Exploring the Integration of Language Models in Everyday Life: Insights from US Adults

According to a recent study conducted by the Center for Imagining the Digital Future, language models have achieved unprecedented levels of adoption among adults in the United States. This technology has surpassed traditional barriers to entry, showing an even distribution of use across different demographic groups—women now engage with these tools as frequently as men, irrespective of education or income levels.

ChatGPT leads the market, being favored by 72% of users on the OpenAI platform. Following closely is Google’s Gemini, with 50% of users preferring it, and Microsoft’s Copilot, which holds 39%. Other notable competitors include Meta’s LLaMa at 20%, xAI’s Grok at 12%, and Anthropic’s Claude at 9%. This aligns with usage figures reported by Similarweb.

Most users engage with multiple AI platforms, with 58% having tried two or more language models. However, only a small number take advantage of premium features; while 20% have access to paid versions, just 4% personally pay for these subscriptions.

The intersection between language models and AI image generators is significant. Two-thirds of language model users (67%) have experimented with image creation tools like DALL-E, Midjourney, Adobe Firefly, or ImageFX. A notable portion are regular users: 18% generate images using AI daily, and 12% do so several times a week.

The study revealed that 51% of users primarily utilize language models for personal learning and planning, with only 24% using them mainly for work-related tasks.

Professional use varies greatly by field. Leaders, academics, and licensed professionals, such as teachers, report high levels of engagement. Surprisingly, retail and service workers demonstrate a notable level of involvement at 62%.

Demographically, families with children under 18 are more likely to use these tools (61%) compared to households without children (50%).

Nearly half of the respondents (49%) believe that language models are more intelligent than they are. There is a noticeable gender gap in this perception: 30% of female users think that language models are «significantly smarter» than them, while only 20% of male users feel the same.

Voice interaction has emerged as a major trend, with 65% of users conversing with AI systems that respond with realistic voices. This practice is particularly common among low-income families (76%) and Black users (83%).

While 76% of participants express satisfaction with language models, the study uncovered serious concerns. Almost a quarter of respondents (23%) report making significant errors due to incorrect information provided by the models, and 21% feel they are being manipulated by these systems.

Approximately half of those surveyed feel that language models simplify tasks too much or encourage laziness, while more than a third express feeling overly dependent on this technology. A similar proportion considers the use of language models to be somewhat deceptive or report discomfort with these interactions.

The research also indicates that language models currently influence important life decisions across various domains. Over four in ten users (41%) consult language models for career advancement decisions, while 37% turn to them for health-related inquiries.

Job transitions are another area where language model systems are frequently sought for assistance, with 28% of users relying on them when changing careers. Financial decisions also showcase significant reliance on language model systems, as 25% of users trust these systems for important monetary matters. In housing choices, users seek guidance from language model systems as well (18%).

This study was conducted by SSRS using its Opinion Panel platform from January 21 to 23, 2025. A total of 500 adults with experience using language models participated (498 through online interviews, 2 through phone interviews; 473 in English, 27 in Spanish). Researchers analyzed responses from 939 participants to assess the overall prevalence of language model usage among adult residents of the United States.