Verified · Jun 29, 2026
Anthropic just released the most detailed one-year view of how Claude is used
3 sourcesAnthropic's June Economic Index 'Cadences' report has three chapters: usage rhythms (7 a.m. news, 6 p.m. recipe requests 2.3x the daily average, April 14 tax conversations 8x the May daily average); artifacts and compute (93% of conversations produce an artifact; Claude Code shows 0.37 points higher autonomy on average); and a perception survey of about 9,700 users (86% / 82% / 69% report gains in speed, scope, and quality).
Why now
The report PDF, appendix, and full method are public as of June 26, 2026, with figures you can cite directly.
Why it is worth publishing
It is a data-dense official report that can fuel three or more separate videos, and each headline number (93%, 8x, 86/82/69) carries its own angle.
Evidence basis
'How AI is actually used' is a perennial creator topic, and the report covers rhythms, output structure, and self-reported impact at the same time, which lets creators push back on both the 'AI is useless' and 'AI can do everything' narratives.
“Anthropic just dropped its June Economic Index report — 93% of Claude conversations actually produce an 'artifact'.”
Angle
Split one report into three charts and one sentence, so viewers understand Claude use from rhythm, output, and self-reported impact at the same time.
Format
Long-form explainer
Demo idea
Pick one usage-rhythm chart (Figure 1.4 tax spike), one artifact-type breakdown (Figure 2.1), and one self-reported impact bar chart (86/82/69), then close with a single sentence: 'When, what, and how well people say it works'.
Platform notes
Always label the sample as Claude-only, about 9,700 surveyed, and the report window as June 2026 — do not generalize single-vendor numbers to 'all LLMs' or 'globally'. When quoting the 'AI will do most of my work next year' perception, pair it with the report's own note that reported exposure exceeds observed exposure, and link to Chapter 3. Cite figures by number (Figure 1.4, 2.1, 3.2) rather than from memory.
Usable claims
- Anthropic published the June 2026 Anthropic Economic Index report, themed 'Cadences', on June 26, 2026, with the report PDF, appendix PDF, and full method disclosed on the research post.
- Anthropic's classifier identified 93% of Claude conversations as producing an artifact, with the most common artifact types being explanations (17%), documents/reports (15%), and guidance (11%).
- On April 14, tax-related conversation clusters were eight times as common as on the average day in May and remained about as high on April 15, a usage spike linked to the US tax deadline.
- Across six dimensions, people with a higher share of automated Claude sessions feel more optimistic about AI's effect on their job outcomes next year, and 86% / 82% / 69% of survey respondents reported gains in speed, scope, and quality of work respectively.
Evidence pipeline
From the news
Breakdown
This breakdown opens the June Economic Index report chapter by chapter: Chapter 1 on usage rhythms (7 a.m. news, 6 p.m. recipes 2.3x the daily average, April 14 tax conversations 8x the May daily average); Chapter 2 on artifacts and compute (93% of conversations produce an artifact; Claude Code 0.37 points higher autonomy); Chapter 3 on the perception survey of about 9,700 users, including the 86/82/69 self-reported gains and the report's own 'reported exposure exceeds observed exposure' caveat.
Sources
Risks
- Always cite the report window, the Claude-only sample, and the survey size; never quote a number as if it applied to all LLMs or all countries.
- Pair any 'AI will do X% of my work' quote with the report's own note that reported exposure exceeds observed exposure, and link to Chapter 3.
Demo ideas
- Turn Figure 1.4's April 14 tax spike into a 30-second 'the day Claude got used the hardest' clip
- Break the 93% artifact figure into three buckets (explanations 17% / documents 15% / guidance 11%) and visualize it as an output pie chart
- Show the 86/82/69 self-reported gains as a 3-bar chart and label the bars 'self-reported', so viewers know it is perception, not an objective metric