EXECUTIVE SUMMARY

• Capital discipline has become the dominant media frame. Perscient's semantic signature tracking the density of language asserting that businesses increasingly doubt large AI spending posted the single largest move in the signature set, propelled by a sell-off in big tech shares, Meta's raised capital-spending guidance, Ken Griffin's "almost certainly" bubble remark, and Financial Times reporting on delayed or blocked data center projects. Coverage frames the question as one of financing structure and allocation credibility rather than whether the technology works.
• Doubt about spending is not the same as doubt about the product. Our signature tracking the density of language asserting that promised AI efficiency improvements have not occurred sits near its average level, and the signature tracking claims that AI powers sustained market gains and economic expansion was flat and below its mean. Media are simultaneously carrying contrarian voices — Weiss Multi-Strategy's "hedge fund leverage popped, not the bubble" argument and IoT Analytics' dissent — and uncomfortable adoption statistics, with Gartner reporting that only 13 percent of CIOs claim realized AI value. The result is a press environment that questions how AI is being paid for while mostly leaving efficacy claims unchallenged.
• Benefit and broad-prosperity stories remain the quietest part of the conversation, though they are stirring. The signature tracking language connecting AI to efficiency improvements and universal basic income is the weakest reading in the set, yet it recorded the second-largest monthly gain, driven by OpenAI's proposed "Public Wealth Fund" and a 2026 debate that has moved from cash transfers toward public ownership stakes. Distribution-of-gains coverage is appearing chiefly as a pre-IPO positioning story rather than as evidence of delivered benefit.
• Synthetic content is now the structural backdrop of all AI coverage. Our signature tracking the density of language claiming that AI is producing, writing, or creating news content remains the strongest reading in the entire set and strengthened further, while the companion signature on AI-generated fake videos as a societal threat holds at an elevated level. The US midterm cycle supplies the proximate driver — 220 AI-generated political ads backed by more than $137 million, roughly 70 percent of them undisclosed — alongside Brazil's runoff, Israel's online campaign, and an Iranian operation that placed fabricated stories through imaginary reporters.
• Taken together, the two threads describe a narrative center of gravity moving from promise to proof. Financial coverage demands evidence that capital expenditure converts to returns, while information-integrity coverage supplies abundant, concrete evidence of AI's visible effects in the form of deepfakes and synthetic news. Because benefit language stays muted and synthetic-harm language stays loud, the most available proof points in circulation are adverse ones — a asymmetry that shapes how both investors and regulators are likely to read the technology in the weeks ahead.