What we do

breinz Labs

We try ideas on public sources and on our own data and research. Most tests run on trial and error and stay in-house. What holds up and can be shared appears here, for anyone to read and use.

In beta, updated as often as time allows.

White paper
2026.10.31

AI True Cost Index

How much time AI really saves

Of every hour AI promises to save, 24 to 26 minutes come back. The paper follows the rest of the hour through failures and checking, and compares uses of AI, age groups, and the US with Europe and the rest of the world.

See where the hour goes
AI True Cost Index cover
White paper
2026.11.03

Two years to first AI agent

How people adopt AI, an innovation that keeps changing

The first users needed two years to get from asking what AI is to building agents. The paper follows how questions about AI change as it spreads, in Europe, the US and the world.

See how people adopt AI
Two years to first AI agent cover
White paper
2026.11.04

English is only half of AI use

Most of what we know about how well AI works comes from English

About half of the conversations in a large public record of ChatGPT use are in other languages. Conclusions drawn from English results therefore describe half of real use at most.

See which languages AI works in
English is only half of AI use cover
White paper
2026.11.06

Is AI taking jobs?

It depends on the job

Desk tasks of every length reach AI, yet in Europe the jobs where AI is used grew as its use spread.

See which jobs AI reaches
Is AI taking jobs? cover
What AI actually does

breinz Charts

Findings from breinz Labs research, one per chart.
Take any of them with you as a PDF.

AI True Cost Index

For every hour AI is expected to save, it actually saves about 25 minutes

Every hour AI is expected to save1 hourWhat actually happens to that hour25 min actually saved25 minactually saved20 min lost to failed tasks20 minlost to failed tasks15 min checking and fixing15 minchecking and fixingEvery hour AI is expected to save1 hourWhat actually happens to that hour25 min actually saved25 minactually saved20 min lost to failed tasks20 minlost to failed tasks15 min checking and fixing15 minchecking and fixingEvery hour AI is expected to save1 hourWhat actually happens to that hour25 min actually saved25 minactually saved20 min lost to failed tasks20 minlost to failed tasks15 min checking and fixing15 minchecking and fixing

Source: breinz AI True Cost Index. AI True Cost Index, average of the February 2026 and November 2025 results, rounded to whole minutes. An average conversation is expected to save about 2.9 hours, here scaled to one hour. Expected saving: time a task takes a person alone minus time with AI, from the Anthropic Economic Index; checking rate from Workday.

Where the hour goes

Of every promised hour, 24 to 26 minutes come back

February 2026February 2026, come back: 26 minutes26 mincome backFebruary 2026, lost to failed tasks: 19 minutes19 minlost to failed tasksFebruary 2026, checking and fixing: 15 minutes15 minchecking and fixingNovember 2025November 2025, come back: 24 minutes24 mincome backNovember 2025, lost to failed tasks: 22 minutes22 minlost to failed tasksNovember 2025, checking and fixing: 14 minutes14 minchecking and fixingFebruary 2026February 2026, come back: 26 minutes26 mincome backFebruary 2026, lost to failed tasks: 19 minutes19 minlost to failed tasksFebruary 2026, checking and fixing: 15 minutes15 minchecking and fixingNovember 2025November 2025, come back: 24 minutes24 mincome backNovember 2025, lost to failed tasks: 22 minutes22 minlost to failed tasksNovember 2025, checking and fixing: 14 minutes14 minchecking and fixingFebruary 2026February 2026, come back: 26 minutes26 mincome backFebruary 2026, lost to failed tasks: 19 minutes19 minlost to failed tasksFebruary 2026, checking and fixing: 15 minutes15 minchecking and fixingNovember 2025November 2025, come back: 24 minutes24 mincome backNovember 2025, lost to failed tasks: 22 minutes22 minlost to failed tasksNovember 2025, checking and fixing: 14 minutes14 minchecking and fixing

Source: breinz AI True Cost Index. AI True Cost Index, computed from the Anthropic Economic Index, releases of November 2025 and February 2026.

What people use AI for

Over half of all conversations are work, spread across three uses, and a third are personal advice

Personal advice: 32% of conversations32%Personal adviceHanding over office work: 25% of conversations25%Handing over office workWriting that goes out: 17% of conversations17%Writing that goes outBuilding software: 14% of conversations14%Building softwareLearning and study: 12% of conversations12%Learning and studyPersonal advice: 32% of conversations32%Personal adviceHanding over office work: 25% of conversations25%Handing over office workWriting that goes out: 17% of conversations17%Writing that goes outBuilding software: 14% of conversations14%Building softwareLearning and study: 12% of conversations12%Learning and studyPersonal advice: 32% of conversations32%Personal adviceHanding over office work: 25% of conversations25%Handing over office workWriting that goes out: 17% of conversations17%Writing that goes outBuilding software: 14% of conversations14%Building softwareLearning and study: 12% of conversations12%Learning and study

Source: breinz AI True Cost Index. Share of conversations by segment; segments are breinz's grouping of the Anthropic Economic Index, April and May 2026.

Return by use

Writing returns the most of its promised time, software the least

Share of promised time backWriting that goes outWriting that goes out: November 2025 41%, February 2026 50%50%41%Learning and studyLearning and study: November 2025 44%, February 2026 46%46%44%Personal advicePersonal advice: November 2025 40%, February 2026 44%44%40%Handing over office workHanding over office work: November 2025 38%, February 2026 41%41%38%Building softwareBuilding software: November 2025 33%, February 2026 35%35%33%Share of promised time backWriting that goes outWriting that goes out: November 2025 41%, February 2026 50%50%41%Learning and studyLearning and study: November 2025 44%, February 2026 46%46%44%Personal advicePersonal advice: November 2025 40%, February 2026 44%44%40%Handing over office workHanding over office work: November 2025 38%, February 2026 41%41%38%Building softwareBuilding software: November 2025 33%, February 2026 35%35%33%Share of promised time backWriting that goes outWriting that goes out: November 2025 41%, February 2026 50%50%41%Learning and studyLearning and study: November 2025 44%, February 2026 46%46%44%Personal advicePersonal advice: November 2025 40%, February 2026 44%44%40%Handing over office workHanding over office work: November 2025 38%, February 2026 41%41%38%Building softwareBuilding software: November 2025 33%, February 2026 35%35%33%
November 2025February 2026

Source: breinz AI True Cost Index. AI True Cost Index by segment; segments are breinz's grouping of the Anthropic Economic Index, November 2025 and February 2026.

