AI Isn't Only Homogenising Culture. It's Also Lowering the Cost of Building It.

Part of what I actually do at The Changing Booth is curation, going looking for digital spaces worth someone's time and attention, filtering out the noise, deciding what's worth surfacing on a blog nobody is paying me to run. Most weeks that means me finding something. Recently, something found me instead.

A fellow founder got in touch after his platform showed up as a recommendation inside AI-generated search results that linked back to a post on TCB's blog. Someone had asked an AI assistant something like where to find independent art online, and the answer pointed to something I'd written months earlier, which then pointed onward to his platform. That's a genuinely new discovery path. It didn't depend on his SEO budget or mine, or either of us paying for reach, or either of us having a following large enough to get picked up by an algorithm optimising for engagement. It depended on the content actually being specific and relevant enough for a language model to surface it in response to a real question.

The platform is GalleryTalk, which gives independent artists and small galleries their own three-dimensional exhibition space online, for under seven euros a month, fully mobile, no technical setup required. Over four thousand artists have built spaces there. Building the equivalent of that five years ago would have meant hiring a developer, paying for hosting, and maintaining something that broke every time a browser updated. It's on TCB's recommendations now, which is where this whole story started for me, but the product itself isn't really the point.

What struck me was the pattern behind it. The mainstream conversation about AI and culture follows a fairly consistent script at this point: algorithms flatten taste, recommendation engines reward whatever generates the most short-form, high-dopamine engagement, human curation gets squeezed out of the loop entirely, and everything online eventually starts to look and sound the same. I don't think that concern is wrong. I see versions of it constantly, in exactly the platforms that concern describes. But from inside TCB, building something small with limited resources, I keep running into a different pattern that doesn't fit that script at all.

What I actually see is solo founders and small teams using AI as infrastructure, not as the thing deciding what an audience sees, but as the thing that makes it financially possible to build something narrow enough to matter to a small audience in the first place. Five years ago, a founder wanting to give independent artists a proper online exhibition space would have needed a development team, a much larger budget, and enough scale to justify both. Today the cost of building that specific, non-mass-market thing has dropped enough that one person can do it. That's not AI deciding what culture looks like. That's AI lowering the barrier to entry for the kind of cultural platform that was never going to be financially viable at scale in the first place.

The art world has always been one of the more closed, gatekept spaces there is. Representation by the right gallery, presence in the right cities, the right dealer relationships built up over years, sometimes generations. None of that changes because AI made it cheaper to build a website. Platforms like GalleryTalk don't dismantle those gates. What they do is create entry points that exist entirely outside them, new mediums for showing work, new conversations about culture that don't require anyone's permission to happen.

I don't think GalleryTalk is a one-off either. The same pattern shows up anywhere a cultural niche used to require real capital or real industry connections just to exist online. A small press putting out three books a year, a niche music label, a regional theatre company that wants an audience beyond its own city, none of these were ever going to justify a custom-built platform or a development team. They either found a workaround, usually a generic template that didn't fit what they were actually doing, or they didn't bother building a digital presence worth visiting at all. What's changed isn't that any of these fields became less niche. It's that the cost of building something specific enough to actually serve a niche dropped to a level a single person can absorb.

The dominant framing around AI right now is almost entirely about efficiency: saving time, cutting costs, optimising for revenue, doing more with fewer people. That story is true, and it's the one that gets funded, reported on, and repeated. But there's another story running alongside it that gets far less attention, about access rather than efficiency, about niche conversations that can now exist because the cost of starting one dropped low enough for an individual to absorb it, about closed cultural spaces getting cracked open at the edges by people building things that were never primarily about the money.

What if AI isn't only the thing that homogenises culture and centres everything around monetisation, but also a vehicle for diversity of thought, access and equity within culture?