Transparency
How we use AI at lowbrowse
Yes, we use AI, and we also use spell checkers, calculators and other tools. AI is a tool too, a very powerful one, but still a tool. It is very good at some tasks and much less good at others. We use it for the things it does well and keep it away from the things it doesn't.
For example, AI is great at drafting content based on structured input. That is why we invest heavily in connecting to primary sources, databases and official wires. Once we have a solid record of an event and its claims, we let AI draft our news, alert and policy cards.
Before you see any of them, a human checks each card against its sources and approves it. Insight cards work differently. They are written from start to finish by people outside lowbrowse, and where the author allows it, AI translates them into our other languages.
How a news card gets made
Every news card starts with an article from an official news wire whose content we license. We don't scrape websites or repackage other outlets' content. Everything we start from has already been through a professional newsroom before it reaches you.
AI tools pull in the previous day's articles and group together the ones that cover the same story. Claims are then pulled from the original articles and verified, and AI writes the news card in plain language, taking out sensational or loaded wording and any opinion, adding context from primary sources and putting it into the lowbrowse style. Before the card goes any further, a program checks that every quote on it appears word for word in the original article. Then a person reads the card, changes the wording where it needs it and decides whether it goes into the edition.
Alert and policy cards
Alert and policy cards don't start from news articles at all. They're built from primary sources, such as official alert data or parliamentary records, and AI turns that material into a card. A person then checks each claim against the source it came from and only approves the card once everything matches.
Insight cards
Insight cards are written and analysed entirely by people. The only thing AI does here is help with translation, so that the same card can appear in our other languages.
Where people come in
Nothing is published until a person has approved it. That person decides which stories make it into the edition, fixes the wording wherever the draft gets it wrong and signs off every card. Each claim on a card links to the source it rests on, along with the exact quote, so you can check it yourself. And if you think we've got something wrong, every card has a dispute button.
What we mean by "not an AI summary"
A lot of apps now offer AI summaries of the news. Usually that means a language model reads a handful of articles and writes a summary that nobody checks line by line. We don't publish those. Every lowbrowse card is built from claims that have been checked against their source and quote, and a person approves it before it goes out.
Why we're telling you this
Plenty of people are uneasy about news written mostly by machines. In the Digital News Report 2025, 54% of respondents in Germany said news made mostly by AI made them somewhat or very uncomfortable 1. The German Press Council added rules on AI to its press code in 2024 2. We think you should be able to see where AI is involved in what you read, so this page sets it out for every type of card.
Sources
- Reuters Institute, Digital News Report 2025, German results as reported by MVFP on 30 June 2025
- Deutscher Presserat, Pressekodex