Insights

content

A photo shoot without a studio. AI content in e-commerce: what works and what damages your brand

Sep 22, 2026 · 9 min read · Human

[post image]

We built a clothing brand whose entire product photo shoot was created with AI. Here’s how it turned out, where the limits are, and how to keep your brand from turning into generic mush.

In 2023, we wrote on this blog that AI won’t replace the human eye in expert content. We stand by that. But there’s one area of e-commerce where AI has changed the economics of an entire category of work, and it would be dishonest to pretend otherwise. We’re talking about product content: photos, video, color variants, social media assets.

A photo shoot for a collection of thirty T-shirts used to mean a day in the studio, models, a stylist, a photographer, retouching, and then a second day for the “lifestyle” versions. For a small brand, that cost decided whether the collection would happen at all. For SUAP.M, a clothing brand for dog owners, we created the entire digital photo shoot with AI: models, dogs, styling, backgrounds, campaigns. A catalog of more than thirty products, plus assets for the store, Instagram and ads, in a time frame and on a budget a traditional shoot could never match.

What works

  • Product photos on a plain background and packshots in many color variants from a single master. Here AI is simply better: faster, cheaper, more consistent.
  • Lifestyle scenes where the product is the hero and the person or animal is the backdrop. A dog bed in a living room, a T-shirt on the street, an accessory on a table. As long as the product in the photo is that product, not an approximation of it.
  • Product video and short social formats: showing the cover, the fastening, the texture, close-ups. Formats that used to get skipped because they weren’t worth shooting.
  • Making-of material that becomes content in its own right: people are curious about how it was made and happy to watch.
  • Localizing for different markets: the same scene with different surroundings, a different season, a different type of interior for a different country.

What damages your brand

Genericness. Models who look like every model from every other brand. Interiors that are “pretty” the way everything generated without an idea is pretty. Photos where the product has a different color, a different seam, a different texture than in reality, because the tool “improved” it its own way. The customer will notice when unboxing, and then write about it in a review.

The second problem is the lack of a guiding hand. AI content without a creative director is like catering without a chef: everything is edible and nothing tastes good. At SUAP.M, we started with a brand book, a palette, the type of person and dog, the character of the scenes. Only then did we generate anything. Half of the material went in the bin, because it wasn’t the brand, just a picture.

The rules we stick to

  1. The product must be accurate. If the tool alters the product, we don’t publish. For products with complex textures, we combine a real photo with a generated setting.
  2. The style is written down before the first image is created. A brand book, references, a list of things we don’t do.
  3. A human approves every asset. AI proposes, a person decides. It’s the same rule we apply to product descriptions and campaigns.
  4. We’re honest with customers. We don’t pretend the photos come from a real shoot if someone asks. Most people don’t ask, because what matters to them is the product, not the process.
  5. We measure what we’d measure for any other content: product page conversion, social media engagement, campaign results. If AI assets sell worse than real ones, we go back to the studio.

What we want to achieve and how we’ll know

At SUAP.M, we measure AI content the same way as any other content. Launch goals:

  • A full catalog of more than 30 products with a complete set of assets (packshot, lifestyle, video) ready within 2 days of brand book approval, instead of 4 weeks with a traditional shoot.
  • Content cost per product no higher than 60% of a studio shoot for the same collection.
  • Product page conversion of at least 1.8%, comparable to pages with real photos at similar clothing brands.
  • Returns because the product didn’t match the photo below 2% of orders. This is the most important measure of how honest the assets are.
  • Cost per click in Meta campaigns no higher than PLN 1.50.

Results after the first quarter: time to launch 2 days, content cost 60% of a studio shoot, product page conversion 1.8%, photo-related returns below 2%, CPC PLN 1.50.

What this changes in planning

The biggest change is in the order of things. Content used to come last because it was expensive, and many brands launched with five photos, promising themselves they’d add the rest once they started making money. Now content can be complete from day one, which means the store, the campaign and social media can launch together. For an experiment meant to test a new brand in a few weeks, that’s the difference between possible and impossible.

Running an online store?Let’s start with a 30-minute conversation about your catalog, campaigns and margin.Explore AI eCommerce

Contact

Want to earn your audience’s trust?