Personal experience with general-purpose LLMs is mixed. In terms of frequency, people reported using LLMs a few times rather than regularly. The most popular use is searching for answers and recommendations, with a third (33%) of the UK public indicating they have used these technologies at least a few times. This is followed by educational purposes (21%) and everyday tasks such as writing emails (21%). Two-fifths (40%) of the UK public have used LLMs for one or more of the tasks we asked about.
When compared with existing research, these figures suggest an upward trend in the use of LLMs. For example, a 2024 survey across six countries including the UK found that on average 24% of people used generative AI tools for getting information, 9% used it for writing emails and 8% used it for educational purposes. The usage figures in our survey may even be a conservative estimate as it is possible that some people in our sample have used AI without being aware of it, due to existing integration of AI in some search engines. As these tools become more integrated in search engines, we can expect usage to increase.
As opposed to everyday tasks, Figure 2 shows that few people have used general-purpose LLMs for entertainment purposes (14%), supporting job applications (11%) or guidance on issues such as legal disputes or taxation (8%). A considerable proportion of the UK public are also closed off to using LLM-based AI tools for some of the applications we presented. This was most prevalent for supporting job applications, where 39% of people would not want to use LLMs for this.
Those with fewer digital skills and those on lower incomes are slightly more likely to be closed off to the use of LLMs for all the tasks we asked about than those with higher levels of digital skills and those on higher incomes, with this difference being statistically significant. For example, of those not open to using general-purpose LLMs for supporting job applications, 27% do not have basic digital skills and 39% have low incomes (equivalised monthly household income of £1,500 or less). This is in contrast with those that have used LLMs for supporting job applications, or are open to using them for this, where only 16% do not have basic digital skills and 33% have low incomes.
This limited adoption might be related to a range of reasons. First, it might be indicative of an apprehension towards using a general tool for a specialised task, such as getting legal guidance, suggesting personal red lines in terms of in which context AI tools are deemed appropriate. People may feel, for example, that some tasks require human expertise. Second, it may relate to concerns around access and opportunity (for low-income or digitally excluded groups), with some feeling apprehensive about the role of emerging technologies in domains such as legal advice or job applications and/or their ability to use these tools. While our work offers important preliminary insights into public experiences with general-purpose LLMs, we are unable to unpack reasons for limited adoption with our data due to limitations of survey length. Future research should track public experiences with such emerging technologies in more detail.
