NGO report Credible

Large Language Models in the UK: Public Use, Trust, and Attitudes

Survey report on how UK adults use large language models, how much they trust them across domains, and their attitudes toward them, based on 2,002 respondents recruited through Prolific in December 2025 and matched to the UK population on age, gender and region. Almost all respondents had used an LLM, a fifth had used an AI companion platform at least once, and a substantial minority of regular users turn to LLMs for personal advice, deep conversation or companionship. The report also measures trust in LLM advice on personal relationships and other domains and the extent to which interactions feel human-like.

Read at oii.ox.ac.uk Archived copy

Publisher

Oxford Internet Institute, University of Oxford

Published

9 Jul 2026

Added

1 month ago

DOI

10.5287/ora-pr556xxyx(opens in a new tab)

Key Findings

92.7% of respondents had used an LLM at least once
20.2% of the whole sample had used an AI companion (character-based platforms such as Replika or Character.AI) at least once; among companion users, 7.6% use one daily, 20.8% several times a week and 15.6% about once a week
Among regular LLM users, 31% use them for personal advice very often or occasionally, 24.9% for deep and meaningful conversation and 14% for role-play and companionship
38% of respondents report trusting LLMs at least somewhat on matters relating to personal relationships
Survey data are openly available via the Open Science Framework (osf.io/agyfb)

Methodology Notes

PDF dated July 2026 (creation 2026-07-02; OII news item 2026-07-09, used as the publication date). n = 2,002 UK adults, online, December 2025, recruited through Prolific's representative-sample tool matched on age, gender and region. The executive summary describes the sample as nationally representative while the limitations section states that Prolific recruitment is not probability-based and the results should not be treated as fully representative of the UK population; both statements are carried here. Curator read the PDF (34 pages).

Sources

Report PDF (Oxford Internet Institute)(opens in a new tab) (primary)

Open data (OSF)(opens in a new tab)

arXiv preprint version (v1, 23 September 2026)(opens in a new tab) (23 Sept 2026)

Oxford University Research Archive record (DataCite DOI 10.5287/ora-pr556xxyx)(opens in a new tab) (20 Jul 2026)

Archived snapshot (Wayback Machine)(opens in a new tab) — preserved against link rot

Topics

AI companionship Human-AI relationships Industry landscape high stakes guidance Dependency & parasocial attachment

Authors

Florence E. Enock, Helen Z. Margetts

Cite This

APA

Florence E. Enock, Helen Z. Margetts. (2026). Large Language Models in the UK: Public Use, Trust, and Attitudes. Oxford Internet Institute, University of Oxford. https://www.oii.ox.ac.uk/wp-content/uploads/2026/07/Large-language-models-in-the-UK-public-use-trust-and-attitudes_FINAL.pdf