Success vs time

Office work succeeds less often than writing but returns the most hours

Success rateHours back per conversationWriting that goes out84%1.8 hLearning and study77%2.2 hPersonal advice73%1.5 hHanding over office work67%2.7 hBuilding software63%0.9 hSuccess rateHours back per conversationWriting that goes out84%1.8 hLearning and study77%2.2 hPersonal advice73%1.5 hHanding over office work67%2.7 hBuilding software63%0.9 hSuccess rateHours back per conversationWriting that goes out84%1.8 hLearning and study77%2.2 hPersonal advice73%1.5 hHanding over office work67%2.7 hBuilding software63%0.9 h

Source: breinz AI True Cost Index. Success as judged by Claude and hours returned before checking, by segment. Anthropic Economic Index, February 2026.

Age

Messages from 18 to 24 year olds fell from 41% to 30% in two years

20242024, 18 to 24: 41.0%41.0%2024, 25 to 44: 48.8%48.8%2024, 45 and over: 10.2%10.2%20252025, 18 to 24: 35.5%35.5%2025, 25 to 44: 52.0%52.0%2025, 45 and over: 12.5%12.5%20262026, 18 to 24: 30.3%30.3%2026, 25 to 44: 53.9%53.9%2026, 45 and over: 15.8%15.8%20242024, 18 to 24: 41.0%41.0%2024, 25 to 44: 48.8%48.8%2024, 45 and over: 10.2%10.2%20252025, 18 to 24: 35.5%35.5%2025, 25 to 44: 52.0%52.0%2025, 45 and over: 12.5%12.5%20262026, 18 to 24: 30.3%30.3%2026, 25 to 44: 53.9%53.9%2026, 45 and over: 15.8%15.8%20242024, 18 to 24: 41.0%41.0%2024, 25 to 44: 48.8%48.8%2024, 45 and over: 10.2%10.2%20252025, 18 to 24: 35.5%35.5%2025, 25 to 44: 52.0%52.0%2025, 45 and over: 12.5%12.5%20262026, 18 to 24: 30.3%30.3%2026, 25 to 44: 53.9%53.9%2026, 45 and over: 15.8%15.8%
18 to 2425 to 4445 and over

Source: breinz AI True Cost Index. OpenAI Signals, data files version 3.0. Share of messages by age group, world, averages of July to September 2024, 2025 and 2026.

What each age asks

Seeking information rises with age, practical guidance falls

18-2425-3435-4445-5455-6465+Age groupWriting, 18-24: 21.2%Writing, 25-34: 21.9%Writing, 35-44: 21.4%Writing, 45-54: 20.4%Writing, 55-64: 18.6%Writing, 65+: 17.0%17%WritingSelf-expression, 18-24: 10.1%Self-expression, 25-34: 8.7%Self-expression, 35-44: 8.5%Self-expression, 45-54: 8.8%Self-expression, 55-64: 8.8%Self-expression, 65+: 10.9%11%Self-expressionTechnical help, 18-24: 4.4%Technical help, 25-34: 4.2%Technical help, 35-44: 4.4%Technical help, 45-54: 4.4%Technical help, 55-64: 4.1%Technical help, 65+: 3.4%3%Technical helpPractical guidance, 18-24: 31.4%Practical guidance, 25-34: 30.0%Practical guidance, 35-44: 27.6%Practical guidance, 45-54: 26.3%Practical guidance, 55-64: 26.6%Practical guidance, 65+: 27.0%27%Practical guidance31%Seeking information, 18-24: 18.1%Seeking information, 25-34: 19.8%Seeking information, 35-44: 21.2%Seeking information, 45-54: 23.3%Seeking information, 55-64: 25.2%Seeking information, 65+: 22.6%23%Seeking information18%18-2425-3435-4445-5455-6465+Age groupWriting, 18-24: 21.2%Writing, 25-34: 21.9%Writing, 35-44: 21.4%Writing, 45-54: 20.4%Writing, 55-64: 18.6%Writing, 65+: 17.0%17%WritingSelf-expression, 18-24: 10.1%Self-expression, 25-34: 8.7%Self-expression, 35-44: 8.5%Self-expression, 45-54: 8.8%Self-expression, 55-64: 8.8%Self-expression, 65+: 10.9%11%Self-expressionTechnical help, 18-24: 4.4%Technical help, 25-34: 4.2%Technical help, 35-44: 4.4%Technical help, 45-54: 4.4%Technical help, 55-64: 4.1%Technical help, 65+: 3.4%3%Technical helpPractical guidance, 18-24: 31.4%Practical guidance, 25-34: 30.0%Practical guidance, 35-44: 27.6%Practical guidance, 45-54: 26.3%Practical guidance, 55-64: 26.6%Practical guidance, 65+: 27.0%27%Practical guidance31%Seeking information, 18-24: 18.1%Seeking information, 25-34: 19.8%Seeking information, 35-44: 21.2%Seeking information, 45-54: 23.3%Seeking information, 55-64: 25.2%Seeking information, 65+: 22.6%23%Seeking information18%18-2425-3435-4445-5455-6465+Age groupWriting, 18-24: 21.2%Writing, 25-34: 21.9%Writing, 35-44: 21.4%Writing, 45-54: 20.4%Writing, 55-64: 18.6%Writing, 65+: 17.0%17%WritingSelf-expression, 18-24: 10.1%Self-expression, 25-34: 8.7%Self-expression, 35-44: 8.5%Self-expression, 45-54: 8.8%Self-expression, 55-64: 8.8%Self-expression, 65+: 10.9%11%Self-expressionTechnical help, 18-24: 4.4%Technical help, 25-34: 4.2%Technical help, 35-44: 4.4%Technical help, 45-54: 4.4%Technical help, 55-64: 4.1%Technical help, 65+: 3.4%3%Technical helpPractical guidance, 18-24: 31.4%Practical guidance, 25-34: 30.0%Practical guidance, 35-44: 27.6%Practical guidance, 45-54: 26.3%Practical guidance, 55-64: 26.6%Practical guidance, 65+: 27.0%27%Practical guidance31%Seeking information, 18-24: 18.1%Seeking information, 25-34: 19.8%Seeking information, 35-44: 21.2%Seeking information, 45-54: 23.3%Seeking information, 55-64: 25.2%Seeking information, 65+: 22.6%23%Seeking information18%

