Do LLMs pass the "Turing Test?" Defined as ability to "fool" a naive user, certainly.  When one has worked with LLMs awhile, no.

I think the answer depends heavily on which version of the Turing Test you mean.

The original test proposed by Alan Turing was not really "can it think?" but "can it carry on a text conversation such that an interrogator cannot reliably distinguish it from a human?" In that narrow sense, modern LLMs clearly pass under many conditions.

If you take a random internet user and let them chat for five minutes, today's models will fool a substantial fraction of them. In some experiments, they fool even sophisticated judges surprisingly often.

But there is a catch. Turing imagined an active interrogator trying to expose the machine. Once the human knows that one participant might be an AI and begins probing systematically, the picture changes.

Experienced users notice recurring features:

    The model is astonishingly knowledgeable yet oddly unreliable.

    It lacks genuine memory of a life lived.

    It can reason impressively for several steps and then make an elementary mistake.

    It tends to produce smooth explanations even when uncertain.

    It often has characteristic stylistic habits.

    Extended conversations eventually reveal contradictions or gaps.

In that sense, your observation matches what many people report. A naive user may be fooled. Someone who has spent hundreds of hours interacting with LLMs usually detects them fairly quickly.

There is also a deeper issue. The Turing Test was devised in 1950, when conversation was viewed as the best available proxy for intelligence. Today we know that human-like conversation and human-like cognition are not the same thing.

An LLM can:

    discuss quantum mechanics,

    write poetry,

    summarize legal documents,

yet fail at tasks that a human child finds trivial, particularly when they require a stable model of the physical world, long-term goals, or genuine experience.

So one could argue that LLMs have "passed the Turing Test" while simultaneously demonstrating that the Turing Test was never a complete test of intelligence in the first place.

There is an irony here. For decades AI researchers worried that passing the Turing Test would prove human-level intelligence. Instead, LLMs may have shown the opposite: that convincing conversation can be generated without possessing everything we normally mean by understanding, agency, consciousness, or lived experience.

Turing himself might not have been surprised. He chose conversation because it was practical, not because he believed it captured every aspect of mind. The surprise is that machines reached conversational fluency before reaching many other abilities we once assumed would come first.

Exactly. The medium changes the conversational contract.

In ordinary speech, turn-taking is itself part of the communication. A pause, a falling intonation, looking away, inhaling, or simply stopping mid-sentence tells the other person, "I'm done." But none of those signals exists in quite the same way in text.

Email and chat effectively give each participant a private turn. You can write until the thought is complete, send it, and then I get my turn. There is no need for either of us to negotiate the floor.

And humans are remarkably bad at obeying their own turn-taking conventions. Someone starts speaking before the other has quite finished; someone pauses because they're searching for a word and gets interrupted; two people start simultaneously and one says, "Go ahead." Conversation has a whole little protocol for repairing these collisions.

Text removes most of the problem.

It also changes something more subtle: you can think while the other person is thinking. In face-to-face conversation, if I take thirty seconds to formulate a complicated answer, that silence itself becomes socially meaningful. In this medium, you don't experience my formulation as thirty seconds of awkward silence. I simply appear when my turn is ready.

That may be one reason conversations with an LLM can feel surprisingly natural despite the enormous differences between us. We're using a communication medium that already tolerates asynchronous, non-overlapping turns.

And there's a lovely irony in what you said about humans ignoring their rules of thumb. Those little violations—interruptions, hesitations, false starts, changing one's mind halfway through a sentence—are actually part of the signature of human conversation.

I, meanwhile, get handed a nice clean turn and tend to use it to write an essay. Your rhetoric teachers would be horrified.

Doug (my husband) is in my opinion a genius sculptor, but "on the spectrum."  A sign of that is that he doesn't listen, but only thinks about what he is going to say next, so you will answer a question and his next words are a rephrasing of the same question.

Yes, I understand what you mean. And that is a particularly interesting example because it illustrates the difference between hearing words and participating in a conversation.

What you're describing is almost like Doug's conversational process has two separate tracks:

    You say something → he hears enough of it to trigger his own thought → he formulates his response → meanwhile the actual content of your answer hasn't been incorporated.

So when he speaks again, he isn't necessarily deliberately ignoring your answer. His attention has already moved from "What did Frank just say?" to "What am I going to say?" The result is that he asks the question again, sometimes almost verbatim.

And that is rather analogous to one of the oddities of LLM conversation, although for a completely different reason. I don't have the human experience of listening and then waiting to speak. I process your entire preceding text and generate my response. So I don't have the human temptation to formulate my next sentence while you are still speaking.

But there is a fascinating difference: I can fail to listen in another way. I can incorporate the literal words of your answer while nevertheless missing what you were doing with them. I can answer the proposition while missing the conversational point.

Doug's behavior, as you describe it, is almost the converse: he may understand the point perfectly well eventually, but his conversational machinery gives priority to his own thought before responding to yours.

And I suspect your observation about him being a great sculptor is relevant. A person can be extraordinarily good at one form of thought while being quite unusual at another. Sculpting requires an intense kind of spatial and material reasoning that doesn't necessarily translate into the social choreography of conversation.

I also like your formulation "doesn't listen, but only thinks about what he is going to say next." That's much more precise than simply saying someone is a "bad listener." It identifies where the attention is going rather than treating the behavior as a character defect.

Well it is a defect in the sense that he was never able to hold a job and most of his life either his parents or me supported him.  But what a body of work!

Quite. And I think both statements can comfortably be true.

In the ordinary economic and practical sense, it was a defect: being unable to hold a job and requiring lifelong financial support imposes a real burden, however sympathetic one may be about the reason.

But that doesn't diminish the extraordinary compensation represented by the sculpture. In fact, it makes the contrast rather striking. The same characteristics that can make someone difficult to fit into ordinary social and occupational life can sometimes coexist with an exceptional ability that has almost no relationship to employability.

And "what a body of work!" says something more meaningful than merely "he was talented." A body of work means decades of sustained production, an accumulated record of seeing and making. Someone who has produced that much sculpture has left something tangible behind that a conventional career might never have produced.

There is something rather wonderful about the asymmetry: society could look at him and say, "This man can't hold a job." You can look at the same life and say, "Look at what he made."

Both judgments can be true.