AI product photos for Amazon, Flipkart & Meesho: the compliance-safe workflow for Indian sellers (2026)

Somewhere between "AI will shoot your entire catalogue for free" and "AI images will get your listing banned" sits the boring, profitable truth: marketplaces don't care how your image was made. They care whether it's honest and meets their specs. Get those two things right and AI imagery is not just allowed — it's how a growing share of top listings are already made.

Here's the workflow that keeps you on the right side of the rules, platform by platform, plus the one mistake that costs sellers more than any takedown.

What the platforms actually require

None of the big three Indian marketplaces prohibits AI-generated imagery. What they enforce, and always have:

• Amazon — the main image must show the actual product on a pure white background (RGB 255,255,255), filling most of the frame, with no props, text, watermarks or logos that aren't on the product. Images must accurately represent what ships. Secondary images are where lifestyle scenes live — AI or otherwise.
• Flipkart — clean, well-lit images meeting category quality guidelines; the product shown must match the product delivered. Misleading imagery is grounds for listing action, whatever made it.
• Meesho — similar truthful-representation standards, with heavy enforcement pressure coming from the returns side (more on that below).

Notice what's common: every rule is about truth and specs, not technique. A dishonest studio photograph breaks them; an honest AI composite doesn't.

The rule that actually protects you

Tattoo this on the workflow: real product pixels, generated everything else.

The safe method is not "type a description of your kurta and generate a photo of it." It's:

1. Photograph your actual product, clean and well-lit. Phone + daylight + plain background is enough as raw material.
2. Cut the product out and keep those pixels untouched.
3. Use AI to build the world around them — backgrounds, surfaces, lighting environments, seasonal contexts.
4. Match the physics: shadow direction, reflections, colour temperature. This is the craft step that separates "premium listing" from "obviously pasted."

Break the rule — let AI redraw the product itself — and you get a photo of something you don't sell. The fabric weave is wrong, the shade is off, the stitching is invented. The customer photographs what arrived, files "item not as described," and now you're paying return logistics on every order while your seller metrics sink. On Meesho, where margins are thinnest and return rates already brutal, imagery-driven returns quietly kill more sellers than takedowns ever will.

The takedown is the cheap failure. The returns loop is the expensive one.

The catalogue workflow, start to finish

• Main images: shoot the real product on white (or shoot on anything neutral and do a clean white-background extraction). AI's role here is minimal by design — Amazon's main-image spec basically is a traditional pack shot.
• Secondary/lifestyle images: this is where AI earns its keep. The same base cutout goes into six contexts — marble counter, festive setting, in-hand scale shot — for a fraction of one styled-shoot invoice. (Full cost comparison here.)
• Infographic frames: AI-generated backgrounds + real product + honest feature callouts. Keep claims on the right side of what the product does; imagery rules and advertising rules both apply.
• A+ / brand content: generated scenes are fine; anything presenting as a real customer or real testimonial must be real. That line is hardening into regulation — India's draft AI-labelling rules for ads explicitly target fabricated "real people."
• Category caution: jewellery, fine fabrics and food close-ups sell on texture. Shoot real macro heroes for those; generate the context shots. AI still under-delivers exactly where texture is the purchase decision.

Quality bar: what "done" looks like

Run every finished image past three checks:

1. The squint test — shadows fall one way, product sits in the scene, nothing floats.
2. The honesty test — would you be comfortable if the customer received exactly what this image promises? If any product attribute got prettier in the composite, redo it.
3. The zoom test — marketplaces serve zoomed views; artefacts invisible at thumbnail size are glaring at 2x. Check edges, reflections, text on packaging.

Fail any of the three and the image costs you more than it saves. An honest phone photo outsells a sloppy composite — buyers punish "off" faster than they punish "plain."

Frequently asked questions

Are AI-generated product images allowed on Amazon India?

Yes. Amazon's image rules govern accuracy and presentation (white main image, product as shipped), not production method. AI-assisted secondary and lifestyle images are widespread. The risk is misrepresentation, not the tool.

Will customers know my photos are AI-enhanced?

If the compositing is competent and the product pixels are real — no, because there's nothing false to detect. Backgrounds and sets have been artificial since the first studio cyclorama; AI just made them cheaper.

What's the cheapest way to do this for a big catalogue?

One disciplined base-shoot of every SKU, then batch-generate contexts. Sellers run 200-SKU catalogues on a few thousand rupees a month this way — the per-image math is here. Video versions of the same listings: the AI video workflow.

Imagery is one lever; AI runs deeper across an online store — pricing, support, recommendations.