One in five British adults now uses AI news tools or services to keep up with current affairs, according to Ofcom’s latest news-consumption research. The same study found a sharp trust gap: 56% of adults said they trust news written by AI less than news written by a person.
AI news is becoming another gateway to journalism
Ofcom found that 21% of UK adults had used an AI app or service to access news in the previous month. ChatGPT was the most widely used individual AI service for that purpose, reaching 13%. Use was concentrated among younger adults and people in more affluent households, but the overall figure is already large enough to make AI an important part of the route between publishers and readers.
This does not mean one in five adults has abandoned newspapers, television or news websites for a chatbot. Ofcom’s research shows people assemble news from many places. Adults used around 12 news sources on average, rising to 18 among people aged 25 to 34. AI joins that mixture as an intermediary that can answer a question, summarise events or surface material from elsewhere.
The distinction between publisher and intermediary is increasingly important. Ofcom says 73% of adults access news through online intermediaries such as social media, video platforms, search engines, aggregators and AI tools, while 57% access news directly through publishers’ websites or apps. Those categories can overlap because the same person can use both.
Ofcom’s full 2026 news-consumption research places that AI figure inside a wider shift towards online intermediaries. The significance is not that AI has replaced publishers, but that another layer can now select, compress and reframe reporting before a reader reaches the original source.
Trust has not kept pace with use
The most striking tension in the findings is that adoption and confidence are moving on different tracks. Fifty-six per cent of adults said they trust news written by AI less than news written by a person. Ofcom’s survey does not prove that a particular AI summary is less accurate, or that human-written news is automatically reliable. It measures public attitudes towards authorship.
That scepticism may be rational in part because AI can compress several sources into one fluent answer while making provenance harder to see. A reader may receive a confident paragraph without immediately knowing which claims came from an official statement, which came from a news report and which were inferred by the model. Good news products have to make that chain visible rather than relying on fluency as a substitute for sourcing.
LiveAIWire has separately examined the human and editorial pressures created by generative AI in journalism and why AI answers need testing beyond a headline hallucination rate. Both issues become more important when an intermediary summarises evidence before the reader sees the underlying reporting.
Disclosure by itself may not solve every trust problem. LiveAIWire recently covered a preregistered experiment in which telling people a chatbot was AI did not make it less persuasive. The effect changed only when users were told about the system’s persuasive purpose. For news, the parallel question is whether a simple AI label gives readers enough information to judge the evidence behind an answer.
The internet has already overtaken television as a route to news
Ofcom says 82% of adults accessed news online during the previous month, compared with 79% through television. It estimates adults spend about 72 minutes a day consuming news, with 34 minutes online and 18 minutes on television. Television remains particularly important for older audiences, but the centre of gravity has moved towards digital distribution.
Social and video platforms are a major part of that change. Sixty per cent of adults use them as a route to news, rising to 78% among 16 to 24 year olds and children aged 12 to 15. AI does not arrive in a world where everybody visits a publisher homepage. It arrives after years in which algorithms have already become powerful selectors of what people encounter.
That makes the design of AI news experiences consequential. A chatbot can answer the exact question a reader asks, which is useful, but it can also narrow the field of view. A front page exposes a reader to stories they did not know to request. A personalised answer may be efficient while offering fewer accidental encounters with unfamiliar subjects.
Publishers now have to optimise for being understood by machines and people
For publishers, the growth of AI intermediaries creates a practical challenge. Their reporting needs to be clear enough for readers, but also structured and sourced so automated systems can identify what was actually reported. Precise headlines, unambiguous dates, direct evidence links and clear attribution are not merely search-engine techniques. They reduce the chance that a summary detaches a claim from its context.
There is also a commercial question. If a reader gets the useful part of a story from an AI response, the publisher may receive no visit even when its reporting supplied the underlying information. The long-term relationship between AI services and news organisations will therefore depend on licensing, referral traffic, attribution and product design as much as on model accuracy.
Trust will matter on both sides. LiveAIWire has also reported that warnings can lower trust in an AI without reliably reducing its influence. People can be sceptical of a system and still let it shape their judgement. News literacy in an AI environment will require more than asking whether somebody says they trust the tool.
Ofcom’s figures are a snapshot, not a straight trend line
Ofcom notes that its news survey methodology changed this year. That means tempting year-on-year comparisons need caution. The 21% figure is best read as a current estimate of reported use under the 2026 survey design, not as proof that AI news consumption rose by a particular number of percentage points from an earlier measure.
The study also asks whether people used AI to access news, which can include several behaviours. Somebody might ask a chatbot for a summary of a developing event, use an AI feature embedded in another service or encounter an AI-generated answer while looking for information. Those experiences are not identical, and future research will need to separate them more precisely.
What is already clear is that AI has moved beyond being merely a subject of the news. It is becoming part of the news distribution system itself. One in five adults is enough to affect how publishers think about reach, attribution and trust, even while most news consumption still happens through other routes.
For readers, the sensible habit remains familiar: treat an AI summary as a route into evidence rather than the final authority. Check the named source for consequential claims, notice when a figure lacks a date or denominator and distinguish an official announcement from analysis. AI can make information faster to reach. It does not remove the need to know where that information came from.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.