Provenance
The Print Is Real. The Room Isn't.
Some of the room photographs on this site are not photographs. They are AI-generated visualisations of what one of my prints might look like on a real wall. The print itself, the photograph you would actually receive, is unchanged: it is the same image I made in the field, printed the same way it has always been printed. Only the room around it is invented.
I could have said nothing. Nothing requires a shop to explain how its lifestyle photography was made, and most people looking at a nicely lit lounge with a print above the sofa would have no reason to ask. I've decided to say something anyway, and I want to explain why.
Why I'm Saying So
AI · Consumer Trust
A Question The Industry Hasn't Answered
Picture a lounge that doesn't exist: a sofa no one has sat on, light falling through a window that was never built, a print above it that is completely real. That's the room you're looking at more often than you think. AI-generated imagery and video are moving into everyday marketing faster than most audiences can track, and I don't think that shift is neutral. A room set, a lifestyle shot, a listing photo, any of it can now be generated in minutes, and increasingly is, with no consistent standard for disclosing it. For a creative industry built on the labour of photographers, stylists and artists, that matters: the same tools lowering the cost of an image are also eroding the distinction between work that was made and work that was generated, and the industry has been slow to agree on how, or whether, to mark that difference. I think it needs to. As the tools improve, the visual gap between a photograph and a rendering narrows, while the gap in how that difference gets communicated does not, and closing it shouldn't be optional. That's the space where I've chosen to place a small, deliberate marker of my own.
A New Capability
Small Business · AI
Why This Matters Beyond Me
There's a real creative pleasure in this part of the process too, and I don't want to leave it out. My own sense of home has always come from being outdoors, from the land itself, rather than from any interior. But imagining a room around a print, working out what light does to a wall, what a home could feel like around a piece of work, is its own kind of making. It sits closer to staging a shoot than to faking one: a stylist arranging a real room for a real photograph is still deciding what home could look like, and no one calls that dishonest. The only difference here is that mine happens inside software instead of a lounge, which is exactly why the label matters as much as the image does.
For small businesses and independent creators, that capability changes what's possible. Showing an idea properly, a print on a real wall, a room that doesn't exist yet, used to require a studio, a stylist and a budget only bigger brands could justify, or something I'd schedule four times a year with intense planning and real cost behind it. Now a business with none of that can show its work with the same clarity and ambition as one that has it, and I think that's worth celebrating. It puts something genuinely useful into hands it was never built for. But that only holds if it's used honestly: the same openness that makes this trustworthy for me has to hold for anyone using these tools, or what should level the field just becomes another way of hiding what's real.
There's a more technical layer to this too, one most people would never think to check. Every AI-generated image, mine included, carries an embedded metadata tag identifying it as AI-produced, written into the file itself at the moment it's made; most businesses and individuals carry that same tag without ever realising it's there. That tag can be removed. Some businesses do exactly that, stripping it out so the image looks and behaves like an ordinary photograph with no trace of how it was made. I could do the same. Realising that made me want to do something more useful than leave the tag in place and hope you notice: teach you how to tell for yourself, whether or not you ever have a tag to check.
A Short Guide
If you're ever unsure whether an image, mine or anyone else's, is real, here's what I'd look for.
- Start with the caption and alt text. A business being upfront about AI use should say so plainly, right next to the image, not buried in a policy page. If a caption seems to be avoiding the question, that's worth noticing too. On desktop, right-clicking an image and choosing Inspect will usually show you its alt text even when nothing is visible on the page itself, and some platforms, Pinterest among them, now label AI-modified images automatically.
- Check the file's metadata if you can. Downloading an image and opening its details (on most computers this is right-click, then Properties or Get Info) sometimes surfaces a field called Digital Source Type, part of the IPTC photo metadata standard, or a Content Credentials icon under the newer C2PA standard that some cameras and platforms now support. Two values worth knowing: "trained algorithmic media" means the image came entirely from a generative model, and "composite with trained algorithmic media" means AI-generated elements were combined with real photography. If you're comfortable with a command line, a free tool called ExifTool will show you this data directly, including fields a basic file browser doesn't surface. Not every platform preserves this data, and plenty of ordinary editing tools strip it out entirely, so not finding a tag proves nothing. Finding one, though, is a clear answer.
- Look at where surfaces and objects meet. This is still where most AI-generated interiors give themselves away. A skirting board that doesn't quite line up where two walls meet, a shadow falling in a direction the light source in the room doesn't support, a rug or tiled floor whose pattern warps or repeats too perfectly toward the edge of the frame, a plug socket or light switch that's misshapen or simply missing. Look closely at anything with fine repeating detail, herringbone flooring, woven cushions, venetian blinds, since that's where the pattern tends to break down first.
- Check any text or fine detail in the scene. Book spines, the label on a candle, numbers on a clock face. AI image tools are still notably bad at rendering legible text, so a blurred, warped or nonsensical word tucked into the background of a room is one of the more reliable tells going.
- Look at reflections and glass. Windows, mirrors, glossy tabletops and anything else that should be reflecting the room around it are where AI-generated images most often show their seams, a reflection that's slightly wrong, or missing entirely where one should exist.
- Look closely at hands, ears and teeth if a person appears in the image. These are still the areas generative models get wrong most often: an extra or fused finger, an ear that doesn't match its pair, teeth that are too uniform or don't quite sit against the jawline. Interior shots without people won't have this tell, but it's one of the fastest checks going when a person is in frame.
- Try a reverse image search. Pasting the image into a reverse image search tool will occasionally turn up the exact same room, or one very like it, being used elsewhere, sometimes a sign the image is a stock render rather than a photograph of anywhere real.
- Run it through a dedicated AI-detection tool for a second opinion. A number of free services will analyse an image and return a likelihood that it was AI-generated. Treat the result as a data point rather than a verdict: these tools are improving quickly but still produce false positives and false negatives, particularly on images that mix real photography with generated elements.
- Ask. If a business is genuinely honest about its process, a direct question, "is this room real", should get you a direct answer. If it doesn't, that's an answer in itself.
I don't think using AI in my process is the interesting part. I think what happens to that information afterwards is. That question, what happens to an object's history once it leaves the hands that made it, is the one I've been sitting with for a while now, first as a gallerist watching work change hands with less and less of its story intact, and now as research. It's also the starting point for Provena. I've spent the last year and a half studying it, and I'm now in the final stages of looking closely at its impact, a way of keeping that kind of information attached to a piece as it moves through the world. I'm not ready to talk about it properly yet. But the instinct behind it and the instinct behind this small disclosure are the same one: if something matters about how a piece was made, it shouldn't be left to chance whether that information survives.
For now, that means a plain rule on this site. If a room is AI-generated, it says so. If a photograph is mine, it's mine, unaltered, printed the way it always has been. Nothing more complicated than that.