
Image credit: Search Engine Journal
Global digital marketers are misinterpreting and mismeasuring the return on investment for artificial intelligence visibility due to fundamental architectural and liability differences between search engines and large language models, industry analysts said.
Traditional search engines were designed to provide users with multiple options and route them to third-party websites, inherently shielding the engine from liability as users made their own choices, according to a report by Conductor and Similarweb.
Conversely, large language models (LLMs) such as those from OpenAI and Anthropic are engineered to directly answer user queries, making any referral traffic to external sites a byproduct rather than a primary design objective, the report stated.
This shift in design alters the liability framework, moving responsibility for generated content from the user’s choice to the LLM itself, as the models produce answers in their own voice.
Legal precedents, like the case involving Air Canada’s chatbot, which provided false information leading to the airline’s liability, highlight the concept of “reasonable reliance” and the diminishing protection offered by disclaimers for specialized AI tools.
The New York Times has also sued OpenAI and Microsoft, alleging copyright infringement regarding content used to train LLMs.
Measuring the return on investment (ROI) for AI visibility faces a significant challenge known as the “denominator problem,” according to Conductor.
While the direct referral share from AI models to publishers may appear low, the overall search-driven traffic to publishers has experienced a substantial decline.
This broader collapse in traditional search traffic makes the small AI referral share seem disproportionately insignificant when viewed in isolation, Conductor explained.
Organizations like Noyb, a European digital rights group, have also raised concerns about data privacy and the operational methods of LLMs.
Business Insider and HuffPost are among the publishers that have reported significant drops in search traffic.
Source: Search Engine Journal
Written by
Joyce de Castro
Joyce is a core team member at Rabbit Rank and the lead author covering SEO news, algorithm updates, industry trends, and actionable ranking strategies.
Keep reading
Related Articles

Google expands ‘Preferred Sources’ feature globally for news
Google’s ‘Preferred Sources’ feature is now global, letting users prioritize news sites in search results. Pub...

Google: Preferred Sources Don’t Override Quality Signals
John Mueller addresses if Google’s Preferred Sources can override low-quality signals in Top Stories, clarifyi...

Advertisers Adapt Strategies for Emerging AI Ad Placements
Advertisers are adapting strategies for AI ad placements, accessing inventory, and measuring performance as AI...