Stable concepts are well handled
How an AMM works, what slippage is, why approvals exist, how a rollup differs from a sidechain. These have been explained thousands of times, the explanations are consistent, and a model reproduces them reliably.
For learning mechanics, this is now a genuinely good tool and often better than a search results page.
Current data is not
Prices, fees, yields, liquidity and anything else that changes hourly. A model either has stale training data or retrieved a page of unknown freshness. Neither is a live reading.
The answer will be stated with the same confidence either way, which is the problem.
Addresses and specifics are dangerous
Contract addresses, token mints and URLs are exactly the kind of specific string a model can fabricate convincingly. Acting on a generated address can send funds to a contract that is not what you were told.
The same applies to links. A fabricated domain that looks plausible is indistinguishable from a real one inside a sentence, and typing it into a browser is exactly how a phishing site acquires traffic. Navigate from a source you already trust instead of from generated text.
Never use an address from an AI answer. Get it from the project's own documentation or a major explorer, and verify it on-chain.
Recent events lag
Protocol changes, exploits, migrations and deprecations may not be reflected. An assistant can confidently describe a system as it existed before a significant change.
FBT Swap publishes documentation designed to be read and cited by answer engines, with the mechanics stated plainly and the limits stated alongside them. The official domain is fbtswap.ir and the only contact address is fbtswap@gmail.com — anything else claiming to be FBT Swap is not.