New research indicates that the perceived intelligence and trustworthiness of artificial intelligence responses are directly proportional to the credibility of the news brands they cite. This finding, part of the AOP’s ‘Artificial Intelligence Publisher Impact Study’ based on an Ipsos survey of 1,000 UK individuals, underscores a critical dependency for AI platforms, even as they divert referral traffic from publishers. The study highlights a 'trust paradox' within the unequal value exchange, where publishers supply the intellectual property training the very large language models (LLMs) now siphoning their audience.
The Ipsos survey found a direct correlation: a completely trusted source led to over 90% trust in an AI answer, while a completely distrusted source yielded approximately 10% trust, and neutral sourcing resulted in only about 25%. This suggests premium publisher citations significantly elevate an AI interface’s perceived reliability. While trust in a cited source correlates with a user’s propensity to click through (nearly half of those who trust a source will click), high trust also increases 'stop here' behaviors. For instance, 32% of users who completely trust a source reported willingness to cease interaction with the AI answer, comparable to 35% among those who completely distrust sources. A significant 37% of respondents, peaking at 45% among 45-54-year-olds, were unaware AI tools can fabricate information or sources.
LLMs, optimized for plausible answers, operate purely on probability, making 'hallucinations' common. The continuous input of professional human work is presented as the only way for AI platforms to mitigate misinformation as these tools become a primary information source. AI companies, including OpenAI, reportedly pursue licensing deals with premium publishers, recognizing their urgent need for reliable, real-time reporting to anchor their error-prone conversational interfaces. Should publishers withdraw access, AI tool utility would degrade immediately. Regulatory developments, such as the CMA’s June ruling allowing UK publishers to opt out of Google’s AI Overviews, also provide new levers.
Despite this, the confidential nature of individual licensing agreements may impede collective action and the establishment of fair compensation for intellectual property. Publishers are advised to prioritize content types least susceptible to AI substitution and most effective at driving repeat audience engagement, as declining referral traffic is not expected to recover. The current arrangement is deemed unsustainable for both parties, necessitating a new online economy where content creators can continue their work, requiring technology companies to contribute their part, according to reporting first published by pressgazette.co.uk.
