60% of AI Search Answers Are Confidently Wrong

Published on March 13, 2025

Snapshot: A new study from the Tow Center for Digital Journalism highlights how AI-powered search tools are failing to properly credit original sources, raising serious concerns about content visibility and information reliability. While traditional search engines drive traffic to news publishers, generative AI search models repackage content in conversational formats—often misattributing or fabricating sources.

Source: The Tow Center / CJR

 

The research, which tested eight AI search tools across 1,600 queries, found that more than 60% of responses contained incorrect citations, with some models bypassing publisher restrictions to access content. The study also found that premium AI models, despite their cost, were more likely to generate confidently incorrect responses than free versions. Even AI models with formal content licensing agreements frequently cited syndicated or unauthorized versions of articles instead of original sources.

These findings underscore the urgent need for AI platforms to improve citation accuracy and respect publisher rights. As AI search tools become more prevalent, their impact on news traffic, brand credibility, and public trust in information is a growing concern.

Full story: Columbia Journalism Review

Other news

Tariff Uncertainty Looms Over Ad Budgets

March 11, 2025
The latest round of trade tariffs could have far-reaching consequences for advertising spend, with auto, retail, and tech brands expected to be hit hardest.

CMO Role is Evolving, Tenure and Influence on the Rise

March 14, 2025
The latest Spencer Stuart CMO Tenure Study highlights growing stability in marketing leadership, with CMOs increasingly advancing to higher positions. However, the role itself is changing, often merging with other functions like growth, customer experience, and revenue.