How Botto plays fair
Every game here is played against an AI. That only means something if the AI cannot cheat — so here is exactly what stops it.
The game lives on our server, not in your browser
The real state of a game — which cards are where, what the secret word is, whose turn it is — is held on our server and checked there. Your browser is shown a view of it, and so is Botto. Neither of you is trusted to report what happened; every move you make is re-validated against the real board before it counts.
That is also why you cannot cheat. It cuts both ways on purpose.
Botto is only ever told what a player in its seat could see
When it is Botto's turn, the server builds it a fresh description of the game from the public record — the moves both of you have already made, in the open. There is no code path that puts a hidden value into that description. Not "we ask it not to look": the information is not in the message.
In a game with a secret — the hidden phrase, the sealed word, the crossword solution — that secret never leaves the server until the moment the game reveals it to both of you at once.
Its memory is a rule, not a database
Botto does not keep notes between turns. Each turn, the server replays what has happened in the open and applies the limits of the level you chose: how many positions it can hold in mind, how often a remembered thing comes back wrong, how fast older memories fade.
So Chill Botto genuinely forgets, and Genius Botto genuinely remembers. The difference between the levels is a real constraint on what it is allowed to know — not an instruction to play badly that it can quietly ignore.
It gets no free moves
Every move Botto makes goes through the same legality check yours does. If the AI behind it fails, times out or answers with nonsense, it gets one correction and no more — after that the game itself plays a plain legal move on Botto's behalf. It never gets to keep trying until something lands in its favour.
The referee is not Botto
Some games need a judgement call — is that a real word? In those, the ruling goes to a separate referee: its own request, with its own rulebook, no personality, and no idea which of you played the word. It applies one standard whether you are on Easy or Genius.
Where a check can be made in code instead, it is: words are checked against a fixed dictionary, and crossword answers against the solution the server generated before play started.
What this does not mean
Botto is not perfect, and it is not pretending to be. At the higher levels it will beat you often. At the lower ones it will miss things an attentive person would not. Both of those are the design working.
What it will never do is see a card you have not turned over.