Only 69% of links generated by ChatGPT were real and correctly attributed, raising new concerns about how B2B buyers validate brands in AI-mediated discovery
PAN, a global, integrated, data-driven marketing and PR agency for B2B tech and healthcare brands, released new original research showing that a significant share of B2B citations surfaced by ChatGPT are inaccurate, misattributed, or entirely fabricated, introducing a new credibility risk for brands as buyers increasingly rely on AI-generated summaries to evaluate vendors and solutions.
PAN analyzed more than 11,000 links generated by ChatGPT across executive-level research prompts, originated as a part of its C-Suite Signals research program, designed to reflect how senior B2B leaders investigate their market and technology solutions.
Marketing Technology News: MarTech Interview with Omri Shtayer, Vice President of Data Products and DaaS at Similarweb
Each citation was reviewed and classified as:
real and correctly attributed,
misattributed, or
fully hallucinated.
In total, only 69% of links surfaced by ChatGPT were both real and correctly attributed.
The remaining 31% of citations either pointed to incorrect sources/domains (19%) or referenced sources that did not exist at all (12%).
The findings show that AI systems still frequently imply legitimacy through citations that appear authoritative on the surface, even when the underlying sources cannot be verified. As a result, perceived credibility can be created — or undermined — without brands having any direct visibility into how those references are formed.
Marketing Technology News: Martech Architecture For Small Language Models: Building Governable AI Systems At Scale
For brands, this represents a new and rapidly emerging trust gap.
“For our clients, credibility is no longer something you build once and protect later. It’s something you have to earn, and re-earn, every time a buyer or an AI system looks you up,” said Darlene Doyle, Chief Client Officer at PAN. “When nearly a third of AI-generated citations can be inaccurate or fabricated, clients need partners who are in it with them, helping make their proof easier to find. At PAN, we help brands close that credibility gap by aligning PR, content, web, and executive visibility so what buyers see, what AI summarizes, and what a comPANy can actually deliver all match.”
Buyers are increasingly using generative AI as an early research layer to understand markets, compare vendors, and validate claims before visiting comPANy websites, speaking with sales teams, or engaging analysts. When AI-generated answers surface inaccurate or fabricated citations, brands can be unintentionally associated with unreliable or misleading sources (or omitted from credible ones) in ways that materially influence shortlists and perception.
Write in to psen@itechseries.com to learn more about our exclusive editorial packages and programs.