TikTok Shop permits AI-generated images on product detail pages, and it enforces against them aggressively when they mislead. The line is not whether AI was used. It is whether the image shows the buyer something other than what will arrive in the box.
Malaysian sellers have adopted AI product imagery faster than they have read the rules governing it, and listings are being removed for faults the seller did not know were faults. Imagery sits alongside listings, content and fulfilment in how TikTok Shop works.
The four rules everything else follows from
Truthful and consistent. The image must match the actual product, SKU, title, attributes, packaging list and specifications. You cannot render a 2D item as a 3D object, turn plastic into metal, leather or glass, or present a single item as a set.
No exaggeration. AI must not inflate size, material, quantity, function, effect, durability or usage scenario. Extreme before-and-after results, one-spray-cleans-everything claims, a small product covering a large space – all prohibited.
Verifiable claims. Selling points around size, material, quantity, function and effect must be supported by the real product, its specifications or test evidence. You cannot create a claim through an image that you could not make in words.
Disclosure. Where AI-generated scenario images are used, say so. The platform’s suggested wording is straightforward: *”AI-generated scenario images are for display only. The actual product and specifications shall prevail.”*
What gets listings pulled
The enforcement patterns worth knowing, because they are specific.
*Scale.* Using a person, furniture, a door or a vehicle to make a product look bigger than it is. A 10cm toy car photographed at apparent life size is a violation. The compliant version shows the product against a ruler.
*Material.* Plastic rendered as solid wood. PU rendered as genuine leather. Thin fabric rendered as thick wool. Alloy or imitation stones shown as real gold or real diamonds.
*Function.* Effects the product does not have – glowing, rotating, projecting, self-repairing, floating. Electronics implying cameras, ports, battery capacity, waterproofing or 5G they do not possess.
*Simulation products.* Artificial flowers shown as fresh flowers. Food models shown as edible. Plush pets shown as real animals. The title, category and image must all make clear the item is artificial, decorative, a model or a toy.
*False trust signals.* This is the one that carries the most risk. AI-generated doctors, pharmacists or dermatologists recommending a product. Fabricated certifications, lab results, awards, media coverage, expert endorsements or buyer reviews. Health and beauty sellers are the most exposed and the most frequently caught.
Categories under stricter scrutiny
Health and beauty, food, mother and baby, personal care, supplements, electronics and home improvement are held to a higher standard because the consequences of misleading a buyer are worse.
In health and beauty specifically: no AI-edited faces, skin, hair, teeth or body shape used as proof of efficacy, no implied medical treatment, and no AI professionals endorsing anything. What you may show is packaging, ingredient information, usage steps and honest scenarios.
In food: quantity, weight and packaging must match the SKU. Do not enlarge produce, do not generate food that looks better than what ships, and do not use AI to conceal damage or defects.
What happens if you get it wrong
Enforcement ranges from product delisting or deactivation through product freezing, formal warnings and account health deductions. Repeat violations escalate. A product that keeps attracting the same finding can end up with reduced visibility for progressively longer periods.
If you are hit, the required remediation is clear: replace or remove the misleading images, upload real product photographs covering the main image, details, size, material and packaging, correct the title and specifications, add the disclosure for any remaining AI scenario images, and either evidence or remove high-risk claims.
One thing to get right before anything else – do not delete the content if you intend to appeal it properly. Once deleted, the violation cannot be contested.
A checklist before you publish
Does the product in the image match what ships? Has AI altered size, colour, material, thickness, quantity or accessories? Is at least one real photograph included? Does each SKU image match its SKU? Does the packaging list match what is shown? Are function and effect claims evidenced? Are simulation or decorative items clearly labelled? Is AI scenario imagery disclosed?
Eight questions. Most removed listings fail at least two of them.
Common questions
Can I use AI backgrounds on my main product photo?
Yes, provided the background does not alter the product’s appearance, scale, material, colour or quantity.
Is AI retouching allowed?
Within limits. Cleaning up lighting is fine. Changing what the product looks like is not.
Do I need the disclosure on every image?
On the product details where AI-generated scenario images are used. It is cheap insurance.
We run social commerce management for Malaysian brands, including listing compliance. Tell us the brand and the goal.