Amazon Rolls Out New Image Generation Tools for Seller Ads
Will this drive higher click-through rates? Amazon says yes.
Amazon has introduced a new image generation tool for advertisers that sell on its platform. The examples enable an advertiser to pull in product images from their Amazon.com listings and then generate backgrounds that place their images in everyday settings. Amazon said in its product announcement today:
In a March 2023 survey, Amazon found that among advertisers who were unable to build successful campaigns, nearly 75% cited building ad creatives and choosing a creative format as their biggest challenges.
“Producing engaging and differentiated creatives can increase cost and often requires introducing additional expertise into the advertising process,” said Colleen Aubrey, senior vice president of Amazon Ads Products and Technology. “At Amazon Ads, we are always thinking about ways we can reduce friction for our advertisers...”
To that end, Amazon Ads has launched image generation in beta—a generative AI solution designed to remove creative barriers and enable brands to produce lifestyle imagery that helps improve their ads’ performance. For example, an advertiser may have standalone images of their product against a white background, like a toaster. When that same toaster is placed in a lifestyle context—on a kitchen counter, next to a croissant—in a mobile Sponsored Brands ad, click-through rates can be 40% higher compared to ads with standard product images.
Practical Applications of Generative AI
Andy Jassy, Amazon’s CEO, outlined the company’s three-layer generative AI strategy in an earnings call earlier this year. The compute layer relies on AWS infrastructure along with NVIDIA GPUs and Amazon’s home-grown AI-optimized ASICS for training and inference jobs. Foundation models, such as large language models and image generation models, are in the middle layer. You can think of this as the capability layer, services layer, or the engine for generative AI-enabled applications.
The top layer is the application layer. This is where the new Amazon Ads image generation originated. During Amazon’s second quarter 2023 earnings call, Jassy commented:
Inside Amazon, every one of our teams is working on building generative AI applications that reinvent and enhance their customers' experience.
Shooting for Small Wins
Adding image backgrounds is not an earth-shattering invention. Indeed, Meta announced the same feature in May, along with text variation, outcropping, and an advertiser AI sandbox. Amazon is playing catch-up with many of its tech giant peers. However, the feature does show tangible progress by Amazon in embracing generative AI features. It also appears to validate Jassy’s statement that every team is working with generative AI.
Last month, Amazon introduced generative AI tools for writing ad copy for seller listings. Voicebot.ai wrote at the time:
Amazon is rolling out a new generative AI feature for third-party sellers to compose better product listings on its marketplace. The tools leverage large language models to turn a brief prompt defining the item into a full post, including title, description. and bullet points summarizing its details.
Amazon frames the new feature as a way of streamlining the process of creating and posting items for sale on its website, kickstarting the process for those who find the blank text box overwhelming…Using the AI text generator should shorten the time and energy typically required for sellers to manually put together high-quality listings.
Background image generation and seller ad listing copy also illustrate a pragmatic approach to generative AI. Add the technology where it can provide immediate value to users. A 40% higher click-through rate for products placed in a “lifestyle context” offers practical value to Amazon advertisers and should help drive more sales, which would make everyone happy.
While other companies are focused on making a big splash with flashy products, Amazon and Meta seem to be content with small wins that they hope will add up to big value.