aimodelnews
A person seen from behind at a kitchen table in morning light, holding a smartphone beside a mug of tea and a folded newspaper
AI-generated image, not a photograph of a real person or place.

WHAT THE RESEARCH FOUND

Check the answer,
then the source.

AI assistants now answer questions about the news.

Public service broadcasters in 18 countries put them to the test. Their journalists found that many answers sounded confident but had problems with sources, dates or quotes, things a reader can often check in a minute.

45% had a significant issue

Check an answer

THE SHORT ANSWER

Useful, but not
yet reliable.

Two studies led by the BBC asked AI assistants real news questions.
Journalists who knew each topic marked the answers.

SOURCES

Follow the link

Sourcing was the biggest problem: 31% of answers had significant sourcing issues, such as claims the cited source did not support, or no source at all.

What to check

DATES

Is it still true?

Out-of-date information was one of the most common errors. In May 2025, assistants still named Pope Francis as Pope, weeks after his death.

See the numbers

QUOTES

Compare the words

In the BBC's first study, 13% of quotes taken from BBC articles were either changed or not in the article cited at all.

How to check
A stack of old folded newspapers tied with string on a wooden shelf, their print too blurred to read
AI-generated image, not a photograph of a real person or place.

FOUR QUICK CHECKS

Treat it as a lead,
not the last word.

These checks follow the four criteria the broadcasters used to judge answers: accuracy, context, fact versus opinion, and sourcing.

  1. Open the cited source

    Click through and find the claim on the page. If there is no source, or the page does not say it, treat the claim as unconfirmed.

  2. Check the date

    Look at when the source was published. Outdated information was one of the most common errors the journalists found.

  3. Separate fact from opinion

    Ask whose view a statement is. A good answer says clearly who holds an opinion rather than presenting it as fact.

  4. Go to a trusted outlet

    For anything important, read the story from a news organisation or official source that corrects its mistakes in public.

TWO NUMBERS WORTH KNOWING

Confident answers,
frequent slips.

One figure on how often answers went wrong,
one on how many people already use AI for news.

45%

AI news answers with a significant issue

Journalists from 22 public service media organisations in 18 countries, working in 14 languages, evaluated more than 3,000 responses from four AI assistants. When smaller issues were included, 81% of responses had a problem of some form.

ACCURACY · BBC AND EBU

15%

Under-25s using AI chatbots for news weekly

Across 48 markets, 7% of people said they use AI chatbots for news each week, rising to 15% of under-25s. In the UK, weekly use was just 3%.

USE · REUTERS INSTITUTE

The BBC's first study, of 100 news questions asked in December 2024, found significant issues in 51% of answers. The October 2025 round showed some improvement but still high levels of errors.

WHERE ANSWERS GO WRONG

What should you
check first?

Pick a part of an AI news answer to see how often journalists found a significant problem with it. The bar shows how common the problem was in the international study.

Answer check

Choose what to check
The sources31% had issues

The most common significant problem: claims not supported by the cited source, no source given, or sourcing claims that could not be verified. Open the link and look for the claim yourself.

Names, numbers, dates20% had issues

One in five answers had significant accuracy problems, including invented details and outdated information. Check the key fact against a recent report.

The context14% had issues

Some answers left out context or perspectives a non-expert reader would need to understand the story. A short answer to a complicated story deserves a second look.

Direct quotes12% had issues

Of the answers that contained a direct quote, 12% had significant problems with the accuracy of that quote. Compare the words with the original article.

Source: BBC and EBU, News Integrity in AI Assistants report (October 2025), share of all responses with significant issues, and of responses with direct quotes for the quotes figure.

Figures from the BBC and EBU News Integrity in AI Assistants report (October 2025). Answers change over time and between assistants.

Hands on a laptop keyboard next to an open notebook and pen, the screen softly blurred, as someone checks a source
AI-generated image, not a photograph of a real person or place.

WHO FIXES THE MISTAKES?

News outlets correct.
Assistants may not.

The BBC noted that while AI assistants often carry a disclaimer about possible inaccuracy, there is no mechanism for them to correct errors, unlike professional news outlets that acknowledge and correct their mistakes.

When people want to check whether something is true, the Reuters Institute found they most often turn first to news outlets they trust (38%), official sources (35%) and fact-checkers (25%), rather than social media (14%).

Questions

Questions people ask

How accurate are AI assistants about the news?

In the October 2025 BBC and EBU study, 45% of more than 3,000 answers had at least one significant issue, and 81% had some kind of issue. The study covered four leading AI assistants in 14 languages.

What was the most common problem?

Sourcing. 31% of answers had significant sourcing problems, such as a cited source that did not support the claim, missing sources, or incorrect attributions. Accuracy issues affected 20%.

Do the assistants admit when they do not know?

Rarely. Of 3,113 questions asked in the international study, only 17 were refused, about 0.5%. The researchers say assistants seem more willing to answer whether or not they can answer well.

How many people get news from AI?

The Reuters Institute Digital News Report 2025 found 7% use AI chatbots for news each week across 48 markets, and 15% of under-25s. In the UK the figure was 3%.

Are people comfortable with AI-made news?

Mostly not. The Reuters Institute found audiences more comfortable when humans stay in the loop, and expect AI to make news cheaper to make but less trustworthy. In the UK, comfort with mainly AI-made news was 11%.

Sources

  1. BBC Media Centre: AI assistants misrepresent news content 45% of the time (October 2025)
  2. BBC and EBU: News Integrity in AI Assistants report (2025)
  3. BBC and EBU: News Integrity in AI Assistants toolkit (2025)
  4. BBC: Representation of BBC News content in AI assistants (February 2025)
  5. Reuters Institute: Digital News Report 2025, executive summary

Updated

ABOUT THIS NAME

Interested in leasing aimodelnews.com?

aimodelnews.com is part of the DoodleSale portfolio. It isn't in an auction right now, but you can ask about leasing it.

Ask about this name