Phony AI ad tech—how brands can avoid misleading marketing products

A popular meme invaded the internet in 2020 posing a seemingly simple question: Is it cake? These social media videos featured everyday items—from a pair of Crocs to toilet paper rolls. A sharp knife would then enter the frame and carve into it to reveal whether it was cake or the item it appeared to be. Today, marketers are asking a similar question: Is it AI or is it a fake? Just like the viral meme, marketers need a way to separate phony AI services from real ones, and companies that use actual AI from ones that don’t. Brands are being flooded with offers from vendors with AI-based products that promise to improve marketing and work efficiencies. Generative AI helps create images and writes copy, and machine learning makes decisions in marketing plans. But some of these products are not as sophisticated as they seem, according to Sol Rashidi, a data, analytics and AI expert and former chief analytics officer at Estée Lauder. “Specifically, within the marketing functions [at brands],” Rashidi said in an interview with Ad Age. “There is a lot of noise and chatter, and a lot of vendors, software providers and companies going to market saying, ‘We’re doing AI, we’re doing AI.’ And it’s getting very confusing.” Also read: Top 5 AI activations from brands in 2023 In February, the U.S. Federal Trade Commission even issued a warning to companies telling them to “keep your AI claims in check,” so as not to oversell it. “If you think you can get away with baseless claims that your product is AI-enabled, think again,” wrote Michael Atleson, attorney, FTC Division of Advertising Practices. Still, AI has powerful potential in marketing, and it’s not all bunk. Platforms including OpenAI, Google, Meta, Amazon, Microsoft and others have put AI into their advertising products. Sophisticated machine learning is being deployed to manage internet ad campaigns, target ads and measure their outcomes, while generative AI is already being used to create ads. Ad tech providers are touting AI-powered capabilities in their ad servers as a way to find new customers by analyzing patterns and reaching segments of internet users brands might overlook if not for AI’s keen analytical eye. Last month, marketing group MMA Global held workshops with Rashidi and Rex Briggs, the recently appointed chief AI officer at Claritas, to help brands understand when AI is delivering as advertised. MMA Global also recently established the AI Leadership Coalition with more than 100 brands, including Google, Amazon, Meta, McDonald’s, Wendy’s, Target, Mastercard, General Motors and others. It’s not the only industry group grappling with the influx of AI offerings. Earlier this month, Meta announced the AI Alliance, a research association that will work on standards for AI and its implementation in corporate and academic settings. As many as three out of four marketing products billed as AI really aren’t, Rashidi said, and marketing executives often can't tell the good from the bad. “Executives are struggling,” Rashidi said. “Their leadership teams are struggling and the teams reporting into those teams are struggling.” AI learns on its own “Does it really have things like unsupervised learning in it or is it really just statistics dressed up as AI, or rules dressed up as AI,” Briggs said. “Everyone wants to put themselves under the AI banner now.” “Unsupervised learning” means that the machine can process and make sense of large data sets without any human input. In contrast, a less sophisticated method of analysis uses datasets that have been categorized manually, and the machine plays with that data based on pre-set rules written by humans. For example, there are ad servers that use rules dictated by humans to determine when and where to run an internet ad, which is not true AI, Briggs said. If there is a rule that says, “If you are in this city you’ll get this ad,” then “it’s not unsupervised learning,” Briggs said. A lot of ad tech products appear to be AI, but they’re really just very complex Excel spreadsheets, Rashidi said. Marketers have to determine whether it is “a hamster running the wheel; is it an Excel spreadsheet with a lot of macros and [database] … with a massive decision tree. Or is it fundamentally AI, which is code where you set up a bunch of prompts and basic rules, but it learns through experience,” Rashidi said. See our 2023 Marketers of the Year here AI at work Rashidi designed a map that outlines some of the telltale signs of when a product is using true AI. One example illustrates the difference between a camera and “computer vision.” If a program can identify what it sees, then that’s “computer vision,” which is a characteristic of AI. Computer vision is a prominent part of marketing as brands analyze images across the internet and study how their brands and iconography are used on social media sites and in other settings. Rashidi’s AI roadmap also tells marketers to distinguish between AI that can recognize patterns in data and then make decisions based on those patterns—reasoning. True AI uses “natural language processing” to analyze text and “machine learning” to respond in a coherent way, according to Rashidi. “For instance, one of the primary checkpoints revolves around the system’s ability to ‘read’ and ‘see.’” Rashidi wrote in “Is it AI.” “These sensory abilities determine if the system has foundational capacities akin to human senses. If a system can ‘see,’ further questions arise, such as whether it can identify patterns across what it perceives or if it's merely acting as a camera. If a system can ‘read,’ it's vital to determine if it's merely translating or if it's using Natural Language Processing (NLP) to understand and respond to text.” AI marketing blog How brands are using ChatGPT, DALL-E 2 and other tools Read more here Paying the AI price Marketing organizations are eager to try new AI tools, but they also are hesitant, according to Jenny Kelly, Deloitte Digital’s head of content. Deloitte recently surveyed 650 marketers and found only a quarter had deployed generative AI in their strategies, but the companies feel pressure to implement these technologies or else they could lose out to nimbler rivals. “It’s a Blockbuster [moment],” Kelly said, referring to the former video rental chain that collapsed, in part, because it failed to pivot to streaming. Brands are still deciding “‘What do we have to do to still be relevant in the market,’ so if they don’t adopt AI, they won’t be relevant.” In marketing, generative AI is changing fields such as stock imagery, defined by companies such as Adobe and Shutterstock, which have developed new AI image generators. Email marketing is also being affected by text generators, Kelly said. Advertisers worry that machines will be able to outperform people, soon. There will be “game changers” such as a “no-touch ad generation,” Kelly said, where marketers just tell machines what creative they want and the campaign goals. The AI transition can be expensive, Rashidi said, which makes it even more important for companies to do their homework. “So marketers who aren’t technology-based, marketers who aren’t AI-based, who are drinking from a firehose of information,” Rashidi said, “how do we help them, assist them in identifying what is AI versus not AI. And then more importantly, distinguish whether or not they should be paying the premium for it or not paying the premium for it.”

Phony AI ad tech—how brands can avoid misleading marketing products

Featured Stories