What 700 AI-generated buying responses reveal about how the market actually perceives JE Dunn.
Across 700 simulated buying responses, three patterns emerge with devastating consistency.
314 of 700 responses contain language like "due to the confidential nature of semiconductor projects" or "specific client details are not publicly available." This is the AI's way of saying: I can't verify JE Dunn's claims because there's nothing for me to cite.
More than a third of responses use phrases like "JE Dunn may have experience" or "it is possible that JE Dunn has worked on." The AI is uncertain — not because JE Dunn lacks the experience, but because the evidence isn't structured in a way LLMs can confirm.
838 of 2,319 analyzed citations point back to jedunn.com. When over a third of all source evidence comes from JE Dunn itself, the AI treats the information as less authoritative — the digital equivalent of "source: trust me."
Turner Construction appears unprompted in 25.7% of all responses — but the distribution across stakeholders reveals where the threat is most acute.
The source authority hierarchy reveals why LLMs treat Turner's claims with more confidence than JE Dunn's.
| Source Category | JE Dunn Citations | Turner Citations | Gap |
|---|---|---|---|
| Unique Third-Party Domains | 8 | 15+ | –7+ domains |
| First-Party Source Dependence | 36.1% | ~18% | 2× higher self-citation |
| Verifiable Semiconductor Case Studies | 1 (LG Chem) | 4+ | –3+ case studies |
| Industry Award Citations | Deloitte Best Managed (under-leveraged) | Multiple ENR, AGC, ABC | Award gap |
| Executive Thought Leadership Pieces | Minimal | Frequent (panels, bylines) | Presence gap |
| "Confidential client" Mentions | 116 appearances | Rare | Credibility friction |
This is the single most important metric in the assessment — and the one JE Dunn can act on fastest.
When a buyer asks an AI "What semiconductor fabs has JE Dunn built?", the model searches its training data for verifiable evidence. When it finds references to "confidential clients" instead of named projects, it does what any cautious analyst would do — it hedges. The AI generates responses like "JE Dunn has reportedly worked on semiconductor projects, though specific details are not publicly available."
That hedge is confidentiality friction. It's the AI telling the buyer: I can't confirm this.
In a competitive evaluation, friction is a disqualifier. When Turner's project references are publicly citable and JE Dunn's are confidential, the AI presents Turner with authority and JE Dunn with uncertainty — even if JE Dunn's actual track record is equal or better.
The buyer never sees JE Dunn's real capability. They see the AI's confidence gap — and that gap shapes the shortlist before any human conversation happens.
The friction rate varies dramatically by stakeholder — revealing where the information gap hurts most.