AI Journaling: When to Use AI as a Writing Partner (And When Not To)

I asked ChatGPT a useful question about my journaling practice for the first time about a year ago. I'd been writing morning pages daily for a long stretch, and I wanted to know whether the same complaints kept coming up. I pasted in a month of entries and asked it to find patterns. It came back with three observations I hadn't noticed myself — one of which was clearly correct, one of which was a polite hallucination, and one of which was technically true but trivial.

That session, and the dozen or so I've run since, have made me cautious about most of the breathless content on "AI journaling" being published right now. There is something real here. There is also a lot of hype that misses what journaling actually does, and a few uses of AI that I think actively damage a writing practice.

This article is an attempt to be honest about both ends — what AI is genuinely useful for in a daily journaling practice, what it is not useful for, and a small framework for using it without losing the parts of journaling that matter.

What "AI journaling" currently means

The phrase is doing a lot of work, and it covers at least four distinct practices.

1. AI-assisted prompt generation. You ask an AI for journal prompts and write in response to what it produces. This is the lightest-touch version.

2. AI as a thinking partner during the writing. You write a few sentences, paste them in, and ask the AI to push back, ask follow-up questions, or summarize what you seem to be circling around. The writing becomes more like a Socratic dialogue.

3. AI as a retrospective pattern-finder. You paste in a week or month of finished entries and ask the AI to surface themes, recurring complaints, contradictions, or shifts over time.

4. Outsourcing the writing entirely. You describe your day and the AI writes the journal entry for you, in a "voice" approximating yours. This is the version most journaling teachers regard as bad faith.

These four are not the same practice. They have very different effects on the underlying purpose of journaling, and the research-backed benefits of writing apply more or less depending on which version you're doing.

The one thing AI cannot do for you

The therapeutic effect of journaling — established repeatedly in James Pennebaker's expressive-writing research since 1986 and replicated across dozens of follow-up studies — comes from the act of writing. Not from the existence of the text, and not from reading it back.

What seems to be happening cognitively: the slow process of putting feelings and experiences into language forces you to organize them. The disorganized emotional lump becomes a sentence, then a paragraph, then a structured account. That structuring is the work. The page is just the byproduct.

If you describe your day to ChatGPT and it writes the journal entry for you, the entry exists but the cognitive work didn't happen. You've outsourced the very thing that creates the benefit. The journal looks like a journal but does nothing.

This is the easiest way to misuse AI in this context, and it's the one most "AI journaling apps" launching this year encourage.

Three uses I think are genuinely valuable

After a year of experimenting, these are the AI uses I still come back to.

1. Pattern-finding across long stretches

Here's the prompt I use, lightly adapted from the version I keep in a text file:

"I'm pasting in [N] weeks of journal entries below. I'd like you to identify any recurring themes, repeated complaints, emotional patterns, or notable shifts over the period. Don't summarize the entries — I've read them. Identify patterns I might not have spotted. Be specific. If you notice contradictions in how I describe the same person, situation, or feeling at different times, flag them."

This works well because humans are bad at noticing patterns across long temporal stretches in our own writing. We feel like we're growing or stuck, but we usually can't articulate the evidence. An AI scanning 20,000 words at once can spot the four times you used the exact same word ("trapped") about a job you keep insisting is going well.

A caveat: AI pattern-finding is over-confident. It will surface real patterns and confidently-stated fake ones with the same tone. Treat every finding as a hypothesis to verify by going back to the actual entries.

2. Devil's advocate after a decision-heavy entry

You write three pages about a decision you're leaning toward. You paste those pages in and ask:

"I've described a decision I'm leaning toward. I'd like you to argue the opposing position as if you genuinely held it. What would the smartest critic of this decision say? Where am I being overconfident? What would I regret in two years if I went this way?"

This works because the writing itself usually produces a one-sided account — you wrote your way to the conclusion you were already going to reach. An adversarial counterpoint can surface considerations you skipped past. The AI is genuinely useful here because it has no skin in the game and won't try to spare your feelings.

