In 2001, readers of USA TODAY could encounter something unusual beside the news: a live probability generated by other readers trading on NewsFutures. Twenty-five years later, Agenda Pública has begun embedding live, AI-powered forecasts from The Forecasting Machine in its analysis. The technology has changed dramatically, but the editorial proposition remains recognisable—help readers see not only what happened, but what may happen next and with what degree of confidence.
Émile Servan-Schreiber, founder of NewsFutures and Chief Scientific Advisor at The Forecasting Machine, recently connected these two moments in a retrospective. He described the USA TODAY partnership as the first integration of a prediction market with a mainstream media website, and Agenda Pública as the first media outlet to integrate live AI-powered forecasts. Those priority claims are his account; the underlying integrations themselves are documented by the organizations involved.
NewsFutures put probabilities inside the news
Servan-Schreiber and Maurice Balick developed the first NewsFutures prediction market in 2000. Hypermind’s official history records that USA TODAY later hosted the U.S. version under the name Sports Exchange. A surviving screenshot from January 2002 shows live probability forecasts placed alongside relevant articles rather than isolated on a specialist trading page.
The model invited readers to trade with play money on clearly defined future outcomes. A price between zero and 100 could be read as the group’s implied probability, moving as participants absorbed new information. The interface turned audience participation into a measurable forecast and made changes in collective belief visible while a story was still developing.
What prediction markets added to journalism
Traditional reporting explains events through facts, context, expert analysis, and competing interpretations. A prediction market adds a different object: a probability that can be updated and ultimately scored. It does not replace reporting. It gives readers a concise view of how informed expectations are changing and creates a record that can be compared with the outcome.
- Visible uncertainty: a probability communicates more information than a binary prediction or an undefined claim that something is likely.
- Continuous updating: the forecast can move when new evidence changes the balance of possible outcomes.
- Accountability: a resolved question reveals whether confidence was justified and supports calibration over many forecasts.
The NewsFutures approach also produced researchable data. In a 2004 study covering 208 National Football League games, Servan-Schreiber and co-authors compared NewsFutures’ play-money prices with a real-money market. Both markets showed predictive power, and the study found no significant accuracy advantage for the real-money version. The result helped establish that disciplined incentives and aggregation can matter even without cash wagering.
Agenda Pública brings synthetic forecasts into analysis
On 7 June 2026, Spanish policy publication Agenda Pública announced that it would add forecasts from The Forecasting Machine to selected major analyses. Its editorial team framed the integration as a complement to long-form explanation: analysis can clarify causes, incentives, and consequences, while probabilistic scenarios can help readers distinguish what is possible, probable, or merely speculative.
An early implementation appeared in an analysis of Giorgia Meloni, Matteo Salvini, and the European Union’s fiscal rules. The page includes a dedicated Forecasting Machine widget alongside the publication’s reporting, demonstrating that the partnership is a live product integration rather than only an announcement.
From collective foresight to artificial foresight
NewsFutures and The Forecasting Machine use different sources of judgement. The earlier market depended on people seeking information, trading against one another, and collectively setting a price. The newer system can scan open sources continuously and combine the assessments of an ensemble of AI models at machine speed.
The shared discipline is more important than the shared vocabulary. Both approaches require a precise question, explicit possible outcomes, a forecast horizon, documented evidence, and a resolution rule. Without those elements, a number may look quantitative while remaining impossible to evaluate.
Editorial judgement remains essential
Embedding a forecast does not transfer editorial responsibility to a market or a model. Editors still decide which questions are relevant, whether the resolution criteria are fair, how prominently to display a probability, and what context readers need. They must also make clear when the forecast was generated and whether it has changed since publication.
This is why Agenda Pública’s emphasis on transparency matters. Forecasts are most useful as an additional analytical layer: one that makes assumptions explicit, shows uncertainty, and can be revisited as events unfold. Used carelessly, a precise-looking percentage can create false confidence. Used with context, it can sharpen the questions that reporting asks.
A 25-year thread of forecasting innovation
The path from NewsFutures to The Forecasting Machine was not a simple product upgrade. NewsFutures evolved into Lumenogic and then Hypermind, extending prediction markets into organizational and public-sector forecasting. Hypermind later spun out The Forecasting Machine to focus on AI-powered, continuously updated forecasts. Across those stages, the recurring objective has been to make uncertainty measurable and usable.
The Agenda Pública integration brings that history back to the place where the public encounters uncertainty every day: the news. Twenty-five years after probabilities first appeared inside USA TODAY stories, the same editorial ambition now has a new engine—one designed to turn open evidence and multiple model judgements into forecasts that readers can inspect, challenge, and track.
