Contents

According to Simon Willison’s November 2, 2026 report, the U.S. Census Bureau changed the wording in its business survey from “artificial intelligence” to “superintelligence.” Anyone using its AI adoption figures should disclose this change separately from the earlier expansion in the question’s scope.

Key Points#

  • Citing Executive Order 14434, the survey introduced the phrase “superintelligence or SI” starting with the October 19, 2026 survey.
  • The results released on October 22 use the previous AI wording. The first results using the new wording are scheduled for November 5.
  • A November 2025 change broadened the question from AI use in production to AI use in any business function, preventing direct comparisons.
  • Among businesses with 250 or more employees, the share reporting use under the broader question rose from about 30% in February 2026 to about 40% in October.

The Current Question#

The Business Trends and Outlook Survey (BTOS) asks businesses whether they used certain technologies during the preceding two weeks. The latest question reads:

> In the past two weeks, did this business use superintelligence or SI (previously referred to as artificial intelligence or AI) in any of its business functions?

Examples include data analysis, generative tools for marketing or coding, customer service agents, forecasting, pricing, and computer vision for quality control. These are common business uses of AI. A “yes” response does not indicate that the business uses systems with capabilities exceeding human intelligence.

According to the Census Bureau’s explanation quoted by Willison, the wording change stems from Executive Order 14434. This establishes the official reason for the survey revision, but does not show whether respondents will interpret the new term consistently.

The Bigger Break Came in 2025#

An earlier revision changed the question’s scope. Before November 17, 2025, businesses were asked whether they used AI “in producing goods or services.” After the revision, they were asked about AI use “in any of its business functions.” The bureau announced this revision on December 3, 2025.

The broader wording includes uses outside production, such as marketing and administrative work. A business could therefore change its answer from “no” to “yes” because the question changed, even if its actual technology use remained the same.

This matters when comparing figures. In September 2025, about 12% of businesses with 250 or more employees answered “yes” to the narrower question. In February 2026, about 30% answered “yes” to the broader question. Because the survey measures different ranges of activity, these percentages do not support a claim that adoption more than doubled.

What the Figures Show#

Comparisons within 2026 are more useful. Under the “in any of its business functions” wording, the share of businesses with 250 or more employees reporting use rose from about 30% in February to about 40% in October. That is an increase of about 10 percentage points.

These figures measure reported use during a two-week period. They do not measure spending, deployment scale, productivity gains, or return on investment. A business using generative tools for limited tasks and one running AI throughout its operations can both answer “yes.”

The October 22 results also cannot reveal the effect of the new “superintelligence” label. According to Willison, the published figures used the previous AI wording. As of November 2, the first results reflecting the new wording were scheduled for release on Thursday, November 5.

How to Read the Next Release#

Treat the terminology change as something that could affect measurement and needs to be tested. The parenthetical explanation, “previously referred to as artificial intelligence or AI,” may help respondents understand the question. Whether it prevents confusion remains unknown.

BTOS has a target population of 1.2 million businesses and surveys a panel of 200,000 every two weeks. Each business participates once every 12 weeks, so respondents encountering “superintelligence” will generally be comparing it with a question they saw several months earlier.

When updating charts or reports, keep the following in mind.

  • Mark November 17, 2025 as the date the question’s scope changed.
  • Identify the first release using the October 19, 2026 wording.
  • Compare results within periods using the same question, and check uncertainty estimates before highlighting small changes.
  • When describing changes after the terminology switch, avoid attributing a cause until supporting evidence is available.

Next Steps#

Before using the November 5 release to update claims about trends, annotate AI adoption charts with both changes to the question’s wording.

Source: https://simonwillison.net/2026/Nov/2/census-bureau-super-intelligence/

Contents