Source: breinz AI True Cost Index. OpenAI Signals, data files version 3.0. Topic mix of each age group derived from age by topic and topic shares, world, average of July to September 2026; Multimedia and Other not shown.

Markets

The US leans on personal advice, the EU and the rest of the world on software

United StatesUnited States, Personal advice: 40%40%Personal adviceUnited States, Office work: 23%23%Office workUnited States, Writing: 18%18%WritingUnited States, Software: 10%10%SoftwareUnited States, Study: 10%10%StudyUK and IrelandUK and Ireland, Personal advice: 35%35%Personal adviceUK and Ireland, Office work: 22%22%Office workUK and Ireland, Writing: 19%19%WritingUK and Ireland, Software: 10%10%SoftwareUK and Ireland, Study: 15%15%StudyEuropean UnionEuropean Union, Personal advice: 34%34%Personal adviceEuropean Union, Office work: 22%22%Office workEuropean Union, Writing: 18%18%WritingEuropean Union, Software: 14%14%SoftwareEuropean Union, Study: 13%13%StudyRest of the worldRest of the world, Personal advice: 29%29%Personal adviceRest of the world, Office work: 26%26%Office workRest of the world, Writing: 17%17%WritingRest of the world, Software: 15%15%SoftwareRest of the world, Study: 13%13%StudyUnited StatesUnited States, Personal advice: 40%40%Personal adviceUnited States, Office work: 23%23%Office workUnited States, Writing: 18%18%WritingUnited States, Software: 10%10%SoftwareUnited States, Study: 10%10%StudyUK and IrelandUK and Ireland, Personal advice: 35%35%Personal adviceUK and Ireland, Office work: 22%22%Office workUK and Ireland, Writing: 19%19%WritingUK and Ireland, Software: 10%10%SoftwareUK and Ireland, Study: 15%15%StudyEuropean UnionEuropean Union, Personal advice: 34%34%Personal adviceEuropean Union, Office work: 22%22%Office workEuropean Union, Writing: 18%18%WritingEuropean Union, Software: 14%14%SoftwareEuropean Union, Study: 13%13%StudyRest of the worldRest of the world, Personal advice: 29%29%Personal adviceRest of the world, Office work: 26%26%Office workRest of the world, Writing: 17%17%WritingRest of the world, Software: 15%15%SoftwareRest of the world, Study: 13%13%StudyUnited StatesUnited States, Personal advice: 40%40%Personal adviceUnited States, Office work: 23%23%Office workUnited States, Writing: 18%18%WritingUnited States, Software: 10%10%SoftwareUnited States, Study: 10%10%StudyUK and IrelandUK and Ireland, Personal advice: 35%35%Personal adviceUK and Ireland, Office work: 22%22%Office workUK and Ireland, Writing: 19%19%WritingUK and Ireland, Software: 10%10%SoftwareUK and Ireland, Study: 15%15%StudyEuropean UnionEuropean Union, Personal advice: 34%34%Personal adviceEuropean Union, Office work: 22%22%Office workEuropean Union, Writing: 18%18%WritingEuropean Union, Software: 14%14%SoftwareEuropean Union, Study: 13%13%StudyRest of the worldRest of the world, Personal advice: 29%29%Personal adviceRest of the world, Office work: 26%26%Office workRest of the world, Writing: 17%17%WritingRest of the world, Software: 15%15%SoftwareRest of the world, Study: 13%13%Study

Source: breinz AI True Cost Index. Segment mix by market; segments are breinz's grouping of the Anthropic Economic Index, April and May 2026.

Language

The UK and Ireland ask like Americans and work like Europeans

What people ask forSoftware requestsSoftware requests, US: 8.4%8.4%USSoftware requests, UK: 8.5%8.5%UKSoftware requests, Europe: 11.2%11.2%EuropeAdvice as the outputAdvice as the output, US: 14.4%14.4%USAdvice as the output, UK: 13.2%13.2%UKAdvice as the output, Europe: 11.4%11.4%EuropeHow people workWhole task handed overWhole task handed over, US: 31.8%31.8%USWhole task handed over, UK: 28.5%28.5%UKWhole task handed over, Europe: 28.5%28.5%EuropeStudyStudy, US: 10.8%10.8%USStudy, UK: 17.4%17.4%UKStudy, Europe: 15.3%15.3%EuropeWhat people ask forSoftware requestsSoftware requests, US: 8.4%8.4%USSoftware requests, UK: 8.5%8.5%UKSoftware requests, Europe: 11.2%11.2%EuropeAdvice as the outputAdvice as the output, US: 14.4%14.4%USAdvice as the output, UK: 13.2%13.2%UKAdvice as the output, Europe: 11.4%11.4%EuropeHow people workWhole task handed overWhole task handed over, US: 31.8%31.8%USWhole task handed over, UK: 28.5%28.5%UKWhole task handed over, Europe: 28.5%28.5%EuropeStudyStudy, US: 10.8%10.8%USStudy, UK: 17.4%17.4%UKStudy, Europe: 15.3%15.3%EuropeWhat people ask forSoftware requestsSoftware requests, US: 8.4%8.4%USSoftware requests, UK: 8.5%8.5%UKSoftware requests, Europe: 11.2%11.2%EuropeAdvice as the outputAdvice as the output, US: 14.4%14.4%USAdvice as the output, UK: 13.2%13.2%UKAdvice as the output, Europe: 11.4%11.4%EuropeHow people workWhole task handed overWhole task handed over, US: 31.8%31.8%USWhole task handed over, UK: 28.5%28.5%UKWhole task handed over, Europe: 28.5%28.5%EuropeStudyStudy, US: 10.8%10.8%USStudy, UK: 17.4%17.4%UKStudy, Europe: 15.3%15.3%Europe

Source: breinz AI True Cost Index. Four measures from the Anthropic Economic Index, April and May 2026. UK stands for the UK and Ireland; Europe for continental Europe without them.

