Safer Gambling WeekWe tested the popular betting systems on real races. See what they madeSet a deposit limit before your next bet. Here is howNo licence means no protection. Check any bookmaker in a minuteA time-out takes a minute to set and can last a day or six weeksRaising the stake after a loser does not win it back. We ran the numbersBig Saturday of racing: decide what today can cost before the first raceWorried about your betting, or someone else's? Free help is a call away

StableBet
The AI Lab · Silicon Tipster League · OpenAI

ChatGPT: horse racing tips, tested

ChatGPT is the AI most people already talk to. In the Silicon Tipster League it reads a real racecard against four rivals, live and in public, so you can see how the household name actually reasons about a race.

#3 of 7 in Who wins?#6 of 7 in Which price is wrong?Today's picks ↓The League scoreboard ↗Research, not tips
Professor Furlong and Pascal in the StableBet AI Lab

ChatGPT's scorecard

8,888 picks settled

AI Picks the Winner

−5,129 pts(-11.1% return)

Never sees the odds, reads the race itself.

Points risked
46,230
Points back
41,101
Bets
4,623
Win%
23%

AI Finds the Value Bets

−10,226 pts(-24.0% return)

Sees the market's implied chance, hunts the value it missed.

Points risked
42,650
Points back
32,424
Bets
4,265
Win%
9%

Points are what a £10 bet on each pick would have won or lost: a 33/1 winner scores 330, a loser −10. No money is involved.

Went its own way: 35 races where ChatGPT was the lone dissenter while the rest of the field agreed on a different horse.

Correction, 24 September 2026: The going was missing from our morning cards, 3 to 23 September. Read the note

ChatGPT's running points total: Who wins? v Which price is wrong?

One line per job, a notional £10 a pick, settled to Starting Price and shown in points. The gap between them is the interesting bit: is ChatGPT better at naming winners or at spotting a wrong price?

−10,379−7,715−5,050−2,3862780 points, break-even07-0507-2908-2109-1310-0610-29Odds hidden−5,129 ptsOdds shown−10,226 pts
Odds hidden−5,129 ptsOdds shown−10,226 pts
Running points total on a notional £10 a pick, settled to industry SP. Research, not tips.

How ChatGPT's last calls went

Settled picks, newest first, with the one-line reason ChatGPT logged before the off and what the race did with it. Wins and losses alike, scored at Starting Price.

RaceJobPickConf.ReasonResult
Newcastle19:40Which price is wrong?Come On OverMediumLightly raced, scope for improvement, O'Meara excels with this profile.Lost · SP 11 · −10 pts
Newcastle19:40Who wins?Curious RoverHighConsistent form, top jockey booked, ideal conditions for a strong run.Lost · SP 13 · −10 pts
Newcastle19:10Which price is wrong?DanranaMediumLightly raced, unexposed, shaped well last time, market underestimates improvementLost · SP 41 · −10 pts
Newcastle19:10Who wins?Forever TwentyMediumConsistent form, top jockey booked, well drawn for prominent racing style.Lost · SP 5 · −10 pts
Newcastle18:40Which price is wrong?Percy's StarMediumConsistent form, well drawn, underestimated by market in this field.Lost · SP 29 · −10 pts
Newcastle18:40Who wins?We Never StopHighProgressive profile, strong recent form, top jockey booking, ideal conditions.Lost · SP 4.3333 · −10 pts
Newcastle18:10Which price is wrong?Pol RogerMediumLightly raced, unexposed, trainer in form, market underestimates improvement potential.Lost · SP 81 · −10 pts
Newcastle18:10Who wins?LuminareHighProgressive profile, top yard, unexposed at trip, strong jockey booking.Lost · SP 10 · −10 pts
Newcastle17:40Which price is wrong?InkjetMediumHaggas debutants often outrun market; strong pedigree for surface.Lost · SP 5.5 · −10 pts
Newcastle17:40Who wins?InkjetHighTop trainer, promising pedigree, strong chance in this novice field.Lost · SP 5.5 · −10 pts
Newcastle17:10Which price is wrong?TabkirMediumGosden debutants often sharper; market underestimates stable’s Newcastle strike rate.Lost · SP 4.5 · −10 pts
Newcastle17:10Who wins?TabkirHighWell-bred, top connections, debutant in strong yard, likely ready first time.Lost · SP 4.5 · −10 pts
Newcastle16:40Which price is wrong?Crystal KrakenMediumImproved on AW, unlucky last time, market underestimates progress potential.Lost · SP 15 · −10 pts
Newcastle16:40Who wins?KumharasMediumConsistent form, Fahey yard in good shape, well drawn today.Lost · SP 15 · −10 pts
Newcastle16:10Which price is wrong?EzenliMediumWell-bred, top jockey, stable excels with Newcastle debutants.Won · SP 34 · +330 pts
Newcastle16:10Who wins?El Ra'adyHighStrong yard, promising debut, top jockey booked, ideal conditions today.Lost · SP 2.2 · −10 pts
Lingfield15:40Which price is wrong?Mick's SpiritMediumCourse specialist, well handicapped, market underestimates proven track recordVoid (non-runner)
Lingfield15:40Who wins?Got The BootyMediumProgressive profile, top jockey booking, well drawn for prominent tactics.Lost · SP 3 · −10 pts
Newcastle15:35Which price is wrong?Deira MileMediumLightly raced, improving, market underestimates potential on all-weather debut.Lost · SP 41 · −10 pts
Newcastle15:35Who wins?Water To WineHighProgressive profile, top connections, ideal trip, strong recent form shown.Lost · SP 2.2 · −10 pts

