ELIZA matches your sentence against a list of keywords ("mother", "I am", "because", …), picks a canned template, and mirrors your own words back with the pronouns flipped ("I" → "you"). That is the entire mechanism — no memory of meaning, no model of you, no understanding.
Yet in 1966, people confided in it and asked to be left alone with it. Weizenbaum was disturbed by how readily users attributed empathy to a few hundred lines of pattern matching. This tendency to read mind into mechanism is now called the Eliza effect — and it is worth keeping in view when talking to modern language models, which are vastly more capable but invite the same projection.
Turn on Show the machinery below to watch the illusion being produced: each reply is annotated with the rule that matched and how your words were recycled.