As marketers widely adopt large language model AI tools like ChatGPT, Gemini, and Grok to assist with mix media models, built-in bias is leaving radio in danger of being left out of ad budget recommendations, especially in political, awareness, and national campaigns, unless immediate action is taken to make AM/FM visible within AI ecosystems.
Futuri Media analyzed more than 20,000 AI-generated media mix models and found radio receiving minimal – sometimes zero – share of projected campaign spend.
According to Futuri’s findings, the most widely used large language models are prioritizing channels like YouTube, connected TV, and programmatic display advertising, while sidelining traditional media formats such as radio and broadcast TV. Among the eight LLMs tested, Anthropic’s Claude and Google’s Gemini excluded radio in 100% of their generated media plans.
Even more concerning, radio’s average share of AI-recommended political ad spend was just 3%, compared to 18% for CTV/OTT and double-digit percentages for programmatic display.
The core issue, according to the report, is that AI systems only recommend channels they have been trained to recognize as effective, measurable, and relevant. While digital platforms supply massive quantities of structured data like engagement metrics, impressions, and click-through rates, radio’s performance data is sparse, inconsistently formatted, or entirely absent from the sources AI tools use to learn.
Compound bias is already reshaping how campaigns are planned. As AI platforms guide advertisers toward digital video and programmatic, those channels continue to generate more performance data, reinforcing the AI’s belief that they are superior. Meanwhile, radio loses relevance in the machine’s model of the media world.
Futuri emphasizes that the window to reverse course is short and closing fast – around two years.
To combat this, suggested solutions include publishing case studies and success metrics in AI-ingestible formats, pushing radio performance content into high-authority business publications, and feeding results into the databases used by major MMM vendors.
The full Forensic Analysis on Radio and TV Revenue Loss in 2025 report is available
via Futuri.
“The core issue, according to the report, is that AI systems only recommend channels they have been trained to recognize as effective, measurable, and relevant.”
AI likes things that it can measure in real time. Oh, BTW, clients do too. OTA radio can’t do that, and neither could the newspapers.
Newspapers over the last 20 years either collapsed or turned themselves into digital marketing hubs. Radio overall has been slow to adapt to the digital landscape, and now it’s really beginning to show how far behind the industry is compared to it’s digital competition.
If the machines don’t like your station, it’s best get out from behind your desk and go solidify your relationship with your clients by doing a face-to-face visit. Zoom visits don’t count. People do intangibles, machines can’t. Lunch, thank you cards, ballgame tickets, etc…all the things you already know how to do.
The big advantage we have over the machines is that we can build business and personal relationships with our customers. Let’s go do that.
Pretty simple. The LLM are mainly analyzing the recession of radio listening and revenue over the last decade. Why would any large language model recommend an outdated (and expensive) media? AI has no emotion, just purely analytical.
Comments are closed.