November 12, 2026, (Inside AI) — Microsoft Copilot's NFL prediction model entered Week 10 of the 2026 season with a 14-1 record from Week 9 and a season-long mark of 89-46-1. The chatbot, tasked by USA TODAY Sports with picking every game, initially pulled incorrect schedule data before being asked to reassess. Its corrected Week 10 slate includes a rare endorsement of the New York Giants' opponent, the Washington Commanders, and a surprising pick against the Los Angeles Chargers that hinges on the team's winless record.
The experiment matters beyond football fandom. It offers a public stress test of a large language model's ability to synthesize real-time sports data, injury reports, and historical team tendencies into coherent, falsifiable predictions. Copilot's Week 9 performance, which included correctly calling the New Orleans Saints' overtime upset of the Cleveland Browns, suggests the model can occasionally outperform human intuition on tightly contested games. But its initial schedule error in Week 10 exposes a persistent weakness: LLMs often struggle to distinguish current data from stale training material.
Copilot's Week 10 picks span 14 games, with the model favoring the Commanders over the Giants on Thursday Night Football, the Detroit Lions over the New England Patriots in Munich, and the Kansas City Chiefs over the Atlanta Falcons. It also backs the Houston Texans, Green Bay Packers, Jacksonville Jaguars, Indianapolis Colts, New Orleans Saints, Buffalo Bills, Seattle Seahawks, Los Angeles Rams, San Francisco 49ers, Cincinnati Bengals, and Baltimore Ravens.
One of the more notable selections is Copilot's pick of the Ravens over the Chargers, even without quarterback Lamar Jackson. The model said Los Angeles has "shown it cannot consistently finish games," a claim that aligns with the Chargers' 0-8 start. The Ravens acquired Jacoby Brissett to stabilize their offense, and Copilot's logic suggests that may be enough against a struggling opponent.
Read: AI Chatbots May Narrow Knowledge
Copilot also flagged the Bills vs. Jets matchup as its "strongest selection," citing Buffalo's "elite" offense against a Jets unit that has "struggled to consistently score." That confidence comes despite Buffalo's uneven Week 9 performance against the Minnesota Vikings, a 23-6 win in which the offense failed to find a rhythm.
The model's pick of the Packers over the Vikings was labeled "one of the toughest calls of the week." Copilot sided with Green Bay's "offensive balance" and a playoff-like atmosphere. The Vikings recently traded for Kirk Cousins, a move that raises their ceiling but also introduces uncertainty against a Packers defense that has improved under coordinator Jonathan Gannon.
Copilot's Week 10 output also includes a pick of the Colts over the Dolphins, a game the model described as "ugly either way." It backed Indianapolis because the team is at home and has "been more competitive recently." The Dolphins, meanwhile, placed Malik Willis on injured reserve with a groin issue, leaving their quarterback situation unsettled.
The chatbot's early schedule error is a reminder that AI prediction tools remain dependent on human oversight. When asked to reassess, Copilot corrected its data and produced a full slate of picks. That iterative process mirrors how enterprises deploy LLMs in other domains, from financial forecasting to supply chain planning, where initial outputs often require validation against ground truth.
Copilot's 89-46-1 season record translates to a winning percentage of roughly 65.9%, a mark that would rank among the better human experts in NFL pick 'em contests. But the sample size remains small, and the model's reliance on public data means it may not capture late-breaking injury news or locker-room dynamics that human analysts weigh.
Read: Have AI Disagree With You
For now, Copilot's Week 10 picks offer a mix of chalk and contrarian calls. The model's willingness to back the Commanders over a Giants team that has "exceeded expectations" suggests it values quarterback Jayden Daniels' ceiling over New York's coaching advantage under John Harbaugh. That pick, like all the others, will be settled on the field.
Sofia spent years studying how machines parse human language before pivoting to cover the systems now rewriting it. She reads technical papers the way others read news and looking for what the numbers are actually saying beneath the headlines.
