Meta’s AI ad update—what brands need to know about image and headline automation

Meta released a new generative AI tool for advertisers but it’s still working on some details around how to label AI ads and limiting some of its outputs so advertisers maintain some control over campaigns. The latest fledgling AI ad products exemplify how the industry is working quickly to incorporate automation while at the same time taking precautions to navigate the thorny issues related to generative AI. Last year, Meta started exploring generative AI in its ads platform, giving the technology the ability to write copy and change backgrounds on images. On Tuesday, Meta, which owns Facebook, Instagram and WhatsApp, expanded AI features inside its ad platform, allowing brands to generate full images based on their existing creative assets. The tool also generates headlines as well as ad copy. The latest iteration of the program includes full-image ads and runs on Llama 3, Meta’s latest large-language model. There are some safeguards in the image-generation tool—it won’t create people or products that haven’t been uploaded to the system by a brand. “We focused on being able to expand from some of your existing assets and create variations on that,” said Jon Hegeman, Meta’s VP of monetization, speaking at Meta’s AI-themed event on Tuesday. “We felt that this was the area where it would make the most sense.” Check out the latest Ad Age events and award programs Automation—and anxiety Meta, like rivals Google, Microsoft, Amazon and TikTok, has been working AI into its ad platform to automate campaigns, handle bidding and targeting, and create chatbots that marketers use to interact with the ad platform. Generative AI automates the creation of different ad formats, such as resizing ads to run in a vertical orientation. Generative AI can also instantly test different fonts and ad copy, and it switches up the creative depending on what works best for different audiences. However, AI also has caused some anxiety among marketers, and not just because machines could fill roles traditionally done by humans. Brands are concerned about how their data is handled by large language models, which are difficult to audit. Also, if it’s uncontrolled, generative AI could create mistakes that reflect poorly on a brand. The major platforms have been building safeguards to prevent AI mishaps. For instance, last month, Google launched a generative AI tool in its ad product called Demand Gen, and its policies keep the image generator from depicting people and brands’ logos. Here are 5 brands that are cautious about using AI Meta’s AI launch had similar restrictions, although brands can bring their images of people and products into the platform. “As part of the text prompt feature within Image Generation, advertisers are able to create image variations inspired by their original ad creative, this could include people, generic faces, or hands,” a Meta spokesperson said on Tuesday, adding that “we have strong guardrails that are designed to prevent our systems from delivering low quality or inappropriate ad content that violates our Community Standards and Advertising Standard.” Meta showcased the new tools with two brands that have tested them—Casetify, a phone case merchandising company, and AS Beauty. Last year, the beauty brand pushed about 30% of its Meta advertising into AI-based campaigns known as Advantage+, according to Scott Kramer, AS Beauty’s VP of growth. Advantage+ is Meta’s automated campaign that is geared toward shopping goals such as sales and clicks. Now, with generative AI producing more of the creative assets, campaigns can test more varieties of ads. “[We’re] letting the system optimize for the different strengths of the different signals we know are going to get us the results,” Kramer said, adding that “instead of us trying to do a lot of growth hacking and making lots of changes, let’s focus on creative or different strategies, those areas we know we can lean into.” “Creative diversification is the key,” said Alvin Bowles, Meta’s VP, global business group, Americas. ‘Made with AI’ The message from Meta in its push toward AI is that automation will free advertisers to focus on more important strategic matters. “[It] takes some of the mundane tasks away from the clients and the agencies and the marketers,” Bowles said. Brands such as AS Beauty are cognizant of the potential for AI flubs, though, and Kramer noted the importance of using original assets from the brand as the foundation for AI-enhanced campaigns. “From a brand standpoint, at the end of the day, we’re trying to sell products,” Kramer said. “So, we can’t just have AI spin up and show a product we don’t even have.” Meta is still working through some details about how to label AI-generated content in ads, Hegeman said. This month, Meta started rolling out a “made with AI” label required on organic posts that use generative AI. “We’re in the process of working through some of the specifics in terms of exactly how that works in the context of ads,” Hegeman said. There also are data ramifications to working with large language models in major platforms, which brands are still trying to understand. For instance, when they give data to a platform, brands worry their information could help evolve a large language model that competitors also use to run ads. Hegeman said that Meta was being careful to prevent inappropriate outputs with its AI, and that it avoided sharing any copyrighted or unauthorized material. Meta has terms that govern how marketers share data and their rights when using the AI ad platform. “It is important to note that all inputs and outputs are pooled in aggregate along with billions of other training data inputs when we use them to help train models or improve our systems, not to facilitate competitive conquesting,” Meta’s spokesperson said. “The key thing that we’re finding,” Hegeman said, “is separating out the things that advertisers really want close control of, from things where they would actually be happy to have the system automate to drive the best possible performance.” Check out the latest Ad Age events and award programs

Meta’s AI ad update—what brands need to know about image and headline automation

Featured Stories