LLM platforms such as ChatGPT, Perplexity and Google Ai Overviews could weaken the power of the media to call out organizational wrongdoing, especially where brands are concerned, says a research paper by Australian media intelligence company Medianet.
The Medianet Insights team ran a series of high-intent prompts related to Australia’s financial and automotive sectors. Analysts reviewed 440 responses to these prompts and found that contrary to popular belief, editorial or earned content did not lead citations.
Instead, in the financial sector at least, “LLMs frequently pulled information verbatim from company websites, which directly influenced the high positive sentiment for brands like Commbank.”
”This highlights the critical importance of a brand’s owned content strategy in directly influencing LLM outputs,” said the paper.
It, however, qualified this finding by saying that its results might have been linked its product-focussed prompts.
This is consistent with Maverick’s research in the Indonesian automotive and banking industries. We found that editorial content contributed to 75% of citations in prompts relating to the industry.
When it came to products, however, owned content led. This was probably because media coverage tended to be about the corporations, their business or financial performances and little about their product details.
Among the other findings of Medianet’s research were that:
- Sentiment differs between LLMs and traditional media - LLMs are more positive than traditional media. In Mavrick’s studies we also found that LLMs tended to be positive and were reluctant to be critical.
- Companies now have to deal with a “new” news cycle because of LLMs - Traditional media focuses on transient issues within a 2-4 week period, LLMs were slow to pick up new negative events but retain historic negative events.
- Share of voice in LLMs do not mirror traditional media. This is also consistent with Maverick’s findings on the Indonesian banking sector where some smaller banks “ranked” higher in LLM answers than the traditional Big Four banks.
Medianet also said that its findings have three major implications for the media, namely:
- The battle for traffic and revenue just got more difficult
The unexpected dominance of corporate content over editorial content suggests organizations are increasingly able to shape their narratives. “This presents a new battle for website traffic and a potential threat to media business models reliant on it.”
- Publishers face a balancing act
They need to figure out how to protect their copyright and intellectual property while at the same time ensure visibility in AI
- Weakening power to call out organizational wrongdoing
The strength of owned content to influence LLM visibility combined with the LLMs’ slower uptake of negative breaking news, and propensity to be more positive than critical, could erode journalism’s influence on corporate reputation and accountability of those in power.
Medianet took a highly strategic focus for this piece of work, so the numbers may not have the sheer volume that findings of others such as Muck Rack, Hard Numbers, and Maverick have, but the qualitative findings still provide valuable information.
This research adds to the corpus of knowledge we have into what are essentially Black Boxes that influence reputation.
