AI product photography has gone from gimmick to genuine tool in the space of two years. Sellers who used to pay a studio £150 per SKU can now generate dozens of variations for a few quid each. But the marketplace rules have moved fast too, and the answer to "should I use AI or shoot real photos?" is no longer a simple one.
This guide breaks down where AI photography wins, where it still falls short, and the marketplace policies you need to know before you upload a single AI-generated image in 2026.
How AI product photography actually works in 2026
Modern AI product photo tools fall into three rough categories:
- Background generators — you upload a clean cut-out of your product and the tool drops it into a generated scene (kitchen counter, beach, marble surface). Tools like Photoroom, Pebblely and Claid sit here.
- On-model and lifestyle generators — flat lay clothing photos turned into images of an AI model wearing the product. This is what ASOS used in its 2025 pilot.
- Full-scene generators — these create the entire image from a text prompt and a reference photo, useful for hero banners and ad creatives but riskier for main product listings.
The technology has improved enormously. Shadows match, reflections behave plausibly, and texture rendering on hard goods like ceramics, leather and metal has reached a point where casual buyers genuinely cannot tell. Soft goods, transparent items and food are still tricky.
The cost gap is real
Industry benchmarks for 2026 put traditional studio photography at roughly £70 to £200 per SKU once you include the photographer, retoucher and styling. AI alternatives sit at around £2 to £10 per image and can produce 20 variations in the time a studio shoot takes to set up a single angle.
For a seller listing 500 SKUs across multiple marketplaces, that maths becomes hard to ignore. A studio bill of £50,000+ versus £2,500 for AI is the kind of difference that decides whether a small brand can compete at all.
Does AI imagery actually convert?
The data so far is genuinely promising, with one big caveat: most published case studies come from large brands with strong existing traffic.
- ASOS reported a 340% lift in product page conversion after rolling out AI-generated on-model imagery in 2025.
- AI lifestyle imagery in fashion has shown around 60% conversion improvements in controlled tests.
- 67% of top e-commerce operators now have a dedicated budget line for AI imaging tools.
The underlying truth has not changed though: professional-quality images convert 30 to 40% better than amateur ones, regardless of whether they came from a studio or a server. AI does not magically save bad inputs. A blurry, badly cropped source photo will produce a polished but still wrong AI output.
What Amazon, Shopify and other marketplaces allow
This is where many sellers trip up. Each platform has taken a slightly different stance.
| Platform | AI images allowed? | Disclosure required? | Key rule |
|---|---|---|---|
| Amazon | Yes, with restrictions | Yes, for substantially AI-generated content | Image must accurately represent the physical product shipped |
| Shopify | Yes, no restrictions | Not required | You control the storefront, so consistency matters most |
| Shopee / Lazada | Yes | Not required, but misrepresentation is penalised | Main listing image still needs a clean white or neutral background |
| TikTok Shop | Yes | Encouraged for fully synthetic content | Hero images must show the actual product |
| EU markets (any platform) | Yes | Yes, under the AI Act from August 2026 | Provenance metadata (C2PA) increasingly expected |
Amazon's 2026 update is the strict one
Amazon updated its seller guidelines in 2026 to require explicit disclosure when product images have been substantially generated or modified by AI. The disclosure goes in the product description, prominently, with wording like: "This product image was created using AI technology to demonstrate the item's appearance."
What Amazon permits without disclosure:
- Background removal
- Colour correction
- Lighting adjustments
- AI-powered resizing
What Amazon prohibits outright:
- Lifestyle images that misrepresent product scale
- Generating product features (extra pockets, fake ports, decorative details) that do not exist on the physical item
- Fabricated before-and-after comparisons
- AI-generated review imagery
Penalties are real: listing suppression, reduced search visibility and account warnings for repeat offenders.
Where AI wins and where studio still wins
Use AI for
- Lifestyle and scene variations — same product, ten different room or outdoor settings
- Seasonal refreshes — Christmas, Chinese New Year, summer scenes without re-shooting
- Ad creative — Meta and TikTok ads where you need many variants quickly
- Marketplace-specific aspect ratios — square for Shopee, 4:5 for Amazon, 9:16 for TikTok
- Colour and material variants on products you have already photographed once
Stick with real photography for
- Your main hero / listing image — buyer trust is highest with a real photo on white
- Soft, drapey fabrics and complex textures — silk, lace, knitwear still confuse AI
- Food and beverages — AI struggles with realistic edible textures
- Anything with text, logos or fine detail on the product itself
- High-return categories like apparel where any visual misrepresentation fuels returns
The hybrid workflow most sellers should adopt
The best approach in 2026 is rarely "AI or studio." It is "shoot once, multiply with AI."
- Photograph the product properly once. Clean white background, multiple angles, accurate colour. This is your truth source.
- Use AI for variations. Drop the cut-out into lifestyle scenes, seasonal backgrounds, marketplace aspect ratios, ad creatives.
- Resize and compress for each platform. A 4000x4000 master image needs different output sizes for Amazon (2000px), Shopify (2048px), Shopee (1024px) and Lazada (1000px). Tools like PixelPrep handle this part of the workflow in batches, so you spend your time on creative and not on file conversions.
- Audit your AI outputs. Walk through every generated image and ask: does this represent what I will actually ship? If the answer is "not quite," redo or fall back to the original photo.
- Disclose where required. Add the Amazon disclosure language for any substantially AI-generated images. For EU sales, ensure your tool supports C2PA provenance metadata.
The 2026 AI photography checklist
Before you publish a single AI image, run through this:
- Does the AI image accurately represent the physical product, including colour, scale and features?
- Have I added platform-required disclosures (Amazon, EU)?
- Is the main / hero image a real photo, with AI used only for supporting shots?
- Does the image meet each marketplace's resolution and background rules?
- Have I checked for AI artefacts: warped hands on models, unreadable text, melted product edges?
- Does the file size sit under the marketplace cap (Shopify 20MB, Amazon 10MB, Shopee 2MB)?
- Have I batch-resized the image into every aspect ratio I need?
The bottom line
AI product photography in 2026 is not a replacement for studio work — it is a multiplier. Sellers who shoot one strong real photo and use AI to scale it across platforms, formats and seasons are getting the best of both worlds: studio-grade trust on the main image, plus the cost and speed advantages of AI on everything else. The brands losing out are the ones doing pure AI without the rules in mind, and the ones still paying for re-shoots when a generated variation would do.
Get the foundations right, respect the marketplace rules, and let PixelPrep handle the volume.