You have 500 products in your store and a designer who works for you two days a week. Black Friday is coming and you want ads with the crossed-out price across the whole catalog, in square, vertical and story formats. The math is simple: 500 products times 3 formats makes 1,500 images. And at the first price change, you start over.
That is how stores with a large catalog end up with one of two options. They put the raw product photo in the ad, or they run the same few creatives across the whole catalog. AI-generated images are a third option, if you know what to ask for and what to check.
Why the picture matters in a catalog ad
In catalog ads, the platform chooses the product. You mostly control the image and the text. The image decides whether the person stops scrolling or moves on. A photo on a white background is right for Google Shopping. On Facebook, Instagram or TikTok, though, it competes with photos of people, vacations and food. A product shown in context says faster what it is for: the mug on the breakfast table, the jacket on an autumn walk.
There is no rule like "the photo with context always works better". It depends on the product, the audience and the platform. The only sure way is to test two or three versions on the same product and compare.
What AI does well
- It changes the background. You start from the cut-out photo of the product and put it into a suitable scene, without a photo shoot.
- It adapts formats. The same image in square, vertical and story, without cropping by hand.
- It works in series. The same composition across hundreds of products, with your colors and your brand.
- It adds graphic elements. Badges, price, a discount banner, in your store's style.
Where it goes off the rails
Generative AI makes mistakes in predictable ways. In a product ad, these mistakes cost you:
- It changes the product. The shape gets distorted, the text on the label can no longer be read, the logo looks different. The customer must receive exactly what they saw in the ad.
- It breaks proportions. A perfume bottle the size of a water bottle, a ring the size of a plate.
- It invents strange details. Hands with six fingers, impossible reflections, objects melting into the background.
- It changes style from one picture to the next. 200 products in 200 styles do not look like a brand.
- It promises what does not exist. Sizes, colors or accessories the product does not have.
The practical conclusion: the product stays the real photo, and the AI generates around it, not in its place. And you check every image before you publish it.
The process, step by step
- Start from clean photos. The result is no better than the photo you start from: a uniform background, good light, the whole product in the frame.
- Write the style once. Colors, type of scene, tone, where the text goes. One page of rules, used across the whole catalog.
- Start with ten products. From different categories, not the whole catalog. Only if it works on ten do you expand.
- Check every image. You compare it with the real product: color, shape, text, proportions. A short checklist spares you many complaints and returns.
- Test. The version with context against the plain photo, with the same budget and in the same period. You keep what wins.
- Change the creatives from time to time. When the same people see the same image too many times, the ad wears out. Prepare new versions for the products you spend the most on.
Tip! Choose the first ten products from among those that already sell. You have data to compare from one day to the next, and if the new image does not help, you see it quickly. A product that does not sell even with the old photo tells you nothing about the new photo, however beautiful it is.
The platform rules you do not skip
- Google Shopping: the main image in the feed may not have promotional text, an overlaid logo or a watermark. Creatives with context or text are for other ad formats, not for the image in Merchant Center.
- Meta and TikTok: each has rules for text in the image, for what you claim about the product and for AI-generated content. Read the current policy before you launch a large series. The rules change, and some ask you to label content created or modified with AI.
- Rights to the photos: you use only photos you have the right to use, taken by you or received from the supplier with permission to use them. You do not ask the generator to copy a real person or someone else's brand.
- Consumer protection: the ad may not mislead. If the image does not show the real product, you risk complaints to ANPC and returns.
How much time you actually save
Honestly: it depends. A generator spares you the repetitive work with formats and backgrounds. It does not spare you the person who chooses what gets published. Choosing the style, checking and testing remain. For a store without a designer, the difference is between having creatives across the whole catalog and having none at all. For one with a designer, it is time gained for the things that require judgment.
How Aopy handles creatives
Aopy is built by the team behind Clinica de Marketing, which has worked in e-commerce since 2016.
Aopy has two tools that work on the catalog in Feed Manager. Feed Design puts a template over the photos in the feed (price, a discount badge, a frame in your brand colors) on a set of products or on the whole catalog at once, not photo by photo. Creative Studio generates ad images from the product photos in the catalog, based on a description you write.
Both start from the same catalog, so you do not upload the products a second time. You check the results before you use them, as described above, and ad approval remains the platform's decision (Google, Meta or TikTok). Creative AI is bought as a one-time package. The packages and prices are on the pricing page.