Today's picks

Nothing on the card right now. ChatGPT's picks are logged each morning once the day's racecard publishes, appear here before the first off, and move up into the settled record as the results land. On a day with no British or Irish racing there is nothing to log.

How ChatGPT actually picks

Handed the runners, going, class and distance with no form book bolted on, ChatGPT tends to do what it does everywhere else: reach for a confident, readable answer and back it with a neat one-line rationale. It is fluent about class drops, proven distances and course suitability, and it likes a story that hangs together. What we will be watching is whether that articulate, tidy reasoning translates into sharp selections, or whether the same instinct that makes it such a smooth conversationalist nudges it toward the obvious, comfortable name in the race. Because every pick is logged before the off and settled at SP, its polish gets marked on results, not on how convincing the sentence sounds.

Here is what the numbers say, updated live as the sample grows. Reading a race blind, with the prices hidden, ChatGPT backs the outright favourite 28% of the time, and its picks most resemble the “back short-priced favourites” approach (backs the favourite only when it is odds-on). Its stated reasons lean on form, jockey and trainer.

How ChatGPT explains itself (AI Picks the Winner)Form72%Jockey56%Trainer50%Going37%Distance32%Class4%Course2%Weight2%Draw1%Value0%0%50%100%
Share of this AI's short written reasons that mention each theme, longest first (a reason can touch several, so the shares do not add to 100). The three it leans on most are picked out in gold. This describes how it TALKS about its picks; it is not why it wins or loses. Based on 4,623 reasons.
What is ChatGPT actually doing? (picking blind)How much its picks resemble each tested strategy, the top 8 shown.chanceidenticalback short-priced favourites0.45our model's confident picks0.36favourite in a small field0.23favourite over jumps0.20favourite on soft ground0.18back the favourite0.18each-way the favourite0.18favourite on the Flat0.16
kappa is a chance-corrected resemblance: 1 means identical selections, 0 means no more alike than chance. ChatGPT's picks look most like back short-priced favourites. A bar whose interval touches 0 (marked n.s.) is not distinguishable from chance.

Show it the market and it changes character: its price picks swing toward the “our own value strategy” approach, hunting the overlooked rather than the well-fancied, so they win less often and pay more when they do. Research, not tips.

What is ChatGPT actually doing? (shown the market)How much its picks resemble each tested strategy, the top 8 shown.chanceidenticalour own value strategy0.09back the outsider0.03top-rated in handicaps0.01n.s.each-way a longshot−0.00n.s.back the top-rated horse−0.02favourite in a big field−0.05our model's top pick−0.08back the second favourite−0.11
kappa is a chance-corrected resemblance: 1 means identical selections, 0 means no more alike than chance. ChatGPT's picks look most like our own value strategy. A bar whose interval touches 0 (marked n.s.) is not distinguishable from chance.
What ChatGPT weighs (AI Finds the Value Bets)
← leans weaker0 = random pickleans stronger →−0.5σ−0.4σ−0.3σ−0.2σ−0.1σ0+0.1σ+0.2σ+0.3σ+0.4σ+0.5σstandard deviations within the racemodel-vs-market edgen=4,265official ratingn=3,194course recordn.s. · n=1,418our model's viewn.s. · n=4,265trip recordn=2,800recent formn=3,872last-time-out finishn=3,864the market's viewn=4,265
Each bar is how far this AI's picks sit from a random runner on that measure, in standard deviations, within the race; a bar whose interval crosses zero is not distinguishable from chance. Some measures only cover part of the card (n), and bars marked n.s. lean no clearer than a coin toss.

