Adoption Monitor — Stanford Digital Economy Lab
How fast are individuals and firms adopting AI and related technologies? Many contemporary AI tools are low- or no-cost and available digitally, allowing for rapid adoption. Capturing a rich picture of AI adoption requires drawing on a number of sources. Here, we track individual- and firm-level trends across surveys and countries.
While widespread adoption is neither necessary nor sufficient for broader economic impact, such adoption can indicate growing or shrinking importance of the technology in everyday and economic life.
— the Monitor’s own introduction
TL;DR
The Stanford Digital Economy Lab’s adoption indicator. It collects published surveys rather than running its own: four US surveys of individuals and one executive survey of firms in four high-income countries. This snapshot is the 1 July 2026 update.
- Individual adoption keeps rising. In the Bick, Blandin & Deming survey, 61.8% of US adults used generative AI in May 2026, up from 44.6% in August 2024.
- Daily use is still low. Nearly 90% of users use it weekly, but only about a quarter use it daily: 15.4% of all adults. The page calls the extensive margin (who uses it at all) and the intensive margin (how often) “distinct.”
- Work-use surveys disagree. Gallup (52%) and Bick, Blandin & Deming (45%) show work use still rising toward half of workers. Hartley et al. show it falling, from a 45.9% peak in June 2025 to 32.7% in June 2026.
- Firm adoption is high and expected to grow only a little. In Yotzov, Barrero, Bloom et al. (NBER 34836), 68.6% of firms use some AI technology, and 74.7% expect to within three years. On net, only about 6% of firms expect to move from no AI to some.
- Firms expect to use LLM text generation less. Firms expect more use of every AI application except one: LLM text generation, which they expect to fall from 41.5% to 31.4% of firms. US firms expect it to fall from 53.5% to 28.8%. The page reports this without explaining it.
- US firms lead, but not by much. Current adoption is 77.5% in the US, 71.3% in the UK, 64.7% in Germany and 59.3% in Australia. Executives expect the gaps to narrow.
Key data (snapshot)
Individuals, United States (share of adults; work-use rows share of workers):
| Survey | Measure | Earliest | Latest |
|---|---|---|---|
| Bick, Blandin & Deming (BBD) | Any generative-AI use | 44.6% (Aug 2024) | 61.8% (May 2026) |
| BBD | Used in the last week | 36.1% (Aug 2024) | 54.2% (May 2026) |
| BBD | Used every day | 12.1% (Aug 2024) | 15.4% (May 2026) |
| Pew Research Center | Ever used ChatGPT (only) | 18% (Jul 2023) | 44% (Feb 2026) |
| BBD | Any use at work | 33.3% (Aug 2024) | 45.2% (May 2026) |
| Gallup | Use at work a few times a year or more | 21% (May 2023) | 52% (May 2026) |
| Gallup | … a few times a week or more | 11% (May 2023) | 30% (May 2026) |
| Gallup | … daily | 4% (May 2023) | 15% (May 2026) |
| Hartley, Jolevski, Melo et al. (HJMM) | Use at work | 30.1% (Dec 2024); peak 45.9% (Jun 2025) | 32.7% (Jun 2026) |
Firms, four countries (share of firms, reported by senior executives November 2025 to January 2026; “expected” is in three years):
| Technology | All firms now → expected | U.S. | UK | Germany | Australia |
|---|---|---|---|---|---|
| Any AI technology | 68.6% → 74.7% | 77.5 → 76.6 | 71.3 → 79.2 | 64.7 → 72.5 | 59.3 → 64.3 |
| Text generation using LLMs | 41.5% → 31.4% | 53.5 → 28.8 | 33.5 → 29.8 | 46.0 → 34.3 | 30.1 → 30.9 |
| Data processing using ML | 28.1% → 40.1% | 36.9 → 45.9 | 33.2 → 43.5 | 21.3 → 36.6 | 22.1 → 31.6 |
| Visual content creation | 28.9% → 32.7% | 32.5 → 32.8 | 29.0 → 35.1 | 29.5 → 31.8 | 20.1 → 28.4 |
| Image processing using ML | 19.7% → 29.5% | 26.7 → 31.6 | 24.5 → 32.4 | 14.5 → 27.2 | 11.8 → 24.8 |
| Robotics | 9.1% → 20.7% | 13.0 → 22.3 | 13.6 → 21.3 | 3.9 → 18.4 | 7.5 → 23.6 |
| Autonomous vehicles | 3.5% → 15.9% | 4.5 → 14.5 | 3.0 → 13.6 | 2.6 → 17.5 | 6.6 → 20.3 |
Method, as the page states it
- Individuals. Self-reported adoption from several surveys. They “differ in their sample size, the populations they represent, and elicitation strategy.” Self-reports capture use people recognise as generative AI and “might fail to capture embedded or ambient use, as well as integrated tools like spell checking and autocomplete.”
- Firms. Use of AI in production, reported by company leadership. The page argues this, and the cost firms take on, “could provide a more reliable signal of the value of the technology for economically meaningful tasks,” though adoption and diffusion frictions blur it.
- Sources. Bick, Blandin & Deming (2026, Management Science); Gallup’s AI indicator (April 2026); Pew Research Center (2025); Hartley, Jolevski, Melo et al. (2026, SSRN); Yotzov, Barrero, Bloom et al. (2026, NBER Working Paper 34836). Nicholas Bloom, a co-author of the firm survey, is the page’s featured affiliate.