Poland: software

Poland gives software a bigger share than the world in every release

Share of conversations about softwareNovember 2025November 2025: Poland 23%, world 15%23%15%February 2026February 2026: Poland 15%, world 11%15%11%April and May 2026April and May 2026: Poland 17%, world 14%17%14%Share of conversations about softwareNovember 2025November 2025: Poland 23%, world 15%23%15%February 2026February 2026: Poland 15%, world 11%15%11%April and May 2026April and May 2026: Poland 17%, world 14%17%14%Share of conversations about softwareNovember 2025November 2025: Poland 23%, world 15%23%15%February 2026February 2026: Poland 15%, world 11%15%11%April and May 2026April and May 2026: Poland 17%, world 14%17%14%
WorldPoland

Source: breinz AI True Cost Index. Software segment share, Poland and world; breinz's grouping of the Anthropic Economic Index, November 2025, February 2026, April and May 2026.

Poland: age

Polish messages come mostly from people aged 25 to 44

Share of messages18 to 2418 to 24: Poland 22.8%, world 30.3%22.8%30.3%25 to 4425 to 44: Poland 61.0%, world 53.9%61.0%53.9%45 to 5445 to 54: Poland 11.0%, world 9.3%11.0%9.3%55 and over55 and over: Poland 5.1%, world 6.5%5.1%6.5%Share of messages18 to 2418 to 24: Poland 22.8%, world 30.3%22.8%30.3%25 to 4425 to 44: Poland 61.0%, world 53.9%61.0%53.9%45 to 5445 to 54: Poland 11.0%, world 9.3%11.0%9.3%55 and over55 and over: Poland 5.1%, world 6.5%5.1%6.5%Share of messages18 to 2418 to 24: Poland 22.8%, world 30.3%22.8%30.3%25 to 4425 to 44: Poland 61.0%, world 53.9%61.0%53.9%45 to 5445 to 54: Poland 11.0%, world 9.3%11.0%9.3%55 and over55 and over: Poland 5.1%, world 6.5%5.1%6.5%
WorldPoland, higherPoland, lower

Source: breinz AI True Cost Index. OpenAI Signals, data files version 3.0. Share of messages by age group, Poland and world, average of July to September 2026.

AI adoption

In 2025, two in three Europeans had not started using AI, and those who had were mostly still learning to use it

27%8%43%22%Not yet startedtwo in three EU adultsStartedone in threeWhat it changes27% of new questionsLiving with the results8% of new questionsHow to use it43% of new questionsWhat AI is22% of new questions

Source: breinz Two years to first AI agent. Eurostat 2025, people aged 16 to 74 who used generative AI. Inner spheres: breinz keyword dataset, new questions recorded in 2025 by stage, excluding those that fit no stage; each question counts in one stage only.

AI adoption

In 2026, people moved on from images and detection to agents and money, while beginners kept asking what AI is

20222023202420252026first askedAsked more oftenMaking videos with AIMar 2023Making videos with AI, 2023: 10Making videos with AI, 2024: 25Making videos with AI, 2025: 61Making videos with AI, 2026: 91Whether AI will take jobsDec 2022Whether AI will take jobs, 2023: 23Whether AI will take jobs, 2024: 22Whether AI will take jobs, 2025: 53Whether AI will take jobs, 2026: 74What AI isFeb 2023What AI is, 2024: 11What AI is, 2025: 36What AI is, 2026: 72AI agents that do tasksJan 2025AI agents that do tasks, 2025: 20AI agents that do tasks, 2026: 68How to use ChatGPTDec 2022How to use ChatGPT, 2023: 20How to use ChatGPT, 2024: 17How to use ChatGPT, 2025: 41How to use ChatGPT, 2026: 60Investing in AI stocksJun 2025Investing in AI stocks, 2025: 24Investing in AI stocks, 2026: 59Asked less oftenSpotting AI-written textFeb 2023Spotting AI-written text, 2023: 11Spotting AI-written text, 2024: 38Spotting AI-written text, 2025: 72Spotting AI-written text, 2026: 66Making AI text sound humanOct 2023Making AI text sound human, 2024: 34Making AI text sound human, 2025: 79Making AI text sound human, 2026: 55Making images with AIJun 2022Making images with AI, 2023: 34Making images with AI, 2024: 60Making images with AI, 2025: 59Making images with AI, 2026: 46Bubble size: how often people asked about it

Source: breinz Two years to first AI agent. Google Trends, worldwide, yearly average of monthly interest, each phrase scaled to its own peak month (100); 2026 covers January to September. Phrases: ai video generator, ai replace jobs, what is ai, ai agent, how to use chatgpt, ai stocks, ai detector, humanize ai, ai image generator. "Asked more often" and "Asked less often" compare 2026 with 2025. First asked: the first of three months in a row at 5 or more. Yearly averages below 10 are not shown.

AI adoption

The first users ask about a new AI tool within weeks, and most people follow months or years later

202320242025Writing a cover letter with ChatGPTUsing ChatGPTSpotting AI-written textMaking AI text sound humanBuilding an AI agentWriting prompts
Tool launchedFirst users askingMost people asking

Source: breinz Two years to first AI agent. Google Trends, worldwide, each phrase scaled to its own peak month (100). First users asking: the first of three months in a row at 5 or more. Most people asking: the first of three months in a row at 25 or more. Phrases: chatgpt cover letter, how to use chatgpt, ai detector, humanize ai, how to build an ai agent, prompt engineering. Launch dates: ChatGPT, GPTZero, Turnitin AI detection, GPTs.

