Earned Media, AI and Brand Reputation | Tina Kozak of Franco

PRGN Presents • Season 8, Episode 5 • Originally published as an audio podcast on March 5, 2026. Is public relations dead, as yet another round of breathless headlines claims? Tina Kozak, CEO of Franco in Detroit, joins Abbie Fink and Dr. Adrian McIntyre to argue the opposite, and to talk about earned media, AI and authenticity in brand reputation management. Tina says brands are asking Franco for more earned media than at any time in the last decade, because the large language models behind AI search rely on trusted third-party sources. The same technology makes it easier for bots and bad actors to manipulate a brand’s story, so she urges clients to monitor conversations closely. Abbie argues that the lessons of social media’s shiny early days apply to AI. Adrian asks how firms rooted in traditional PR can take on new disciplines like generative engine optimization. Tina describes Franco’s new AI enablement services, which pair its own strategy with technical specialists while the agency looks after the brand. She closes with a photo of Tony Franco, who founded the firm in 1964. He’s standing on a baseball field in a trench coat and sunglasses, beside the CEO of client Domino’s Pizza. The role has never changed, she says: know the client, the opportunities and the risks, and don’t back away from being visible. KEY TAKEAWAYS • Brands are asking for more earned media than at any time in the last decade, because AI search relies on trusted third-party sources. • Monitoring conversations about a brand matters more as bot attacks and manipulated narratives become easier. • The lessons of social media apply to AI: plan for its risks rather than chase what is new and shiny. • Agencies can lead AI enablement and generative engine optimization work, bringing in technical partners while keeping charge of the brand. • Being proactive and visible, with preparation, beats retreating from public opportunities out of fear of risk. Read the full episode notes and transcript: