70/100

Moderate evidence

Cite
More
EmailLinkedInX

German experiment finds readers trust AI-written news outlets less and accept fewer ads

Study title: AI in the Newsroom: Does the Public Trust Automated Journalism and Will They Pay for It?

Evidence type
Randomized controlled trial
Sample
n = 1,261
Sources
1 sourcePeer-reviewed
Published
2 min read

Study title: AI in the Newsroom: Does the Public Trust Automated Journalism and Will They Pay for It?

A pre-registered online experiment in Germany with 1,261 respondents found that a fictional news outlet was trusted less when it was described as using generative AI instead of trained journalists. The drop was larger for political coverage than for entertainment coverage.

Read the study abstract

Artificial intelligence (AI) is increasingly being adapted for news production tasks in news outlets worldwide. Whether the public trusts news outlets relying on AI to produce news remains unclear. Further, the economic implications (willingness to pay and ad acceptance) and moderators remain unexplored. We conducted a pre-registered 2 × 2 experiment with a quota-based nonprobability sample from Germany (N = 1,261). Respondents were exposed to vignettes about a fictional online news outlet. We manipulated the pr…

The problem

In a pre-registered experiment with 1,261 German respondents, outlets described as producing news with generative AI were trusted less than outlets using trained journalists, especially on political news, and people were less willing to accept ads on them.

In brief

Respondents were also less willing to accept ads on AI-produced outlets, and on outlets covering entertainment. Willingness to pay for the news was not affected by whether AI or journalists produced it.

The authors conclude that AI in the newsroom may reduce outlets' trust and ad revenue, and that the effect may grow depending on audience characteristics. The result comes from vignettes about a fictional outlet, not from real news products.

Main points

  1. The study was a pre-registered 2 × 2 experiment using a quota-based nonprobability sample from Germany (N = 1,261).
  2. Participants read vignettes about a fictional online news outlet.
  3. The production process (generative AI vs. trained journalists) varied between subjects; the topic (politics vs. entertainment) varied within subjects.
  4. Trust in outlets was lower for AI-generated news, particularly political outlets.
  5. Respondents were less willing to accept ads on AI outlets and on outlets covering entertainment.
  6. Willingness to pay was not affected by AI production.
  7. People who believed AIs spread incorrect information trusted AI-generated news less; that belief did not change willingness to pay or ad acceptance.

The full story

Researchers randomly showed German respondents vignettes about a fictional online news outlet that either used generative AI or trained journalists. Every respondent also judged both a politics outlet and an entertainment outlet, allowing comparison across topics.

The outcomes were trust in the outlet, willingness to accept ads and willingness to pay. Trust fell for AI-produced news, with the biggest fall for political content, and ad acceptance was lower for AI outlets and for entertainment outlets. Willingness to pay did not differ.

The team also tested moderators. Respondents who thought AIs spread incorrect information trusted AI-generated news less, but this perception did not moderate willingness to pay or ad acceptance. The opinion climate did not moderate the effect of how content was produced.

Sources

1 on file
  1. Journalism Studiesdoi:10.1080/1461670x.2025.2547301
    Peer-reviewed2025-08-26

Talk it over

Three questions to start on, two more if you have time
In the experiment with 1,261 German respondents, trust fell more for political news than for entertainment news when outlets used generative AI. Why might readers treat the two topics differently?

Start with a question

German experiment finds readers trust AI-written news outlets less and accept fewer adshttps://truvace.com/item/ai-in-the-newsroom-does-the-public-trust-automated-journalism-and-will-they-pay-

  1. In the experiment with 1,261 German respondents, trust fell more for political news than for entertainment news when outlets used generative AI. Why might readers treat the two topics differently?

    Answer in the thread
  2. Respondents read vignettes about a fictional online news outlet, and the sample came only from Germany. How far could that setup tell us about how people treat real news outlets?

    Answer in the thread
  3. Willingness to pay did not change, but trust and ad acceptance did. What finding from real news outlets would change your mind about what AI-written news costs them?

    Answer in the thread
If you have more time
  • Respondents who believed AIs spread incorrect information trusted AI-generated news less. How would you test whether that belief drives trust, or whether something else does?

    Answer in the thread
  • Think of a news outlet you know that discloses how its stories are made. What would you need to see to judge whether readers there would react as the German respondents did?

    Answer in the thread

What readers say