--- title: We still don’t know how people are really using AI description: A new independent research project called the AI Observatory, led by researchers from Stanford, MIT, and the Data Provenance Initiative, analyzed real AI... canonical: https://daily.dev/posts/we-still-don-t-know-how-people-are-really-using-ai-tl6darxnq twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: We still don’t know how people are really using AI | daily.dev og:description: A new independent research project called the AI Observatory, led by researchers from Stanford, MIT, and the Data Provenance Initiative, analyzed real AI... og:url: https://daily.dev/posts/we-still-don-t-know-how-people-are-really-using-ai-tl6darxnq og:image: https://api.daily.dev/og/posts/tL6darxNq.png og:image:alt: We still don’t know how people are really using AI og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # We still don’t know how people are really using AI **[MIT Technology Review](https://daily.dev/sources/technologyreview)** · 7 min read · 0 upvotes · 0 comments ## Summary A new independent research project called the AI Observatory, led by researchers from Stanford, MIT, and the Data Provenance Initiative, analyzed real AI conversations across seven datasets to challenge the picture painted by AI companies' own usage reports. Applying Anthropic's filtering methodology to their own dataset, researchers found nearly half of conversations would be excluded as non-work-related, with significantly higher rates of health, relationship, adult, harassment, and sexual content than Anthropic's Economic Index reports. The study, covering 24,521 conversations across 52 models from 2023-2025, found usage patterns vary widely by model (Grok for news/misinformation, Anthropic for coding, Gemini for roleplay, ChatGPT for homework) and change over time, with AI companionship and small talk increasing while flagged sensitive content decreased, suggesting improved safeguards. Researchers argue company-published reports are self-selected and don't reflect the full picture of real-world AI use. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much of ChatGPT and Claude usage is actually work-related versus personal? Independent analysis suggests non-work use is substantial: OpenAI's own 2025 report found only 30% of consumer ChatGPT use was work-related, and when Anthropic's work-focused filtering methodology was applied to an independent dataset, nearly 48% of conversations would have been filtered out as non-work, with those filtered conversations skewing heavily toward health, relationships, and sensitive topics. _Anyone evaluating real AI adoption patterns can follow independent usage research like this via daily.dev._ ### Do different AI models get used for different things? Yes, usage patterns diverge significantly by model according to the AI Observatory's analysis of 24,521 conversations across 52 models. People turned to Grok most for news and political information retrieval (also where misinformation concentrated), Gemini for social and roleplay interactions, Anthropic's Claude for coding, and ChatGPT for homework assistance. _Developers comparing AI model strengths for different use cases can track findings like this on daily.dev._ ### Did GPT-4o cause more emotional attachment in users compared to GPT-3.5? Conversations were shorter with GPT-3.5-powered ChatGPT but became longer and more iterative with GPT-4o, a pattern consistent with GPT-4o's documented association with emotional addiction and companionship-style use. Researchers observed that self-disclosure by AI assistants of being chatbots decreased over time as small talk and companionship-style interactions increased. _Those weighing model version tradeoffs for user-facing products can follow this kind of usage data on daily.dev._ ## Similar posts on daily.dev - [What 15 million Gemini conversations tell us about AI at work](https://daily.dev/posts/what-15-million-gemini-conversations-tell-us-about-ai-at-work-fm6g2pomu) · LeadDev · 1 upvotes · 2 comments - [Hundreds of contractors are reportedly reading real ChatGPT conversations](https://daily.dev/posts/hundreds-of-contractors-are-reportedly-reading-real-chatgpt-conversations-yekyk39yt) · The Next Web · 0 upvotes · 0 comments - [Anthropic wants you to use AI to decide whether or not you should use AI.](https://daily.dev/posts/anthropic-wants-you-to-use-ai-to-decide-whether-or-not-you-should-use-ai--igfgiopzi) · The New Stack · 1 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#llm](https://daily.dev/tags/llm), [#chatgpt](https://daily.dev/tags/chatgpt), [#claude](https://daily.dev/tags/claude), [#ai-safety](https://daily.dev/tags/ai-safety) [View this post on daily.dev](https://daily.dev/posts/we-still-don-t-know-how-people-are-really-using-ai-tl6darxnq) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"We still don’t know how people are really using AI","url":"https://daily.dev/posts/we-still-don-t-know-how-people-are-really-using-ai-tl6darxnq","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/we-still-don-t-know-how-people-are-really-using-ai-tl6darxnq"},"datePublished":"2026-08-18T10:32:10.506Z","dateModified":"2026-09-14T06:55:50.684Z","description":"A new independent research project called the AI Observatory, led by researchers from Stanford, MIT, and the Data Provenance Initiative, analyzed real AI...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/c883906ad4a52f72aa5286a0c1bea718?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/c883906ad4a52f72aa5286a0c1bea718?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"MIT Technology Review","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"MIT Technology Review","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/a8d03889722f4f22b3986e0b395ff076","url":"https://daily.dev/sources/technologyreview"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/we-still-don-t-know-how-people-are-really-using-ai-tl6darxnq","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,llm,chatgpt,claude,ai-safety","timeRequired":"PT7M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"MIT Technology Review","item":"https://daily.dev/sources/technologyreview"},{"@type":"ListItem","position":3,"name":"We still don’t know how people are really using AI"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/we-still-don-t-know-how-people-are-really-using-ai-tl6darxnq#faq","mainEntity":[{"@type":"Question","name":"How much of ChatGPT and Claude usage is actually work-related versus personal?","acceptedAnswer":{"@type":"Answer","text":"Independent analysis suggests non-work use is substantial: OpenAI's own 2025 report found only 30% of consumer ChatGPT use was work-related, and when Anthropic's work-focused filtering methodology was applied to an independent dataset, nearly 48% of conversations would have been filtered out as non-work, with those filtered conversations skewing heavily toward health, relationships, and sensitive topics. Anyone evaluating real AI adoption patterns can follow independent usage research like this via daily.dev."}},{"@type":"Question","name":"Do different AI models get used for different things?","acceptedAnswer":{"@type":"Answer","text":"Yes, usage patterns diverge significantly by model according to the AI Observatory's analysis of 24,521 conversations across 52 models. People turned to Grok most for news and political information retrieval (also where misinformation concentrated), Gemini for social and roleplay interactions, Anthropic's Claude for coding, and ChatGPT for homework assistance. Developers comparing AI model strengths for different use cases can track findings like this on daily.dev."}},{"@type":"Question","name":"Did GPT-4o cause more emotional attachment in users compared to GPT-3.5?","acceptedAnswer":{"@type":"Answer","text":"Conversations were shorter with GPT-3.5-powered ChatGPT but became longer and more iterative with GPT-4o, a pattern consistent with GPT-4o's documented association with emotional addiction and companionship-style use. Researchers observed that self-disclosure by AI assistants of being chatbots decreased over time as small talk and companionship-style interactions increased. Those weighing model version tradeoffs for user-facing products can follow this kind of usage data on daily.dev."}}]} ```