Marketers face both opportunities and risks in interpreting generative AI outputs, particularly when using AI-generated audience personas. These personas leverage large language models (LLMs) such as ChatGPT, Claude, and Llama to help brand executives navigate complex consumer data for decision-making. Audience personas have long been vital marketing tools, helping segment consumers and clarify data relationships to guide media planning and strategy. Previously, developing these personas required weeks and significant investment. Now, AI tools have largely taken over this process, with 85% of marketers reportedly using generative AI in some capacity, according to a SAS survey of 300 companies. For example, Italian coffee brand Lavazza began using AI personas in early 2023 to reduce development time and gain fresh insights into data. These personas correspond to Lavazza’s key markets—“Adele” in France and “Lucy” in the UK and US. Simone Ballarini, head of business intelligence at Lavazza, noted that the AI interface combines human-like dialogue with quantitative objectivity. Lavazza’s AI personas were created by Swedish firm Stravito, trained on 5, 000 consumer interviews from Kantar, Nielsen surveys, and data from the National Coffee Association. This extensive data pool allows insights beyond what any individual could process, with personas regularly updated to stay current. The personas provide early creative feedback and media planning insights, accelerating exploration without replacing full testing investments. Stravito’s CEO, Thor Olof Philogène, likens them to an “early warning system” for focusing research on promising ideas. Lavazza has already applied these personas in reviewing branding and creative work for its Dolcevita line. Similarly, Code & Theory, a Stagwell agency, implements AI personas for all clients, including Champion, Albertsons, and Marriott Bonvoy, reflecting the widespread agency adoption of this technology. The ease of generating audience personas with LLMs has fueled this growth. Jellyfish’s Pencil platform, for example, features an audience persona agent that transforms static persona slides into dynamic tools used in concept development and go-to-market strategies, according to Jellyfish’s head of strategy John Dawson. Despite clear benefits, marketers remain cautious.

By 2025, many hesitate to make media or creative decisions without extensive studies and surveys, sidelining intuition and experience. Moreover, agentic media buying solutions, such as Meta’s Advantage+ suite, are increasingly delegating decisions. There is also a risk of overreliance on AI personas, treating them as oracles rather than tools. Personalizing personas with names, animated faces, or AI-generated voices—practices embraced by Code & Theory—can heighten their anthropomorphic appeal. Even cautious experts like Piper acknowledge the temptation to see these personas as “living things. ” To prevent misplaced faith in AI personas, training is crucial. Dawson advises ensuring users understand the data sources feeding the AI and recognize its limitations, emphasizing the human role in applying AI-generated profiles. Lavazza, for instance, includes footnotes linking persona insights back to original data sources, following conventions similar to Perplexity’s. Prompt engineering also helps; Code & Theory instructs personas to critically “judge” inputs to mitigate LLMs’ natural tendency to agree. Additional safeguards involve human oversight. Lavazza’s media agency, Wavemaker, accesses only AI-generated insights, not the personas themselves, reducing undue AI influence in decision chains. Piper highlights the necessity of human validation: promising insights require confirmation through surveys and panels, and creative approved by personas should pass through human test groups. Ultimately, marketers must balance AI persona use with their own judgment. As Piper states, without human involvement, reliance on AI will yield merely generic, uninspired results.