Less Chat In ChatGPT? How AI Could Evolve In Ad Tech In 2025

Good news for skeptics: Hopefully, 2025 will be the year of more practical AI use cases for advertising, especially as agencies get more hands-on with the technology.

The more access AI gets to agency workers, the more it will integrate into the advertising industry, said Nicole Perrin, analyst and vice president of business intelligence at Advertiser Perceptions.

Meanwhile, the number of advertisers using AI tools has already increased from 56% to 68% by 2023, according to a survey conducted by Advertiser Perceptions in October.

And the number of advertisers who mostly or completely trust AI-powered ad technology to make campaign decisions without human oversight has also increased, from one in four to nearly half.

Goodbye chatbots?

But advertisers need to better understand what AI can really do before they can make the real leap—which means looking beyond the hyped-up products out there right now.

Chatbots like ChatGPT and Microsoft Copilot and programs like Dall-E for images and Sora for video are good ways for advertisers to get their feet wet, said Amol Waishampayan, chief product officer at ad planning and activation platform fullthrottle.ai.

Similarly, most direct consumer interactions with AI have also taken place in this area, whether through communication with a company chatbot (sometimes with disastrous results, e.g. Air Canada learned earlier this year) or asked for a tool to write an email for them.

But it’s long past time for advertisers to move on and embrace more sophisticated and pragmatic uses of AI, Waishampayan added.

“I’d like to get rid of what I think is very superficial value,” he said, moving toward a stronger integration of AI into media buying, planning and measurement technologies.

In the meantime, however, beyond AI-generated text for e-mail subject lines or calls-to-action, for example, there seems little desire to feed bots anything other than bottom-funnel messaging.

Human-generated content in any form “has more authenticity,” said Akaash Ramakrishnan, co-founder of creative optimization platform AdSkate. He even went so far as to suggest that the desire for this kind of authenticity may cause large language models to be “phased out” from some common textual use cases somewhere down the line.

Alex Collmer, CEO and founder of VidMob, put it more bluntly: “Why should I bother to read something you didn’t bother to write?”

Be specific (but not too specific)

Chatbots aside, artificial intelligence has been part of the ad tech world for over a decade.

Unfortunately, however, there is a tendency to lump all automated tools together into one big bucket.

Still, it’s important to distinguish between “AI” and “ML,” or machine learning, said Wil Schobeiri, CTO of personas company Ogury.

Machine learning is the ability of a computer to identify unsupervised patterns on a large scale or with a computational intensity that humans could not achieve alone, he said, while AI is simply a “marketing term”.

Not that there aren’t practical business cases emerging for both machine learning and generative technologies.

Ad tech companies have primarily used natural language processing capabilities as a way to generate strategic recommendations based on internal company data or audience personas.

And according to Advertiser Perceptions, advertisers are already using AI to more effectively target audiences (58%) and personalize ads (49%). 52 percent use artificial intelligence to inform marketing strategic decisions, a significant increase from 2023.

AI also has the greatest potential to boost creative output, helping with click-through optimization and alleviating the “undifferentiated heavy lifting,” Schobeiri said, meaning essential, high-volume tasks that don’t require human intervention — in other words, busy work.

Even in these more specific use cases, however, advertisers should be careful not to overdo it with AI.

For example, it doesn’t make sense to spend a lot of time, money and computing power on customizing individual consumer experiences, which would lead to “huge sustainability disadvantages,” Collmer said.

“We’re not really that different from each other,” he said. “There will be the right level of personalization to get you the results you need as a marketer.”

Not all that is generative is gold

Meanwhile, the ad tech world will also need to take a more serious look at the current limitations of AI – and the potential for malicious actors to take advantage of it.

For example, Schobeiri worries about what generative technologies, specifically LLM, will do to worsen the proliferation of websites built for advertising.

“Advertisers and DSPs in the supply chain are going to start trying even harder to identify whether the underlying content is high value,” he said, referring to human-generated content that’s not just designed to attract programmatic ad spend.

Another lingering problem will be generative AI’s penchant for “hallucinations,” a colloquial (and technically imprecise) term for providing apparently false or fabricated information.

“There are some potential pitfalls for people using and trusting AI in everything,” Perrin said, advocating a “trust but verify” approach.

Similarly, many sources told AdExchanger that they hope 2025 will be the year AI professionals and companies work more closely with regulators, as well as find better ways to self-regulate as an industry.

More independent third-party sources would be especially helpful. Today, “most agencies get their AI training directly from the companies that sell them on AI,” Perrin said.

“The question is who, if anyone, will emerge as a more credible, neutral source of AI information for our industry,” she said. “Until that happens, a lot of the conversation will end up being pretty vendor-driven.”