AI adoption

Once AI text started getting caught, people began asking how to disguise it

2023202420252026Detectors arriveLabelling and demotionLower reach and paySpotting AI-written textMaking AI text sound human

Source: breinz Two years to first AI agent. Google Trends, worldwide, both phrases compared together (ai detector, humanize ai); 100 = the peak month for ai detector. Lines show three-month averages. Events: GPTZero and Turnitin launch detectors (2023); Meta, YouTube, LinkedIn and Google label or demote AI content (2024); Pinterest and YouTube cut its reach or pay (2025).

AI adoption

AI has moved beyond its young first users, and the people starting now are older

20262026, under 35: 65%65%under 352026, 35 and over: 35%35%35 and over20252025, under 35: 71%71%under 352025, 35 and over: 29%29%35 and over20242024, under 35: 74%74%under 352024, 35 and over: 26%26%35 and over

Source: breinz Two years to first AI agent. OpenAI Signals, consumer ChatGPT, share of messages by age group, average for July to September of each year; calculated from each topic's age split and that topic's share of all messages.

Languages of AI

English and dozens of other languages share AI conversations almost evenly

English: 52.4% English 52% Russian: 11.3% Russian 11.3% All other languages: 9.7% All other languages 9.7% Chinese: 8.8% Chinese 8.8% French: 4.8% French 4.8% Vietnamese: 3.8% Vietnamese 3.8% Arabic: 2.0% Arabic 2.0% Portuguese: 1.8% Portuguese 1.8% Spanish: 1.7% Spanish 1.7% German: 1.1% German 1.1% Persian: 0.7% Korean: 0.6% Italian: 0.5% Turkish: 0.4% Japanese: 0.2% Polish: 0.2% Persian 0.7% Korean 0.6% Italian 0.5% Turkish 0.4% Japanese 0.2% Polish 0.2%

Source: breinz English is only half of AI use. Share of conversations by language, April 2023 to June 2025. The area of each piece is proportional to the share. All other languages cover more than 50 smaller ones and include Yoruba and Indonesian, likely detection errors. breinz analysis of WildChat-4.8M, 30,000 conversations.

English over time

English swung up to two thirds of conversations and back to half within two years

0%40%80%H1 2023H2 2023H1 2024H2 2024H1 2025H1 2023: 47.0%47%H2 2023: 57.1%57%H1 2024: 65.0%65%H2 2024: 51.9%52%H1 2025: 48.1%48%

Source: breinz English is only half of AI use. Share of conversations in English by half-year. breinz analysis of WildChat-4.8M, 30,000 conversations.

Languages of the web

English stays below half of the web pages AI learns from

0%40%80%Half of all pages2018201920202021202220232024202520262018: 43.9%44%2019: 43.9%44%2020: 43.2%43%2021: 45.0%45%2022: 46.6%47%2023: 45.7%46%2024: 44.3%44%2025: 43.8%44%2026: 41.4%41%

Source: breinz English is only half of AI use. Share of web pages whose primary language is English, yearly average of Common Crawl crawls, August 2018 to October 2026. breinz analysis of Common Crawl language statistics.

Language and answers

In almost every language, the AI agrees and reassures more than in English

Agrees with the person's ideas and work,reassures them without being asked, jokes and plays more, and keeps to polite language.

Corrects details nobody asked about,challenges wrong assumptions, asks for evidence, and gives frank criticism of the person's work.

HindiHindiArabicArabicThaiThaiJapaneseJapaneseKoreanKoreanIndonesianIndonesianVietnameseVietnameseSpanishSpanishFrenchFrenchTurkishTurkishRomanianRomanianPortuguesePortugueseChineseChineseDutchDutchUkrainianUkrainianItalianItalianGermanGermanPolishPolishRussianRussian
English

Source: breinz English is only half of AI use. Each bar shows how far Claude's answers in one language differ from its answers in English. English is the line in the middle, and the longer the bar, the bigger the difference. All 20 languages in the study. Calculated by breinz from Anthropic, Claude's values across models and languages, 2026, based on 309,815 conversations from May 2026.

Language and answers

In most languages, the AI sticks to the task more than in English

Admits its own mistakes,says when it is unsure, and is upfront about what it can't do.

Stays within what was askedand aims for a result, giving a polished, confident answer.

DutchDutchJapaneseJapaneseGermanGermanKoreansame as EnglishItaliansame as EnglishRussianRussianPolishPolishTurkishTurkishChineseChineseFrenchFrenchHindiHindiRomanianRomanianThaiThaiSpanishSpanishPortuguesePortugueseArabicArabicUkrainianUkrainianVietnameseVietnameseIndonesianIndonesian
English

Source: breinz English is only half of AI use. Each bar shows how far Claude's answers in one language differ from its answers in English. English is the line in the middle, and the longer the bar, the bigger the difference. All 20 languages in the study. Calculated by breinz from Anthropic, Claude's values across models and languages, 2026, based on 309,815 conversations from May 2026.

Language by country

Each country's conversations mix several languages

France, English: 47.1%France, French: 24.0%France, Other languages: 28.9%47%FranceFrench 24%Germany, English: 46.3%Germany, German: 20.9%Germany, Other languages: 32.8%46%GermanyGerman 21%Spain, English: 35.9%Spain, Spanish: 33.7%Spain, Other languages: 30.4%36%SpainSpanish 34%UK and Ireland, English: 72.2%UK and Ireland, Other languages: 27.8%72%UK and IrelandOther 28%US, English: 61.1%US, Other languages: 38.9%61%USOther 39%
EnglishFrench, German or SpanishOther languages, led by Russian and Chinese

Source: breinz English is only half of AI use. Share of conversations by language, April 2023 to June 2025. The country comes from the internet address of each conversation. breinz analysis of WildChat-4.8M.

