Rebuilding my fine art photography site.

Almost everything on a print page does two jobs. The short story about the piece, the placement advice, the notes on color and atmosphere, the gallery categories: they're there for the visitor deciding whether a print belongs in their home, and every one of them is also written to be found. The placement advice, for example, ranks for searches like "Mediterranean print for living room", which is exactly what someone in buying mode types.

The catch is that none of it can crowd the page. I've seen fine art photography sites stuffed with keywords, and it cheapens the work on display. So the rule became: the search-facing language lives where search engines read it (titles, subtitles, alt text, schema), and the prose stays in my voice. In the gallery, for example, the heading is my line "What Stays" at display size, with "Photography Wall Art by Markus H. Green" above it as a small kicker.

The site was first built on my vocabulary as a photographer: technique words like "intentional camera movement" and "long exposure", and places like Berlin, Porto and Tenerife. In September I checked in Ahrefs whether anyone actually searches that way.

"Wall art" gets 68,000 searches a month in the US. "Fine art print" gets 600.

Every technique term came back with no measurable volume. Place turned out to be a trap for abstract work: you can't recognize the location in the image, so nobody finds it that way. "Berlin wall art", the biggest place term, mostly brings people looking for art on the Berlin Wall.

So I rewrote the search-facing layer of all 22 prints in the buyer's words (style, subject, color, room, size) and left the prose alone. "Giclée print on Hahnemühle" still describes the object. The goal was to be found by the words buyers use without the site starting to sound like a poster shop.

After the head term itself, room is the biggest way people search for art (36,000 searches a month), and the site had no page for it. So I wrote guides for the bedroom, the living room, neutral and minimalist wall art. Each one answers the questions you actually have in front of an empty wall: how wide the print should be relative to the sofa, how high to hang it, horizontal or vertical, one print or a pair. Every print in a guide is shown in the matching room and links to its own page.

They're long, around 1,600 words each, because the results for these searches belong to retailers, and the pages that break through are editorial guides. The rule was that every section answers a real buyer question from a photographer's position, without filler about why art matters.

Earlier guides on paper, dibond, acrylic and framing grew faster than anything else on the site, but they mostly brought in people printing their own photos rather than people buying art. They stay for buyers who need them, and the room guides are aimed at people who are shopping.

Not every good photograph deserves a full print page with a story, mockups and a price. Some work better as part of a sequence. Traces is where those go: place-based photo essays about the Auditorio de Tenerife, Schauinsland above Freiburg and the mountain villages behind Nice, with every photograph still available as a print.

Places are searchable in a way abstract prints aren't, and for most of the summer Traces pages were the only content on the site that ranked. When I expanded them, the rule was my own memories and facts I could verify, nothing invented to fill space.

The site also has essays on how I think about photography: why I make images rather than take them, and why a long exposure can still be true to a place. I write them by talking. AI plays interviewer: it asks me questions about an idea, I answer in voice notes, and it helps turn the transcripts into a structure that I then edit into an article (which I think is one of the greatest gifts of working with LLMs).

The ideas and the voice are mine. The AI asks the questions and keeps the structure, which suits how I think far better than sitting down in front of an empty page.

Buying used to take three steps: find the price on a separate page, open an empty contact form, then describe the print, size and material in your own words. Each step was a place to give up. Now every print page has size and material options with a live price, and "Inquire about this print" opens the form already filled in. A "Save for later" list lets someone ask about several prints at once.

I don't want to run a full online store, and I prefer handling every inquiry myself. Stripe payment links come next only if buyers start asking to pay directly. A Shopify button was ruled out: I had a Shopify store once, and it's a shop with a fee.

For conversations about more than the 22 public prints, I can quickly put together a private client gallery: an unlisted page with a wider selection for one buyer, filterable by place and orientation, feeding the same saved list. Nothing WordPress couldn't do, but setting up a new one is quick.

Every print needs a description, artwork details, placement advice, alt text, an SEO-friendly filename and consistent metadata. Doing that by hand meant I kept putting uploads off, so I built a small Python app with Claude:

The drafts were fluent but samey. One placement phrase ended up on 12 of the 22 prints, and a buyer reading three print pages in a row would notice. I rewrote the descriptions so no four-word phrase appears on more than two prints, and banned the clichés in the app's prompt so new prints don't bring them back. The interface isn't pretty, but it doesn't need to be: it has exactly one user.

I'd wanted to rework my photography site for years. I've had several over time, and this is the first one where the process actually works for me. Being both the photographer and the person who knows how to build the site is quite the fortunate combination: I know what the work needs, and I can try things out myself. So I'm treating it like a portfolio, but also like a product, with experiments to run and changes to measure.

Competing for print searches is a long game, which is why I'm going after the longer, more specific searches. Language models have taken a lot of search traffic, but people still want to look at art before they hang it on their wall, so being found, and suggested by LLMs, still matters. What comes next is mostly more prints and more writing, chosen partly by what I want to make and partly by what people are searching for.