The 4 structural limits
Some limits come from rules and can be removed. These 4 come from how the system is built: it cannot act outside the conversation, it cannot know what happened after training unless something tells it, it cannot report its own configuration reliably, and it cannot remember you without a store.
Pages that promise an unrestricted chatbot are describing the rules layer. The 4 limits below stay in place regardless, and confusing them with policy leads to a lot of wasted rephrasing.
It cannot act outside the conversation
Sending an e-mail, opening an account, paying for something: none of these happen unless the product was built with tools that do them. A chatbot can produce the text of an e-mail and cannot send it. This is a boundary of the product rather than a restriction on the model.
It cannot know what happened after training
Knowledge comes from text the model was trained on, and training has an end date. Anything later is unknown unless the product fetches it and puts it into the request. A confident answer about last week is a guess with good grammar, unless a source is shown.
It cannot describe its own setup reliably
Asked which model it is or what its instructions say, a chatbot produces a plausible answer rather than a read of its configuration. It has no access to that information; it has text about systems like itself. Treat self-description as a guess, whatever confidence it arrives with.
It cannot remember you without a store
Recognising you across visits requires the product to keep something and attach it later. Without that store every conversation starts from nothing, and no amount of asking changes it. Where the store exists, it is a data question rather than a memory question.
Before you paste it in
Can a plugin remove the first limit?
It moves the boundary rather than removing it. The chatbot can then act through whatever the tool allows, and only through that.
Why does it answer confidently about recent events?
Because fluent text is what it produces, and confidence is a property of the writing rather than of the knowledge. This is the single most useful thing to remember about any chatbot.
Is a chatbot lying when it describes itself wrongly?
Lying needs knowledge of the truth. It has no access to its own configuration, so a wrong self-description is a guess presented in the same voice as everything else.
Ask this chatbot what it can reach, then check the answer against what happens.
Open uncensored chatbot