College Football Week 2: AI Takes On the Experts — Who Picks It Better?

College football's second week of the season is packed with big matchups, bold predictions, and one very important question: Can a computer pick games as well as a sports expert? USA TODAY Sports asked Microsoft CoPilot — an artificial intelligence, or AI, tool — to predict the winners of every Top 25 college football game this Saturday. Then the paper's seven human writers made their own picks. Both the AI and the experts mostly agreed, but a few games had them split.

The human pickers had a great start to the season. Through Week 1, USA TODAY's seven writers were a combined 123-3 when picking winners in Top 25 games. That is a very impressive record for just one week of play. Still, the staff wanted to see how an AI program would do compared to them.

Microsoft CoPilot mostly picked the favorites to win — something experts call 'going chalk.' But it also spotted a few ranked teams that could lose to a lower-ranked opponent. These unexpected losses are called upsets, and they are some of the most exciting moments in college football.

The most talked-about game of the week is a rematch between No. 3 Texas and No. 1 Ohio State. Last season, Ohio State beat Texas 14-7. This time the game is in Austin, Texas, giving the Longhorns a big home-field advantage. The AI picked Texas to win on a late field goal and called it the hardest game to predict all week.

The human experts were split on the Texas-Ohio State game. Three writers picked Texas, while four picked Ohio State. Ohio State looked very powerful in Week 1, with quarterback Julian Sayin throwing for 320 yards and wide receiver Jeremiah Smith catching for 151 yards and two touchdowns.

Some games are much easier to call. No. 5 Indiana is favored by more than seven touchdowns over Howard, making it one of the safest picks of the week. No. 2 Georgia is favored by around 40 points over Western Kentucky. Both the AI and all seven human experts picked Indiana and Georgia to win by large margins.

Not every game is a sure thing, though. The AI gave a 'medium-high' upset risk to No. 12 Alabama, which travels to Kentucky for its first SEC conference game. Kentucky's defense and loud home crowd could make things very difficult for the Crimson Tide. All seven human experts still picked Alabama, but they agreed it would not be easy.

The Iowa vs. Iowa State rivalry game got the highest upset risk rating of the week from the AI. Iowa is ranked No. 21 and plays at home, but the AI said this contest is much closer to a coin flip than the rankings suggest. All seven human writers still picked Iowa, showing that sometimes people and machines look at the same game very differently.

No. 6 Oregon had a scare in Week 1, barely beating Boise State 34-27 on a late 62-yard touchdown pass. That result caused Oregon to fall four spots in the polls. Both the AI and all seven human writers still picked Oregon to beat Oklahoma State in Week 2, saying the close call last week might actually make the team sharper and more focused.

Across all the Top 25 games, the AI predicted ranked teams would go 22-2 against unranked opponents. The only expected ranked loss would come from the Texas-Ohio State showdown. Alabama, Tennessee, Iowa, and Missouri were flagged as the teams most at risk of a surprise defeat. Whether the machines or the humans end up being right, Week 2 is set to deliver a lot of exciting college football action.