Our new research "Opening the Black Box: Generative citation analysis and its implications for reputation and issues management" introduces the first in a series of findings that uncover how LLMs reshape the rules of issues management and raise new challenges for newsrooms.
The research was undertaken by Medianet's media analysis arm, Medianet Insights, based on in-depth analysis of 400 LLM-generated responses from platforms such as ChatGPT, Perplexity, and Google AI Summaries and focusing on leading Australian financial and automotive brands. Download the report.
Key takeaways for comms and media professionals
- A “new” news cycle: LLMs are slow to reflect new, breaking negative news (lagging traditional media by weeks or months) but retain damaging historic issues (like the 2019 Royal Commission) for years, creating a persistent reputational risk unlike the traditional 2-4 week news cycle.
- Corporate content dominates: Contrary to expectations, corporate-owned content (company websites) was the dominant citation source (30%) over mainstream editorial media (22%) for financial brands, implying that brands can proactively shape LLM narratives.
- Sentiment is highly favourable: LLM responses are overwhelmingly positive (97-100% favourable for major banks) compared to traditional media (30-60%), suggesting LLMs primarily pull factual, product-focused information from owned sources.
- Media SoV vs. LLM SoV: Share of Voice (SOV) metrics in LLMs do not mirror traditional media. A brand can have low media visibility but high LLM visibility, proving that media engagement does not guarantee LLM presence (and vice versa).
Implications for media and journalism
While these findings pose additional challenges for corporate reputation management, they also imply additional challenges to the core job of media and journalism.
- Declining traffic and revenue: With 45% of citations straight from the brand’s website, organisations seem to be able to shape the LLM narrative. This presents a new battle for website traffic and a potential threat to media business models reliant on it.
- A balancing act: While adapting to the changing revenue model, publishers must now manage considerations of copyright and intellectual property while simultaneously ensuring visibility and feeding the hungry beast of Generative AI by embedding their content into LLM knowledge bases.
- Weakening power to call out organisational wrongdoing: These trends of increasingly owned content combined with the LLMs’ slower uptake of negative breaking news, have redefined the traditional media news cycle. Do these increasing trends mark the erosion of journalism's influence on corporate reputation and accountability for those in power?