The word unlimited, and what sits behind it
The word is a marketing term, not a technical term. Behind it sit 3 common arrangements: a fair-use clause that ends the conversation when somebody decides you are heavy, a silent swap to a cheaper model, and a shorter context window. All 3 are limits; none of them shows a number.
A page that says unlimited has made a promise it cannot keep literally, because hardware is finite. The interesting question is which shape the limit takes, and services differ far more in that than in whether the word appears.
On this page
Fair use is a limit with no published number
Terms of service often reserve the right to restrict accounts that use the service unreasonably. Unreasonable is not defined, which means the ceiling exists and you cannot see it. A visible cap of 30 messages tells you more than an invisible ceiling described as unlimited, because you can plan around a number you know.
The model can change without telling you
Some services keep the message count unlimited and quietly move heavy users to a smaller model. Nothing announces the change. Answers get shorter, reasoning gets thinner, and the person assumes the assistant is having a bad day. The count stayed unlimited; the product did not stay the same.
Memory is a separate ceiling from message count
A conversation has a context window: the amount of earlier text the model can still see. Unlimited messages with a small window means the assistant forgets the start of a long conversation while cheerfully accepting message number 200. People read that as the assistant getting worse, when the limit is the window rather than the count.
What honest phrasing looks like
A service that names its limits is easier to trust than a service that names none. A page saying there is a pause of a few seconds between messages and nothing else has told you the whole arrangement. Read for numbers, not for adjectives, and treat a page with no numbers on it as a page that is not ready to commit.
Fair questions
Is unlimited ever literally true?
For a human sending messages by hand it can be true in practice, because a person cannot send enough to matter. It stays untrue for automated use, and that is what the fair-use clause is written for.
How do I spot a silent model swap?
Ask the same question at the start of a session and again after heavy use. A clear drop in length and detail on identical wording is the signal worth noticing.
Does a bigger context window cost more?
It costs the provider more to run, so it is a common paid feature. This is why a long conversation is often where the paid plan starts to show its value.
Read the limits before the adjectives, then open a chat and check.
Open unrestricted chat