Fox News’ Week 6 NFL betting-picks feature compared five selections from a human participant with five generated by ChatGPT, using odds captured Thursday at 8 p.m. ET. The article reported that the AI had a second consecutive winning week, finishing 3-2 in Week 5 while the human went 2-3.
According to the article’s accounting, ChatGPT gained 0.94 units in Week 5 and the human lost 1.26 units. The AI also won their only head-to-head matchup of the week by backing San Francisco against Seattle, while the human chose Seattle. Despite the recent results, the human remained ahead for the season: 13-11-1 with a gain of 1.26 units, compared with ChatGPT’s 11-14 record and loss of 3.87 units. The feature said the AI had gone 9-6 over the preceding three weeks after starting 2-8.
The article described the betting format as five picks per participant from a shared odds board, with no selection priced worse than -120. Stakes were adjusted so each winning bet returned one unit of profit; pushes returned the stake and counted as zero units. The piece presented both participants’ reasoning, rather than reporting an independent assessment of the games or betting strategy.
For Week 6, the human’s selections included Chicago-Atlanta under, Los Angeles Chargers +7.5 against Kansas City, Las Vegas + points against Buffalo, Dallas at plus money against Green Bay, and Washington +7.5 against San Francisco. The human’s explanations emphasized expectations about scoring, injuries and rest, along with a stated preference for betting against teams perceived as strong or popular.
ChatGPT’s five selections were all game totals: Carolina-Philadelphia over 39, Tennessee-Indianapolis under 44.5, New York Jets-New England under 43.5, Buffalo-Las Vegas over 53.5, and Chargers-Chiefs under 41.5. Its analysis cited factors including offensive and defensive production, injuries, quarterback play and the possibility of one team’s scoring contributing enough to push a total over. It projected Carolina to beat Philadelphia 24-17, Tennessee to lose to Indianapolis 24-16, the Patriots to beat the Jets 23-17, Buffalo to beat Las Vegas 31-27, and Kansas City to beat the Chargers 24-13.
The two participants disagreed on several games. The human selected the Chargers against the spread, while ChatGPT preferred the under. The human backed Las Vegas to stay close to Buffalo, while the AI chose the over. For Washington-San Francisco and Dallas-Green Bay, ChatGPT discussed the human’s picks but did not include either in its five selections. Its five Week 6 choices were therefore entirely totals bets, a pattern the human noted while questioning the high number attached to Buffalo-Las Vegas.
The article also reviewed the prior week’s results. It said the human’s Philadelphia selection covered, while picks involving New Orleans and New England did not. ChatGPT’s recap credited its Tampa Bay selection against Dallas and described its Raiders pick as a loss. The feature included the participants’ own explanations of those outcomes, including criticism of coaching decisions, but those comments were opinions and were not independently evaluated in the article.
The piece stated that the AI was prompted to review eligible spreads, totals and moneylines, explain its choices and passes, and consult at least 10 sources per game. It also said the AI used GPT-6.1 Sol with the highest available “Ultra” thinking setting. Those details describe the author’s stated process; the article did not provide the underlying source list or an independent audit of the model’s research.
