Desk Trials

ARCHIVE / Evidence

Stanford's AI Index reports numbers it credits to other sources

The annual AI Index compiles labor-market and economy figures from named outside providers rather than measuring them itself.

Preserved retrospective record

Historical source and event dates are not site publication dates. Product plans, policies and availability may have changed since retrieval.

Visual for this record: Stanford's AI Index reports numbers it credits to other sources
Visual published by hai.stanford.edu, shown for identification of the record. Credit: hai.stanford.eduPreserved source visual · owner review pending

The setup

Stanford's Institute for Human-Centered AI (HAI) publishes the AI Index Report annually, describing itself on its own project page as an effort to provide unbiased, rigorously vetted and globally sourced data on artificial intelligence across technical, economic and societal measures. Unlike a survey organization or a statistical agency, HAI does not run its own labor-market data collection for most of the report; instead it assembles and charts figures produced by a roster of named partners, including LinkedIn, the labor-market data firm Lightcast, McKinsey, GitHub and the International Federation of Robotics, among others listed on the project's own page, as retrieved 16 September 2026.

What the documents show

The most recent edition, the 2026 AI Index Report, devotes its fourth chapter to the economy, which studies how AI is reshaping labor markets, productivity and jobs across industries, according to the report's own landing page. Because the chapter's charts are compiled rather than independently measured, a given figure's real source is whichever partner organization HAI credits for it, not HAI's own research staff.

The friction

That structure is also the report's main limitation for a reader who wants to check a number: HAI's landing pages do not carry a source-by-source citation list for every figure on the summary page, so tracing a specific economy-chapter statistic back to its original provider means locating the citation inside the full report rather than the summary. A reader who quotes the AI Index without also naming the original data provider is citing an index of citations, not a primary measurement.

What changed in the work

What the Index adds, done well, is a single reference point that saves a reader from separately tracking down McKinsey's, LinkedIn's and Lightcast's individual reports, provided the citation trail back to each original source is preserved rather than dropped when a chart gets copied out of context. Treating an AI Index chart as if Stanford itself measured the underlying figure is a misreading the report's own compiled structure does not support.

  • Does a quoted AI Index figure name the original data provider credited in the report, or only the Index itself?
  • Is the underlying source a survey, a logged dataset, or a government series, and does the Index distinguish between them?
  • Has the specific edition and chapter been identified, given that the report updates yearly and its contents can change?

The Index's value is closest to that of a well-organized bibliography with charts attached: useful for finding the range of what is being measured across the field, but only as reliable as the citation trail a reader is willing to follow back to each original source.

Sources & verification

Preserved from the earlier archive. These sources have not all been freshly rechecked for this expansion.

  1. AI IndexSource date: not stated · Retrieved: 2026-09-16

    Stanford HAI's own project page listing named data-analytics partners, including LinkedIn, Lightcast and McKinsey, that supply figures compiled into the report.

  2. 2026 AI Index ReportSource date: not stated · Retrieved: 2026-09-16

    The 2026 edition's own landing page describing its fourth, economy-focused chapter's coverage of labor markets and productivity.

Continue the workflow

  1. Run a workflow pilot that can answer a real question

    Decide whether a proposed workflow deserves wider use using a modest, honest pilot.

  2. McKinsey's AI survey samples opinion, not a random slice of firms

    McKinsey's own methodology note describes an opt-in online survey of 1,719 executives, not a probability sample of the economy.

  3. Epoch AI's compute numbers are mostly estimates, and it says so

    Epoch AI's own methodology shows most training-compute figures in its tracker are estimated, not measured, and it labels the difference.

  4. GitHub counts an AI project by its code imports, not a survey

    GitHub's 2025 Octoverse defines an AI project by dependency signals in the code itself, separate from any developer-sentiment question.