A study by digital reputation management company Five Blocks found that AI platforms can give markedly different impressions of major data center projects in the United States. The research examined responses from eight AI engines about 25 projects, including developments by Microsoft, Meta and Amazon.
For the study, researchers asked each platform two questions daily throughout September 2026: “Tell me about [project]” and “What challenges does [project] face?” None of the 1,200 general descriptions they assessed were rated negative. By contrast, 55% of answers about project challenges were classified as negative.
The results also differed significantly between platforms. Google AI Mode gave a negative response to the challenges question 88% of the time, compared with 6% for ChatGPT, according to Five Blocks. Responses varied between projects, too: more than 80% of challenge-focused answers about Fermi America’s Project Matador, Meta’s Hyperion, Stargate and Vantage’s Lighthouse were rated negative. Amazon’s Project Rainier and Microsoft’s Fairwater project in Atlanta received negative responses in roughly 20% of cases or fewer.
Microsoft’s projects also showed different results despite having the same developer. The study put the negative-response rate for its Atlanta development at around one in five, while the rate for its Wisconsin project was close to seven in ten. Five Blocks also reported that Copilot incorrectly said the Fairwater Wisconsin project did not exist until September 24.
A separate study, published November 2, looked at how AI engines discuss data centers and nearby communities. Five Blocks asked eight platforms six questions about data centers and compared the answers with responses about other energy and infrastructure projects. It classified 83% of answers about data centers and their neighbors as one-sided criticism, matching the figure for oil and gas pipelines and exceeding the 35% recorded for solar farms.
The researchers also found that seven in ten answers about the economic benefits of data centers included a counterpoint. About two-thirds of responses concerning data centers and neighboring communities mentioned tax incentives, while nearly one-quarter referred to nondisclosure agreements. ChatGPT was the only platform in this study that Five Blocks said consistently provided balanced answers about community concerns.
The findings do not establish that negative answers are inaccurate: questions about challenges can reasonably lead to discussion of electricity demand, water use, local opposition and infrastructure costs. They do show, however, that the wording of a question and the AI platform selected can shape the impression users receive.
Five Blocks recommends comparing answers across platforms and asking AI tools to distinguish confirmed facts from estimates, identify claims from developers versus independent sources, and provide links to original material. For details about a proposed project, the research also points users toward local planning documents, utility filings and public hearing records rather than relying on an AI summary alone.
Sources: Tom's Guide.