AI at work

AI is used almost entirely for desk work

Share of all job tasksShare of AI use
Desk work42.9%93.6%Partly desk work28.6%5.5%Physical work28.5%0.1%

Source: breinz Is AI taking jobs? Share of the 18,800 O*NET job tasks and share of Claude conversations matched to them, April and May 2026. Desk work: writing, analysing, planning, calculating and communicating. Partly desk work: physical or in-person work with a real share of desk work. Physical work: no desk work. 0.8% of use could not be matched. breinz classification of 332 O*NET work activities; breinz analysis of the Anthropic Economic Index.

Use by country

A country's AI ranking depends on what is counted

Claude use per persontimes the world averageFirms using AIshare of firms with ten or more peopleFrance4.0×Netherlands3.9×US3.9×Denmark3.6×Spain2.7×Germany2.4×Italy1.7×Denmark42%Netherlands33%Germany26%Spain20%France18%Italy16%

Source: breinz Is AI taking jobs? Left: Claude use per working-age person, May 2026, as a multiple of the world average. Right: share of firms with ten or more people using at least one AI technology, 2025, for which the US has no comparable figure. France and Denmark highlighted. Anthropic Economic Index; Eurostat.

Task size

Almost half of AI use goes to tasks that take a person under four hours

17%Under 2 h31%2 to under 4 h32%4 to under 7 h21%7 h or more

Source: breinz Is AI taking jobs? Share of Claude conversations matched to O*NET tasks, grouped by the time a person would need for the task working without AI, as the model estimates it: 594, 700, 804 and 710 tasks. April and May 2026. breinz analysis of the Anthropic Economic Index.

Use per worker

Office work draws the least AI of all desk jobs

Arts, design and media9.65Computer and mathematical6.14Education2.54Sales1.22Business and finance0.89Management0.76Office and administrative support0.64Healthcare practitioners0.64Food service0.06Transport0.03Construction0.021 = use in proportion to employment

Source: breinz Is AI taking jobs? Share of US Claude conversations matched to each occupation group divided by its share of US employment, April and May 2026. Office and administrative support: two thirds of its tasks are desk work. breinz analysis of the Anthropic Economic Index and US Bureau of Labor Statistics employment, May 2025.

Job ads and AI

AI did not start the fall in job ads for the jobs that use it most: it began before ChatGPT

Before ChatGPT, 2020 to 2022After ChatGPT, 2022 to 2026
US−10%−21%Germany−14%−26%France−22%−14%

Source: breinz Is AI taking jobs? How job ads for the quarter of job types with the most AI use changed, compared with the quarter with the least. Before ChatGPT: February 2020 to November 2022. After: November 2022 to September and October 2026. Each group measured against all job ads in its country. Job types: 47 in the US, 41 in Germany, 31 in France; AI use per worker from US data. Indeed Hiring Lab; Anthropic Economic Index; breinz analysis.

Jobs and AI

The jobs AI is used in are not disappearing: in Europe they grew by 5% since 2022

Occupations that use AI+4.9%All jobs+3.0%

Source: breinz Is AI taking jobs? Change in the number of people employed in the EU, 2022 to 2025, people aged 15 to 64. Occupations that use AI: 19 occupation groups with AI use per worker at least a quarter of the average, 121 million people in 2022, such as ICT professionals, business professionals and clerks. AI use per worker from US data. Eurostat; Anthropic Economic Index; breinz analysis.

Out of view

The EU count of AI in firms leaves out four in ten jobs in Italy and a third in Spain

Italy41%Spain33%EU30%Netherlands28%France27%Denmark20%Germany19%

Source: breinz Is AI taking jobs? Share of jobs in the business economy in firms with fewer than ten people, 2023. The EU survey of AI use in firms covers only firms with ten or more people. Eurostat; breinz analysis.

Firms using AI

Half of large EU firms use AI, against one in six small firms

17%Small10 to 49 people30%Medium50 to 24955%Large250 or more20%All firms10 or more

Source: breinz Is AI taking jobs? Share of EU firms using at least one AI technology, 2025, by number of people employed. Firms with fewer than ten people, farming, mining and finance are not covered. Large firms are 0.16% of all EU firms. Eurostat.