Reading the snapshot
The corpus now holds adoption figures that differ by a factor of three or more, and the Monitor shows why. It shows several instruments side by side, for individuals and for firms:
| Unit | Figure | Instrument | Source |
|---|---|---|---|
| US adults | 61.8% use generative AI (May 2026) | Household survey, any use | this Monitor (BBD) |
| US adults | 44% have used ChatGPT (Feb 2026) | Survey, one product only | this Monitor (Pew) |
| US population | 28.3%, 24th in the world | Cross-country diffusion estimate from the Microsoft AI Economy Institute (AI Index 2026, pp. 200–201) | AI Index 2026 |
| Firms, US/UK/DE/AU | 68.6% use some AI (late 2025) | Executive survey, use in production | this Monitor (Yotzov et al.) |
| Organisations, global | 88% (2025) | Organisational survey; the 2025 edition’s 78% came from McKinsey’s survey, counting use in at least one business function | AI Index 2026 |
| Firms, OECD | 20% (2025); 50%+ of big firms, ~17% of small | Official firm statistics | Rendtorff-Smith (OECD) on BCG |
| Firms, US | ~10% (use in current production) → ~20% (any business function) | Census survey, before and after a wording change | Hatzius (Goldman Sachs) |
None of these is wrong. They count different things: any use or production use, one product or all, firms weighted by number or by size, executives or employees. A figure about “AI adoption” should not be quoted without its instrument. The Monitor is the first source in the corpus to set them side by side for one country.
The work-use split is the open question. Two surveys show work use still rising and one shows it falling by a third in a year. The page reports this without resolving it. A self-report that falls while use spreads into tools people no longer call “AI” is one possible reading, and the page’s own caveat about embedded use points that way. Anicich & Brouwers give another reason self-reports undercount: employees hide their AI use.
The expected fall in text generation needs explaining. US firms expect the share using LLM text generation to roughly halve while every other application grows. The page gives no reason. Possible readings: firms expect text generation to be absorbed into other tools or agents and no longer reported separately, or the question was read as stand-alone chatbot use. Treat it as unexplained.
Note on the firm survey (added 2026-11-07). The firm figures come from the February/March 2026 versions of NBER w34836 (four countries). The October 2026 revision (2026-10-01-yotzov-barrero-bloom-nber-w34836-firm-data-on-ai) reports 78% across seven executive surveys (N=8,879, May–July 2026). It also recodes US non-response, so firms that skipped the question are no longer counted as non-users, which raises US adoption to 94–95%. The question asks about current use, not use in production as this page puts it, and the category “text generation using LLMs” became “Generative AI”.
Dynamic-capabilities reading
digital-sensing/digital-scouting: a standing scan of how fast people and firms are taking up AI, across surveys and countries.contextual/external-triggers: individual adoption at 62% of US adults is a change in consumer behaviour that firms meet whether or not they adopt AI themselves. The firm survey describes the competitive field: a large majority of firms in four high-income economies already use some AI.
How it connects
- Yotzov, Barrero, Bloom et al., NBER w34836 — the working paper behind the firm figures, now in the wiki in its October 2026 revision (see the note under Method).
- enterprise-ai-adoption — the firm figures, and the spread of adoption figures across instruments.
- technology-adoption-theories — the extensive and intensive margins of adoption, measured separately.
- micro-productivity-trap — adoption without impact; the page’s “neither necessary nor sufficient.”
Related in this wiki
- Hatzius (Goldman Sachs) — the Census wording change, the clearest case in the corpus of an adoption figure made by its question.
- Frey on Trumponomics — “adoption for what?”: high adoption for email does not drive growth.
- AI Index 2026 — a US adoption figure (28.3%) about half the Monitor’s, and an organisational figure (88%) above it.
- Rendtorff-Smith (OECD) on BCG — official firm adoption at 20%, and the gap by firm size.
- Anicich & Brouwers — why self-reported work use may undercount.
- Gillespie et al. (KPMG) — a 47-country survey: 66% use AI intentionally on a regular basis, with emerging economies ahead.
- Canaries Dashboard and Transformation Tracker — the other two indicators ingested with this one.
What was actually ingested
The full page as served on 4 November 2026, and the full data of all four Flourish charts, saved as CSV beside the raw file. The page’s last update is 1 July 2026; its firm summary was last updated 10 June 2026. The underlying surveys were not read; the figures are as the Monitor reports them. The Gallup frequency rows come from the frequency chart, which plots Gallup’s three frequency thresholds separately.
Linked entities and concepts
- Stanford Digital Economy Lab — publisher. Updated in this ingest.
- enterprise-ai-adoption, technology-adoption-theories, micro-productivity-trap — concepts, updated in this ingest.
Dangling (not in author:; deferred): Nicholas Bloom (featured affiliate, firm-survey co-author; also acknowledged in the Canaries paper), Alexander Bick, Adam Blandin, David Deming, Jonathan Hartley, Ivan Yotzov, Jose Maria Barrero, Gallup, Pew Research Center, Connacher Murphy.
Scope and reliability
A curated collection of other people’s surveys. Its value is the side-by-side view and the methodological notes, not new data. It does not reconcile the surveys that disagree. Two of the five sources are peer-reviewed or NBER working papers; Gallup and Pew are established pollsters; the Hartley et al. paper is an SSRN working paper. The firm survey covers four high-income countries and asks senior executives, who may report adoption more generously than official firm statistics do. This is a living source; the snapshot will move as new survey waves are released.