why it was inferior to a real Monet.
The replies were incredible. Thousands of people confidently dissected the “AI slop”: the reflections were “incoherent,” the composition lacked depth, the colors felt “soulless,” and the brushwork was “obviously synthetic.” One person reportedly wrote an 800-word breakdown explaining why the fake Monet failed to capture Monet’s genius.
It was an actual Monet.
Studies now show people consistently downgrade art, writing, and music once they believe AI was involved, even when they preferred the exact same piece moments earlier without the label.
In other words: a large chunk of the current AI backlash is not aesthetic judgment. It’s social signaling. It’s identity. It’s zeitgeist.
That doesn’t mean the backlash is fake. It’s very real. But it also means the “humans will always instantly spot AI” narrative is getting shakier by the week.
And yes, some companies will absolutely misread this and conclude “people secretly love AI content.” Careful.
The real lesson is subtler: audiences increasingly judge context as much as output.
Which means the brands that win won’t be the ones hiding AI usage. They’ll be the ones framing it intelligently.
2. Google Just Redesigned the Search Box. Your Traffic Problem Is About to Get Much Worse.
Google killed one of the internet’s most important design conventions this week: the classic search box.
The new version is multimodal, conversational, expandable, and deeply AI-native. You can now throw PDFs, videos, images, and even browser tabs directly into Search. AI Overviews and AI Mode are merging into one continuous experience. The “ten blue links” era is ending right in front of us.
And frankly, this is terrible news for publishers and most websites.
Google’s new flow is designed to keep users inside Google for as long as possible. Ask a question, get a synthesized answer, ask follow-ups, refine the request, generate visualizations, maybe even spin up mini apps directly in Search. Fewer clicks. Fewer visits. Fewer opportunities for your carefully optimized landing page to matter.
The important shift here is behavioral. For 25 years, Google trained people to compress intent into keywords. Now it wants them to think out loud. Longer queries, richer context, ongoing conversations.
That changes SEO fundamentally because optimization shifts from keyword matching toward semantic usefulness and machine readability.
If your content strategy still revolves around “ranking for terms,” you’re already playing the old game. The question now is: can an AI system easily extract your expertise, your product positioning, your structured data, your narratives? If not, you risk becoming invisible while still technically ranking.
And yes, Google says these AI experiences “increase engagement.” They probably do. Just not for your site.
3. Gemini Omni Is Not Just a Video Model.
Last week, we teased Google’s new Omni model as “a new video model.”
That was technically true. It was also underselling the story quite badly.
Announced officially at Google I/O, Gemini Omni is Google’s first true any-to-any multimodal model. Text, image, audio, video in. Video, image, audio out. One system, one conversational editing surface, one model reasoning across all modalities simultaneously.
For years, AI creative workflows have looked like Frankenstein pipelines: one model for images, another for lip sync, another for motion, another for sound, another for editing. Omni is Google's attempt to unify all of that into a single foundation layer, where the model actually understands relationships between modalities rather than stitching outputs together afterward.
The examples are honestly absurd. Generate a claymation explainer about protein folding. Turn a real-world clip into retro-futuristic sci-fi. Edit a scene conversationally across multiple turns while preserving continuity and physics. Add synchronized music and voiceovers on the fly. It feels less like prompting and more like directing.
The enterprise implications are huge, especially for marketing and L&D teams drowning in asset requests. Training videos, localized ad variants, product explainers, sales collateral, internal onboarding, the list is long and comprehensive.
4. Google’s Universal Cart Is the Clearest Agentic Commerce Signal Yet
Everyone is talking about AI agents helping you shop. Google is one of the first companies actually wiring the infrastructure together.
The newly announced Universal Cart follows you across Search, Gemini, YouTube, and Gmail. Add products anywhere inside Google’s ecosystem, and they persist inside a single AI-native cart that monitors price drops, checks compatibility issues, tracks stock availability, surfaces loyalty perks, and eventually handles checkout.
In other words: Google wants the cart itself to become an intelligent layer.
And because this sits on top of Google Wallet, the system also supports payments, offers, BNPL options, loyalty programs, and, eventually, autonomous purchasing rules via AP2, Google’s new Agent Payments Protocol.
If you run ecommerce, the important part isn’t “AI shopping assistant.” It’s that Google is becoming the orchestration layer between customer intent and merchant fulfillment.
The merchant technically remains the merchant of record. But the customer relationship starts shifting toward whoever controls discovery, comparison, recommendation, and checkout orchestration. That’s a very different power dynamic.
We’ve been talking for months about “agent readiness.” This is what it actually looks like operationally.
5. Figma’s Design Agent Is Extremely Good (And Potentially Awful News for Junior Designers).
For the last two years, most “AI design tools” were basically image generators pretending to understand UI work.
Figma’s new Design Agent feels different.
The key insight is simple: the agent lives inside the actual file. It understands your components, variables, design tokens, naming conventions, libraries, and systems. Instead of generating random pretty screens in a separate chat window, it edits real production work directly on the canvas.
The agent can generate layout directions, remix flows, bulk-edit components, summarize feedback threads, apply spacing changes across massive files, convert designs to dark mode, rename variables, and iterate visually while you keep working. Less “magic AI generator.” More “extremely fast junior designer who never gets tired of cleanup work.”
Which is exactly why junior designers should probably pay attention.
The reality is that a huge amount of entry-level creative work is mechanical. Padding adjustments, variant swaps, documentation, asset resizing, and consistency cleanup. The Figma agent attacks that layer directly. Not concepting or taste, or strategic thinking. The repetitive production layer.
That doesn’t mean designers disappear, but the value shifts upward faster.
Share with a colleague. Earn paid membership months.
Google Just Redesigned the Search Box. Your Traffic Problem Is About to Get Much Worse. - I disagree. Once upon a time businesses used websites to be found and sell. then an industry sprung up disintermediating our offers from our clients. Content. Instead of working I create content. I could only be found if my content was good. Now I have an opportunity to focus on making my offer found, so that when someone searches for 'customer experience consultant' I show up instead of my article, blog, post about the importance of something or another or my listicle. Sure I have to revamp my website AGAIN, but I am looking forward to this one.