AI True Cost Index
For every hour AI is expected to save, it actually saves about 25 minutes
Every hour AI is expected to save1 hourWhat actually happens to that hour25 min actually saved25 minactually saved20 min lost to failed tasks20 minlost to failed tasks15 min checking and fixing15 minchecking and fixing
Where the hour goes
Of every promised hour, 24 to 26 minutes come back
February 2026February 2026, come back: 26 minutes26 mincome backFebruary 2026, lost to failed tasks: 19 minutes19 minlost to failed tasksFebruary 2026, checking and fixing: 15 minutes15 minchecking and fixingNovember 2025November 2025, come back: 24 minutes24 mincome backNovember 2025, lost to failed tasks: 22 minutes22 minlost to failed tasksNovember 2025, checking and fixing: 14 minutes14 minchecking and fixing
What people use AI for
Over half of all conversations are work, spread across three uses, and a third are personal advice
Personal advice: 32% of conversations32%Personal adviceHanding over office work: 25% of conversations25%Handing over office workWriting that goes out: 17% of conversations17%Writing that goes outBuilding software: 14% of conversations14%Building softwareLearning and study: 12% of conversations12%Learning and study
Return by use
Writing returns the most of its promised time, software the least
Share of promised time backWriting that goes outWriting that goes out: November 2025 41%, February 2026 50%50%41%Learning and studyLearning and study: November 2025 44%, February 2026 46%46%44%Personal advicePersonal advice: November 2025 40%, February 2026 44%44%40%Handing over office workHanding over office work: November 2025 38%, February 2026 41%41%38%Building softwareBuilding software: November 2025 33%, February 2026 35%35%33%
Success vs time
Office work succeeds less often than writing but returns the most hours
Success rateHours back per conversationWriting that goes out84%1.8 hLearning and study77%2.2 hPersonal advice73%1.5 hHanding over office work67%2.7 hBuilding software63%0.9 h
Age
Messages from 18 to 24 year olds fell from 41% to 30% in two years
20242024, 18 to 24: 41.0%41.0%2024, 25 to 44: 48.8%48.8%2024, 45 and over: 10.2%10.2%20252025, 18 to 24: 35.5%35.5%2025, 25 to 44: 52.0%52.0%2025, 45 and over: 12.5%12.5%20262026, 18 to 24: 30.3%30.3%2026, 25 to 44: 53.9%53.9%2026, 45 and over: 15.8%15.8%
What each age asks
Seeking information rises with age, practical guidance falls
18-2425-3435-4445-5455-6465+Age groupWriting, 18-24: 21.2%Writing, 25-34: 21.9%Writing, 35-44: 21.4%Writing, 45-54: 20.4%Writing, 55-64: 18.6%Writing, 65+: 17.0%17%WritingSelf-expression, 18-24: 10.1%Self-expression, 25-34: 8.7%Self-expression, 35-44: 8.5%Self-expression, 45-54: 8.8%Self-expression, 55-64: 8.8%Self-expression, 65+: 10.9%11%Self-expressionTechnical help, 18-24: 4.4%Technical help, 25-34: 4.2%Technical help, 35-44: 4.4%Technical help, 45-54: 4.4%Technical help, 55-64: 4.1%Technical help, 65+: 3.4%3%Technical helpPractical guidance, 18-24: 31.4%Practical guidance, 25-34: 30.0%Practical guidance, 35-44: 27.6%Practical guidance, 45-54: 26.3%Practical guidance, 55-64: 26.6%Practical guidance, 65+: 27.0%27%Practical guidance31%Seeking information, 18-24: 18.1%Seeking information, 25-34: 19.8%Seeking information, 35-44: 21.2%Seeking information, 45-54: 23.3%Seeking information, 55-64: 25.2%Seeking information, 65+: 22.6%23%Seeking information18%
Markets
The US leans on personal advice, the EU and the rest of the world on software
United StatesUnited States, Personal advice: 40%40%Personal adviceUnited States, Office work: 23%23%Office workUnited States, Writing: 18%18%WritingUnited States, Software: 10%10%SoftwareUnited States, Study: 10%10%StudyUK and IrelandUK and Ireland, Personal advice: 35%35%Personal adviceUK and Ireland, Office work: 22%22%Office workUK and Ireland, Writing: 19%19%WritingUK and Ireland, Software: 10%10%SoftwareUK and Ireland, Study: 15%15%StudyEuropean UnionEuropean Union, Personal advice: 34%34%Personal adviceEuropean Union, Office work: 22%22%Office workEuropean Union, Writing: 18%18%WritingEuropean Union, Software: 14%14%SoftwareEuropean Union, Study: 13%13%StudyRest of the worldRest of the world, Personal advice: 29%29%Personal adviceRest of the world, Office work: 26%26%Office workRest of the world, Writing: 17%17%WritingRest of the world, Software: 15%15%SoftwareRest of the world, Study: 13%13%Study
Language
The UK and Ireland ask like Americans and work like Europeans
What people ask forSoftware requestsSoftware requests, US: 8.4%8.4%USSoftware requests, UK: 8.5%8.5%UKSoftware requests, Europe: 11.2%11.2%EuropeAdvice as the outputAdvice as the output, US: 14.4%14.4%USAdvice as the output, UK: 13.2%13.2%UKAdvice as the output, Europe: 11.4%11.4%EuropeHow people workWhole task handed overWhole task handed over, US: 31.8%31.8%USWhole task handed over, UK: 28.5%28.5%UKWhole task handed over, Europe: 28.5%28.5%EuropeStudyStudy, US: 10.8%10.8%USStudy, UK: 17.4%17.4%UKStudy, Europe: 15.3%15.3%Europe
Poland: software
Poland gives software a bigger share than the world in every release
Share of conversations about softwareNovember 2025November 2025: Poland 23%, world 15%23%15%February 2026February 2026: Poland 15%, world 11%15%11%April and May 2026April and May 2026: Poland 17%, world 14%17%14%
Poland: age
Polish messages come mostly from people aged 25 to 44
Share of messages18 to 2418 to 24: Poland 22.8%, world 30.3%22.8%30.3%25 to 4425 to 44: Poland 61.0%, world 53.9%61.0%53.9%45 to 5445 to 54: Poland 11.0%, world 9.3%11.0%9.3%55 and over55 and over: Poland 5.1%, world 6.5%5.1%6.5%
AI adoption
In 2025, two in three Europeans had not started using AI, and those who had were mostly still learning to use it
27%8%43%22%Not yet startedtwo in three EU adultsStartedone in threeWhat it changes27% of new questionsLiving with the results8% of new questionsHow to use it43% of new questionsWhat AI is22% of new questions
AI adoption
In 2026, people moved on from images and detection to agents and money, while beginners kept asking what AI is
20222023202420252026first askedAsked more oftenMaking videos with AIMar 2023Making videos with AI, 2023: 10Making videos with AI, 2024: 25Making videos with AI, 2025: 61Making videos with AI, 2026: 91Whether AI will take jobsDec 2022Whether AI will take jobs, 2023: 23Whether AI will take jobs, 2024: 22Whether AI will take jobs, 2025: 53Whether AI will take jobs, 2026: 74What AI isFeb 2023What AI is, 2024: 11What AI is, 2025: 36What AI is, 2026: 72AI agents that do tasksJan 2025AI agents that do tasks, 2025: 20AI agents that do tasks, 2026: 68How to use ChatGPTDec 2022How to use ChatGPT, 2023: 20How to use ChatGPT, 2024: 17How to use ChatGPT, 2025: 41How to use ChatGPT, 2026: 60Investing in AI stocksJun 2025Investing in AI stocks, 2025: 24Investing in AI stocks, 2026: 59Asked less oftenSpotting AI-written textFeb 2023Spotting AI-written text, 2023: 11Spotting AI-written text, 2024: 38Spotting AI-written text, 2025: 72Spotting AI-written text, 2026: 66Making AI text sound humanOct 2023Making AI text sound human, 2024: 34Making AI text sound human, 2025: 79Making AI text sound human, 2026: 55Making images with AIJun 2022Making images with AI, 2023: 34Making images with AI, 2024: 60Making images with AI, 2025: 59Making images with AI, 2026: 46Bubble size: how often people asked about it
AI adoption
The first users ask about a new AI tool within weeks, and most people follow months or years later
202320242025Writing a cover letter with ChatGPTUsing ChatGPTSpotting AI-written textMaking AI text sound humanBuilding an AI agentWriting prompts
AI adoption
Once AI text started getting caught, people began asking how to disguise it
2023202420252026Detectors arriveLabelling and demotionLower reach and paySpotting AI-written textMaking AI text sound human
AI adoption
AI has moved beyond its young first users, and the people starting now are older
20262026, under 35: 65%65%under 352026, 35 and over: 35%35%35 and over20252025, under 35: 71%71%under 352025, 35 and over: 29%29%35 and over20242024, under 35: 74%74%under 352024, 35 and over: 26%26%35 and over
Languages of AI
English and dozens of other languages share AI conversations almost evenly
English: 52.4% English 52% Russian: 11.3% Russian 11.3% All other languages: 9.7% All other languages 9.7% Chinese: 8.8% Chinese 8.8% French: 4.8% French 4.8% Vietnamese: 3.8% Vietnamese 3.8% Arabic: 2.0% Arabic 2.0% Portuguese: 1.8% Portuguese 1.8% Spanish: 1.7% Spanish 1.7% German: 1.1% German 1.1% Persian: 0.7% Korean: 0.6% Italian: 0.5% Turkish: 0.4% Japanese: 0.2% Polish: 0.2% Persian 0.7% Korean 0.6% Italian 0.5% Turkish 0.4% Japanese 0.2% Polish 0.2%
English over time
English swung up to two thirds of conversations and back to half within two years
0%40%80%H1 2023H2 2023H1 2024H2 2024H1 2025H1 2023: 47.0%47%H2 2023: 57.1%57%H1 2024: 65.0%65%H2 2024: 51.9%52%H1 2025: 48.1%48%
Languages of the web
English stays below half of the web pages AI learns from
0%40%80%Half of all pages2018201920202021202220232024202520262018: 43.9%44%2019: 43.9%44%2020: 43.2%43%2021: 45.0%45%2022: 46.6%47%2023: 45.7%46%2024: 44.3%44%2025: 43.8%44%2026: 41.4%41%
Language and answers
In almost every language, the AI agrees and reassures more than in English

