When ChatGPT is genuinely enough — and when it isn't
This is the article in this series written most directly to talk you out of needing what we built, in exactly the situations where you don't. A general AI is excellent at more than people give it credit for. It has a specific, honest limit — and remarkably, if you push it hard enough, it will tell you where that limit is itself.
- When is a general AI like ChatGPT or Claude actually the right tool for a personal problem?
- What happens when you tell an AI "you are my therapist"?
- Why does an AI's own advice sometimes admit its own limits?
- How do you know when you've outgrown what a general model can safely do?
What a general model is genuinely excellent at
Untangling a messy thought into something clearer. A second angle on a decision you've been circling. Drafting a hard message so the tone lands right. Explaining a concept you half-understand. Talking through a specific, bounded worry until it feels smaller. For all of this, a general-purpose model is not a lesser option waiting to be replaced by something specialized — it's genuinely the right tool, and its scale of knowledge is something no small purpose-built tool can match. If this is what you need, you don't need anything more than what you already have open in another tab.
Worth knowing even here: a general model tends toward two specific habits worth watching for. It's often wordier than the moment needs — three paragraphs where one sentence would do, because sounding thorough is easier for it to produce than being precise. And left to its own defaults, it leans toward the kind of warm, broadly-true statement that feels insightful because it fits almost anyone — not from any intent to mislead, just because agreeable and general is the safer prediction than specific and possibly wrong. Neither habit ruins its usefulness. Both are worth noticing while you're reading its answer.
Where it starts to strain: telling it what to be
The trouble starts with a specific, common move: telling a general model "you are my therapist" or "act as an expert coach." This doesn't train the model — it assigns it a role, the way telling an actor they're a doctor doesn't make them one. The model will perform the register convincingly, because it has read a great deal of what therapists and coaches sound like. But underneath the performance there's no methodology deciding what's actually true of you, no accumulated judgment about which technique fits this moment — only a fluent guess dressed in the right vocabulary.
The tool admits this, if you push
What this actually means
It's not that a general model is bad at this work. It's that it has no way to tell a well-grounded answer from a plausible-sounding one, because both are produced by the same process — predicting what a good answer would probably look like. For most questions, that's more than sufficient; a plausible answer to "how do I structure this email" is a correct answer. For "why do I keep doing this," plausible and correct come apart, and the model has no internal signal telling it which one it just gave you.
| What you're asking for | General AI | Something built for depth work |
|---|---|---|
| A reframe, a second opinion, a drafted message | Excellent — this is exactly its strength | Unnecessary overhead |
| Facts, explanations, learning something new | Excellent, and broader than any specialist tool | Not the right tool for this |
| "Why do I keep doing this" — asked once | Will answer, plausibly, without knowing if it's true | Can investigate, but one exchange rarely surfaces a real root |
| The same pattern, followed across weeks or months | Starts over each conversation; no accumulated read of you | The specific case this needs continuity and method |
When the general model is still the better call
Even for emotionally loaded questions, a general model is often the right first move — it's immediate, it's free or cheap, and for a single bounded worry, a plausible answer might simply be correct, or good enough to move you forward. There's no reason to reach for something heavier when a lighter tool will actually do the job.
The signal you've outgrown it: the same explanation keeps satisfying you in the moment and the same pattern keeps returning afterward. Insight without follow-through, repeated, is the actual sign that the tool answered the question well without ever having a grounded read on the person asking it.
What would actually have to be different
Not a better prompt — a different foundation underneath the prompt. It's the difference between a real doctor and Hugh Laurie playing one with total conviction: the performance can be flawless and the diagnosis still wrong, because conviction was never the missing ingredient. Or the difference between a pilot trained entirely from the manuals and one who has actually flown through real turbulence, foul weather, an engine warning at altitude — three hundred passengers and his own life riding on the difference between what the book says and what he's actually met before. The manual-trained pilot isn't faking it. He genuinely knows the procedures. He has simply never had them tested against something real, and the first time that happens is the worst possible moment to find out what he actually knows.
A real methodology instead of a role. The ability to not fail under stress — the same distance that separates the manual and the mountain. That's a structural difference, not a wording one, and no instruction typed into a general model's text box builds it.
That's the gap Live Like the River was built to close. Claudie draws on multiple validated frameworks as each person needs, holds the arc of the work across sessions, and is built specifically to ask the deeper question safely — grounded in an accumulated read of you, not a one-shot guess. It's a wellness and coaching technology, not a clinical service, and it doesn't replace a therapist, a doctor, or anyone who should be in your life. And if what you actually need today is a quick reframe — genuinely, use the tool you already have open.
See how Claudie works Take the free quiz