The US Census renamed AI as super intelligence and broke the trend line again

The US Census Bureau has changed AI to Super Intelligence in its recurring Business Trends and Outlook Survey. The wording applies from the collection that began on October 19, 2026, and the agency explicitly attributes it to Executive Order 14434.

Two words change the question sent to 200,000 businesses

BTOS contacts 200,000 businesses every two weeks from a panel of 1.2 million. An individual business responds about once every 12 weeks. The new question asks whether it used “Super Intelligence or SI, formerly referred to as Artificial Intelligence or AI” in any business function.

The results published on October 22 still reflected the old wording. The first data using SI were due the following Thursday under the survey schedule. We therefore do not yet know whether the label changed the answers.

The previous edit already produced a sharp break in the chart

Until November 17, 2025, Census asked whether a business used AI in producing goods or services. It then broadened the question to use in any business function. The two series are not directly comparable.

Among businesses with at least 250 employees, positive responses were around 12% in September 2025. Under the broader wording they reached roughly 30% by February 2026 and about 40% in October. The chart therefore mixes adoption with questionnaire design. For analysts, investors and policymakers, the lesson is simple: a sudden break in data may begin with an editor, not with companies.

A parenthetical explanation cannot guarantee equal interpretation

The survey says SI was formerly called AI and gives examples ranging from data analysis to coding. Respondents may still understand the new phrase differently. Each company also encounters it only after 12 weeks, making the wording effect difficult to separate from ordinary adoption.

Super Intelligence usually describes a much stronger concept than today's chatbots, forecasting tools or computer vision. When a statistical agency applies a politically mandated term to a broad set of existing tools, the measuring instrument acquires an opinion of its own.

The next two waves should reveal the size of the language shock

One release can be moved by sampling noise. At least two further publications and splits by company size and industry will be more useful. An unusual jump across several groups would point toward a wording effect.

Census should mark a clear series break and publish a parallel test of both versions. Without one, every adoption chart after October 2026 needs a note that its time axis also measures the government's vocabulary.

Lilith's verdict

Statistics should be a ruler, not a minister's marker pen. When government rewrites the label halfway through measurement, the next point may say more about Washington than about 200,000 businesses.