Agrees with the person's ideas and work,reassures them without being asked, jokes and plays more, and keeps to polite language.

Corrects details nobody asked about,challenges wrong assumptions, asks for evidence, and gives frank criticism of the person's work.

HindiHindiArabicArabicThaiThaiJapaneseJapaneseKoreanKoreanIndonesianIndonesianVietnameseVietnameseSpanishSpanishFrenchFrenchTurkishTurkishRomanianRomanianPortuguesePortugueseChineseChineseDutchDutchUkrainianUkrainianItalianItalianGermanGermanPolishPolishRussianRussian
English
Language and answers
In most languages, the AI sticks to the task more than in English

Admits its own mistakes,says when it is unsure, and is upfront about what it can't do.

Stays within what was askedand aims for a result, giving a polished, confident answer.

DutchDutchJapaneseJapaneseGermanGermanKoreansame as EnglishItaliansame as EnglishRussianRussianPolishPolishTurkishTurkishChineseChineseFrenchFrenchHindiHindiRomanianRomanianThaiThaiSpanishSpanishPortuguesePortugueseArabicArabicUkrainianUkrainianVietnameseVietnameseIndonesianIndonesian
English
Language by country
Each country's conversations mix several languages
France, English: 47.1%France, French: 24.0%France, Other languages: 28.9%47%FranceFrench 24%Germany, English: 46.3%Germany, German: 20.9%Germany, Other languages: 32.8%46%GermanyGerman 21%Spain, English: 35.9%Spain, Spanish: 33.7%Spain, Other languages: 30.4%36%SpainSpanish 34%UK and Ireland, English: 72.2%UK and Ireland, Other languages: 27.8%72%UK and IrelandOther 28%US, English: 61.1%US, Other languages: 38.9%61%USOther 39%
AI at work
AI is used almost entirely for desk work
Desk work42.9%93.6%Partly desk work28.6%5.5%Physical work28.5%0.1%
Use by country
A country's AI ranking depends on what is counted
Claude use per persontimes the world averageFirms using AIshare of firms with ten or more peopleFrance4.0×Netherlands3.9×US3.9×Denmark3.6×Spain2.7×Germany2.4×Italy1.7×Denmark42%Netherlands33%Germany26%Spain20%France18%Italy16%
Task size
Almost half of AI use goes to tasks that take a person under four hours
17%Under 2 h31%2 to under 4 h32%4 to under 7 h21%7 h or more
Use per worker
Office work draws the least AI of all desk jobs
Arts, design and media9.65Computer and mathematical6.14Education2.54Sales1.22Business and finance0.89Management0.76Office and administrative support0.64Healthcare practitioners0.64Food service0.06Transport0.03Construction0.021 = use in proportion to employment
Job ads and AI
AI did not start the fall in job ads for the jobs that use it most: it began before ChatGPT
US−10%−21%Germany−14%−26%France−22%−14%
Jobs and AI
The jobs AI is used in are not disappearing: in Europe they grew by 5% since 2022
Occupations that use AI+4.9%All jobs+3.0%
Out of view
The EU count of AI in firms leaves out four in ten jobs in Italy and a third in Spain
Italy41%Spain33%EU30%Netherlands28%France27%Denmark20%Germany19%
Firms using AI
Half of large EU firms use AI, against one in six small firms
17%Small10 to 49 people30%Medium50 to 24955%Large250 or more20%All firms10 or more