Earlier this week I had a new AI experience – I was interviewed by the GPT-5 AI model. It was a tough interview – the questions were perceptive, responsive, and even provocative, or at least more so than the questions I am often asked in podcasts and human interviews. This AI interviewer was pleasant but focused, well-prepared, and persistent, and I engaged with them (it?) in normal speech that was accurately transcribed down to each punctuation mark – for better or for worse!

The interview was about the recent AI in Journalism Futures 2025 project, in which we used GPT-5 Agent Mode to recreate a large, complex scenario development project from 2024 aimed at understanding plausible outcome for journalism in the emerging AI-mediated information ecosystem. That project involved contributions from around 1000 different people, over a period of more than 6 months, and was designed as a fusion of two different scenario development approaches – combining written contributions from many people with a series of focused discussions among 60 people in a highly structured formal scenario development workshop.

The 2025 version followed the 2024 project design exactly, except that the contributions came from 1000 diverse ‘AI personas’ and a smaller number of AI ‘digital twins’ that represented real people in a series of 31 different discussions. This agentic AI project followed the manual 2024 project in every way, resulting in side-by-side comparisons of the same complex knowledge production project completed in two different ways – manually and agentically. It took my colleagues, Shuwei Fang of the Shorenstein Center and Nicklas Stavner from the Tinius Trust, and I just a few weeks and one hundredth of the cost of the 2024 project to complete.

The GPT-5 interview, and the prompt that generated it, are provided below on the Tomorrow’s Publisher website – itself a marvel of AI-generated news and information about the application of AI to news and information. It’s editors, Alan Hunter and Michael Brunt, spend about 30 minutes each day editing the site and maintaining a constant flow of fresh and relevant articles – many sourced from the Noah news wire service. It is becoming more and more difficult to keep track of what AI is doing, because what it is doing is so seamlessly integrated into human experiences alongside the work of humans.

The entire AIJF 2025 project, and especially this interview, brought that realisation home to me in a deeply visceral way. You too will probably soon be interviewed by an AI, as will millions or even billions of others. AI reporters will talk to us, chat with us on phones and video calls, text us, email us, with intention and focus and persistence, in pursuit of specific and particular information. People will be heard, and have their stories told, at a scale that we can hardly imagine.