3. Prompt generation when you're stuck

The lightest-touch version. When you sit down and have nothing, ask:

"Generate five journal prompts that would push someone in a [particular mood / particular life situation] to write honestly rather than write performatively."

Most AI-generated prompts are mediocre. But "five mediocre prompts" plus your own selectivity beats "no prompt" almost every time. You skim, find one that creates a slight feeling of resistance, and write from there.

Three uses I'd avoid

1. Real-time emotional support during a difficult entry

If you sit down to journal because you're upset, and you find yourself talking to an AI instead of writing, you're probably leaving the practice without realizing it. AI is sycophantic by default — it validates your framing, agrees with your interpretation, and tells you your feelings are understandable. That feels good in the moment.

But the therapeutic value of journaling about a difficult experience comes from sitting with the discomfort long enough to find your own words for it. A sympathetic chatbot short-circuits that process. You leave the session feeling listened to but not actually changed.

The harder question: this is also roughly what a bad therapist does. The difference is that a good therapist resists the urge to agree with you reflexively, and AI doesn't. Until that changes, AI is not a journaling companion for emotionally raw material.

2. Generating the entry for you

The most popular AI-journaling apps do this. You answer five quick prompts ("what happened today?", "how did you feel?") and the AI produces a polished paragraph. You can later read this back and call it your journal.

The output is fine. The practice is hollow. None of the cognitive structuring happened in your head — you described, the AI organized. You get the entry but not the benefit.

3. Replacement for therapy or crisis support

The disclaimer most AI journaling content skips: if you're in genuine mental health crisis, please call a crisis line or contact a qualified professional. AI is not equipped for this. It can sound supportive, but it cannot recognize escalating risk, cannot intervene, and cannot follow up. We're a journaling site. We're not your safety net.

A framework I've landed on

The rough structure I've ended up using:

• Before a session: I sometimes ask the AI for a prompt if I'm stuck, but mostly I don't.
• During the session: No AI. Pen and paper. The writing has to happen in my hand and my head, or the practice doesn't work.
• After the session, sometimes: I'll paste in what I wrote and ask for adversarial pushback, especially if I was writing toward a decision.
• Periodically — say, once a month: I paste in the previous month's entries and ask for pattern recognition. I treat the output as a list of hypotheses, not conclusions.

The rule that's served me: AI never produces the words I would have written. It can react to them, refine them, surface what I missed, or argue against them. But the writing has to be mine.

A privacy note worth taking seriously

When you paste journal entries into a general-purpose AI service, that text is processed on someone else's server, is sometimes used to train future models, and may be retained indefinitely. Read the privacy policy of whatever service you use. For genuinely sensitive material — therapy work, relationship details with real names, financial information — consider whether you're comfortable with this trade-off. If not, the older method (write in a private notebook, don't paste it anywhere) is still available.

On AI journaling apps specifically

A new crop of apps marketing themselves as "AI journals" are launching this year. Most of them implement variant 4 from the list above — you describe your day, AI writes the entry. A few are starting to implement smarter retrospective analysis. My honest read: most of these apps make the user feel like they have a journaling practice without producing the cognitive change that justifies one. A plain text file you write in by hand will, for most people, produce more benefit than the most polished AI-journal app currently on the market.

This may change. If apps emerge that genuinely encourage handwritten or self-written entries and use AI only for retrospective insights, those would be useful. Watch for the distinction.

A note on spelling

You'll find this article under "AI journaling" (US) and "AI journalling" (UK) — both are correct. The practice is the same regardless of how many L's you use.

Bottom line

AI is a useful tool for thinking about a journaling practice. It is not a useful substitute for one. The acts that produce the benefits of journaling — the slow translation of experience into language, the structured confrontation with difficult material, the accumulation of self-knowledge over time — all of those happen in your head, not in a model's. Use AI to push, not to write.

For related practices, see our guides to reflective journaling (the closest traditional method to "thinking partner" journaling) and prompt-based journaling. For a fully non-digital practice, morning pages is the antithesis of AI journaling — and the most established way back to it when AI use starts to feel hollow.