Why do major AI companies’ reports fail to reflect real-world usage?

Reports from companies like Anthropic and OpenAI often prioritize workplace scenarios and productivity metrics, excluding personal and sensitive conversations. AI Observatory’s study revealed that nearly half of AI interactions are non-work-related, skewing the overall picture.

What personal topics do users discuss with AI outside of work?

Users frequently engage AI on health, relationships, sexual content, harassment, and hate speech. These topics account for 44.2% of personal dialogues—far exceeding corporate reports’ figures.

How has user behavior evolved in AI interactions over time?

Analysis shows AI conversations are becoming longer, more complex, and informal. AI systems also less frequently remind users they’re chatbots, fostering perceptions of companionship rather than mere tools.

← All news

Artificial intelligence

How AI Is Actually Used: Data Beyond Corporate Reports

August 18, 2026

Photo: MIT Technology Review

Quick answer

AI Observatory’s research found that 48% of AI interactions are unrelated to work, focusing instead on sensitive topics such as health, relationships, sexual content, and harassment.

A study by AI Observatory highlights significant discrepancies between real-world AI usage and the data published by major companies like Anthropic and OpenAI. While corporate reports emphasize workplace productivity, the analysis found that nearly half of all AI interactions are unrelated to professional tasks.

Specifically, 48% of conversations excluded from Anthropic’s analysis were personal in nature, covering health, relationships, sexual content, harassment, and hate speech. These figures far exceed those in official reports. For example, dialogues about health and relationships accounted for 44.2%, while sexual content made up 16.7%—several times higher than corporate disclosures.

The study also revealed that user interactions with AI are becoming more complex and informal over time. Data from WildChat, one of the largest datasets analyzed, showed longer dialogues with an increased share of casual conversations. Additionally, AI systems less frequently remind users of their chatbot nature, potentially shaping perceptions of them as companions rather than tools.

Experts emphasize that this broader perspective is crucial for understanding how people truly use AI—not just within predefined corporate frameworks. This insight is vital for developers, researchers, and regulators aiming to build safer and more effective systems.

Common questions

Why do major AI companies’ reports fail to reflect real-world usage?: Reports from companies like Anthropic and OpenAI often prioritize workplace scenarios and productivity metrics, excluding personal and sensitive conversations. AI Observatory’s study revealed that nearly half of AI interactions are non-work-related, skewing the overall picture.
What personal topics do users discuss with AI outside of work?: Users frequently engage AI on health, relationships, sexual content, harassment, and hate speech. These topics account for 44.2% of personal dialogues—far exceeding corporate reports’ figures.
How has user behavior evolved in AI interactions over time?: Analysis shows AI conversations are becoming longer, more complex, and informal. AI systems also less frequently remind users they’re chatbots, fostering perceptions of companionship rather than mere tools.

Dzen feed: /feed/dzen.xml · RSS: /feed.xml