New research finds large language models source almost two-thirds of content on brand reputation from editorial coverage, while social media barely registers.

Large language models (LLMs) predominantly depend on editorial coverage for information relating to brand reputation, sourcing almost two-thirds of insights (61%) from news outlets, according to new research. In contrast, social media coverage plays a minimal role.

The study by communications consultancy Hard Numbers finds that, while less than 1% information is sourced from social media, owned media proved an influential source, comprising 44% of content. Hard Numbers co-founder and creative director, Paul Stollery, highlights the enduring power of editorial media. “Everyone predicted digital would kill traditional media twice: first with social, then with AI. But they were wrong. What we’re seeing from this research is that AI is becoming an amplifier for editorial authority.

“What’s clear from this research is that the single best way to influence the answers on these platforms is to earn the right to be covered by respected journalists at influential outlets.”  

The research also found editorial content is a critical source for LLMs when assessing the trustworthiness of a company, contributing 65% of responses, and value for money, 72%. Stephen Waddington, non-executive director at Hard Numbers, stresses the importance of understanding how AI processes information. “Understanding these dynamics will be essential for corporate communicators to develop effective digital strategies for an AI-augmented future.”