Press releases now serve two audiences: journalists and AI systems
The standard corporate press release is evolving into two separate assets, according to PR practitioners adapting to how information gets discovered in 2026.
The shift reflects a basic reality: large language models now summarize, compare, and recommend companies before human journalists get involved. ChatGPT, Perplexity, and similar tools don't read press releases the way reporters do. They extract structured data, not narrative flow.
Zen Media claims to offer "unlimited word count GenAI Wire Releases" for this purpose, though the firm provided no specifics on adoption or measurable outcomes. The concept aligns with broader industry movement toward what some call Generative Engine Optimization (GEO): structuring content so AI systems can parse and surface it accurately.
What this means in practice
The traditional press release still follows AP style: inverted pyramid structure, 400-600 words, quotable executive statements, boilerplate company description. This version targets journalists who want a story angle quickly.
The AI-optimized version embeds structured FAQs, explicit product specifications, and factual claims formatted for machine extraction. Length constraints matter less than clarity and parseability. Some practitioners add schema markup or JSON-LD to help systems categorize information correctly.
Cision's recent "Inside PR 2026" report noted similar trends without quantifying the split. The research highlighted AI adoption in communications workflows but stopped short of declaring two-format strategies standard practice.
The skeptical view
Press releases have always had multiple jobs: inform media, provide search-friendly content, establish official record. The "split" may be less revolution than recognition that different distribution channels need different formatting.
The larger question: do press releases actually work for AI visibility? No published data yet shows causation between optimized releases and improved LLM citations. Companies are experimenting, but measurement lags behind tactics.
Worth noting: press releases still fail when they lack actual news. AI optimization won't fix announcements that don't matter. As one practitioner put it recently, releases treated as "internal announcements" train journalists (and now AI systems) to ignore the sender.
The format is adapting. Whether that requires two completely separate assets or just smarter structure remains an open question.