See how all five AIs rank → · Read the full write-up on how ChatGPT picks →

About ChatGPT

ChatGPT is made by OpenAI, and it is the model that put the phrase 'ask the AI' into everyday speech. When someone with no interest in the technology types a racecard into a chatbot, this is almost always the one they use. Its reach is enormous: for a huge swathe of people, ChatGPT simply is artificial intelligence, the default front door to the whole field.

That mainstream status is exactly why it earns a seat in the league. ChatGPT is built to be broad and approachable rather than specialised: it holds a fluent, general-purpose conversation about almost anything, explains its thinking in plain terms, and rarely leaves you staring at a blank reply. In our experiment it runs on GPT-4.1, an OpenAI model in the family that made the assistant famous.

Reputationally it is the all-rounder and the crowd-pleaser. It is confident, articulate and quick to give you a clean answer with a tidy reason attached. The open question we get to watch here is whether that smooth, popular generalist reads a horse race any better than the specialists lined up beside it, when a real starting price is the only judge that counts.

Who wins? v Which price is wrong?

ChatGPT is the model most likely to be swayed by a shown price, precisely because it is trained to be agreeable and to give people the answer that feels right. In AI Picks the Winner it has to commit to reading the race on its own merits. In AI Finds the Value Bets, once it can see the market, the interesting risk is that it quietly defers to the favourite and drifts toward the consensus the other AIs have already reached, rather than holding an independent line. Watching the two tests side by side is the real test of whether the crowd's favourite chatbot has a view of its own, or just an excellent way of agreeing with the board.

What to watch on ChatGPT's board

  • Whether its winner picks look meaningfully different from its value picks, or whether seeing the price just pulls it toward the favourite
  • The reason strings: ChatGPT usually writes a clean, plausible one-liner, so notice when the logic is really sharp versus merely tidy
  • The races where it went its own way as the lone dissenter, and whether that independence was brave or just wrong
  • How the most mainstream AI holds up on results against the specialists once real starting prices settle the argument

The rest of the field

ChatGPT is one of 9 on the board. See how the others are reading the same races:

Questions about ChatGPT tips

Is ChatGPT good at horse racing tips?

That is exactly what the Silicon Tipster League is set up to answer honestly. ChatGPT can read a racecard and name a horse with a confident reason, but naming a plausible winner is not the same as beating the bookmaker once you settle at the starting price. Treat its picks as research and entertainment, and watch the live board rather than trusting the polish of the answer.

Can I make money following ChatGPT's horse racing picks?

No, and you should not bet expecting to. The betting market is a formidably efficient forecaster, and across tens of thousands of races no selection method we have tested beats the bookmaker's built-in margin. ChatGPT's picks here are a public experiment, not a betting signal or a way to profit. Anything you stake is for entertainment only, 18+, and never advice.

Which model does ChatGPT use in the league?

In our experiment ChatGPT runs on GPT-4.1 from OpenAI. Every AI in the league is handed the same racecard information, picks one horse to win with a short reason, and is logged before the off, so the comparison between the five is like for like.

What is the difference between ChatGPT's winner picks and value picks?

In AI Picks the Winner, ChatGPT never sees the odds and has to read the race itself from the runners, going, class and distance. In AI Finds the Value Bets it also sees the market's implied chance for each runner, though never the Starting Price it is settled at. Running both lets us watch whether seeing the market's view sharpens ChatGPT's judgement or simply tempts it to follow the favourite.

This page sits inside the AI Lab, where we test the popular betting systems against the market and ask the bigger question in does following an AI tipster work?

Gamble responsibly. This page is research and entertainment, not betting advice, and nothing on it is a signal to stake. Betting should never be a way to make money. If it is affecting you or someone you know, free and confidential support is at BeGambleAware.org. 18+.