Buyer's guide · checked 14 October 2026
Do online poker bots work? Three questions a buyer should separate
"Does it work" hides three questions. Does the program run and play a table? Does it win money? Does your account stay open while it plays? A trial answers the first one. The second takes far more hands than any trial gives you. The third is decided by the room, and the three rooms whose rules I read on 14 October 2026 all prohibit bots in writing. Each names account closure or blocking, and two put the balance at risk.
Updated 14 October 2026. Rule texts are quoted from the pages as they loaded that day and change without notice. We sell licenses on this site, so read this as a seller's page that tries to be straight about the limits.
1. Does it run
A working build attaches to a client, reads the table and acts without crashing, and ten minutes of a trial is enough to see that. What it does not show is whether the build still works after the room ships a client update. Ask the seller how many days passed between the last room update and their last build, and whether a license covers updates or only the build you downloaded.
Look for a stated supported client version. A page that says "works with all rooms" and names no version has not told you anything you can check. The free-trial page has the terms of four trials, and the market overview lists where sellers show up.
2. Does the account survive
I opened the rule documents of three rooms and copied the sentences about bots. I did not run or test anything against them, and I could not read PokerStars: its pages did not load for me from where I was. This table only reads the rules; it does not measure what any room catches.
| Room | Where I read it | What the text says about bots | Stated consequence |
|---|---|---|---|
| GGPoker | ggpoker.com/network/security-ecology-policy/ | A bot is "software, programs, websites or apps that enable playing without human intervention or reduces the requirement of a human to make decisions and/or execute actions." The policy also lists RTA, solvers, charts and HUDs as prohibited tools. | For bot breaches: "PERMANENT BAN, CONFISCATION OF FUNDS." |
| partypoker | help.partypoker.com, general terms and conditions | "The use of Bot software programs is prohibited and We deem its use detrimental to other players." The terms also say the room is committed to detecting software "designed to enable artificial intelligence to play". | Cheating may lead to the account being "permanently closed" with any balance "at risk of forfeiture or withholding". |
| 888poker | 888poker.com, user agreement | "You may not use any software program, robot or external aid, which is endowed with artificial intelligence or which is not included in the Permitted Aids list." Listed examples include "any type of tool that performs any action on behalf of a player at the table". | The company "may terminate your access to the Services immediately and/or have your account blocked", and may inform third parties of the breach. |
Beyond the ban, the three share little wording. GGPoker's definition is broad enough to cover a partial helper, since it names tools that only "reduce the requirement of a human to make decisions", and it lists confiscation next to a permanent ban. partypoker's passage on detection says its software can scan the list of running applications and processes on your computer. 888poker's list covers real-time advice tools alongside auto-clickers. None of the three gives a number for how many bots it catches, so any ban rate you read, including the range of 15% to 100% by room quoted on our home page, comes from operators' reports and not from a room.
A seller can sell you a program. A seller cannot sell you a room's permission to use it. If someone's page says detection risk is "low", ask what that rests on and over how many accounts. If the answer is a screenshot, you have your answer.
3. Does it win, and how long until you can tell
Poker results are noisy. I use a standard deviation of about 90 big blinds per 100 hands for no-limit hold'em cash games. That is my own working assumption, within the range commonly quoted for six-max and full-ring, and I have not measured it on any bot. Under that assumption the error on a measured win rate shrinks only with the square root of hands played.
| Hands played | Standard error | 95% interval around your measured win rate |
|---|---|---|
| 10,000 | 9.0 bb/100 | ±17.6 bb/100 |
| 50,000 | 4.0 bb/100 | ±7.9 bb/100 |
| 100,000 | 2.8 bb/100 | ±5.6 bb/100 |
| 350,000 | 1.5 bb/100 | ±3.0 bb/100 |
Take a bot that really wins 3 bb/100, which would be a good result. After 10,000 hands its measured rate could sit anywhere from about -15 to +21. To get the 95% interval down to the size of the win rate itself, you need roughly 350,000 hands. At 600 hands an hour that is about 580 hours, or 24 days of play around the clock. A seller's screenshot of a good week is a small sample, and a small sample of a winning program and a losing one look alike.
The same arithmetic runs the other way. If an account is closed after 40,000 hands, you never learn the true rate. This is why "it works" on a vendor page, backed by a few days of results, deserves a shrug. Neither the seller nor the buyer has enough hands to know.
What a trial can settle
Use the trial for the first question only, and write down what you see. Does the build start without extra permissions? Does it attach to the table type you play? How many times did it stall or crash in the first hour? How long does it take to act on a decision, and does that time look like a fixed number? None of these needs a result. They are observations about the program, and you can compare them across sellers.
If a seller answers the second and third questions with the trial's results, the answer does not fit the evidence. Ten minutes of play is on the order of a hundred hands, far below the first row of the table above, so it says nothing about a win rate.
What to ask before paying
- Which room and which client version the current build supports, with the build date.
- What the license covers after a client update: a new build, a wait, or nothing.
- What the seller says happens to the account, in writing. Check whether the reply addresses the room's rules.
- How many hands stand behind a quoted win rate, on how many accounts, and over what period.
- Refund terms that exist before payment, per the buying checklist.
Treat any balance on an account used this way as money you may not get back. The rooms above state that balances can be confiscated or withheld.
Questions people ask
Do online poker bots work? They run. Whether they win is not provable from short results, and whether the account lasts depends on the room, whose rules prohibit them in all three documents I read.
Are poker bots allowed on GGPoker, partypoker or 888poker? No. Each rule text I read on 14 October 2026 prohibits them. The consequences range from closing the account to confiscating funds.
How many hands do I need to know if a bot is profitable? Under a 90 bb/100 standard deviation, about 350,000 to measure a 3 bb/100 win rate to within its own size.
Will the rules change? They do. Read the current text, and I will date any re-run.