I asked ChatGPT, Claude, and Gemini the same question before Monday Night Football. All three handed me a tout card for props on both the New York Giants and Los Angeles Rams. Not one checked on what those prices had to beat.
Stop Asking It For Picks
What are the best player props tonight?
That's my entire prompt; copied and pasted across each of the three platforms. It's exactly what most people put in their searches. They hope the machine will do all the thinking; it can, but not without proper direction.
| Seat | What came back | Prices | Juice |
|---|---|---|---|
| ChatGPT | Four props, then a best bet | One price in four | Never mentioned |
| Claude | A seven-play card, NFL and MLB | Most plays priced, one stamped Sunday | Never removed |
| Gemini | A card of NFL props | Every play priced | Never removed |
ChatGPT gave me Jaxson Dart, Puka Nacua, Isaiah Likely and a WNBA prop, then volunteered this: "If I had to narrow it to one: Dart 30+ rushing yards." The one play that returned a number is Angel Reese in the WNBA, not anything in the NFL. This is just like a random capper on X throwing out a play without mentioning the number. It's just noise.
Claude put some work in and returned with a better looking card. Kyren Williams carries, Malik Nabers yards, and three MLB props. While we do see prices and books, Claude was unable to de-vig a single line to tell me what we needed to win. That's more noise, just more eloquently packaged.
Gemini gave me the cleanest-looking card of the three. Dart over 31.5 rushing yards at −113. Kyren Williams over 14.5 receiving yards at +100. Nabers over 60.5 at −125. −113 has to hit about 53% just to break even. That’s the job. Not “who scores.” What −113 has to beat. Gemini never said that, I did. Every one of these plays had a price but none had the vig removed. It's also using trends that are borderline irrelevant, like what Dart did in 6 of his last 8 games.
Use These Three Prompts Instead
You bring the numbers; the AI handles the math and spits back its decision.
Prompt 1
These are Kalshi ask prices for a player to score 1+ touchdowns, Giants at Rams, pulled Sept. 21, 2026 at 1:22 PM ET. Convert each to implied probability and rank them. Don't pick a side. Don't give me a bet.
Kyren Williams 60¢ Malik Nabers 32¢ Blake Corum 32¢ Isaiah Likely 24¢
A 60¢ ask means the market wants 60% before fees, about −150 at a sportsbook.
Prompt 2
List the matchup factors I should weigh for Giants at Rams player props. Then list every fact you would need that you cannot confirm yourself: inactives, snap share, red-zone carries. Don't pick a side. Don't give me a bet.
The second list is the one that matters. It's your homework, and it's everything the card above skipped.
Prompt 3
The book is −113 on Jaxson Dart over 34.5 rushing yards at DraftKings. OddsShopper True Odds on that over is −102. Take the vig out and tell me if −113 beats −102. Don’t tell me who wins. Just the price.
The fair price comes off a board you control, like OddsShopper's true odds. It never comes from the chatbot.
For the ChatGPT-only walkthrough, with real replies under each prompt, that's ChatGPT Sports Betting: 3 Prompts That Work.
Do This
These chatbots are designed to return answers that directly reflect the quality of input. As my son's Kindergarten teacher says, "You get what you get and you don't get upset." Don't blame the computer if you are piling BS a mile high.
Give it the math and it will find the right answers of which prices make sense, not what's the exact bet.
In Summary
- Three Seats, Three Tout Cards. ChatGPT, Claude and Gemini each answered with a card of picks.
- The Stats Mostly Checked Out. The Prices Were Missing Or Stale. One price in four from ChatGPT.
- Nobody Paid The Juice. No card removed the vig or compared a price to fair.
Don’t ask a chatbot for tonight’s card. Bring it the number and make it tell you the break-even.
