Journaling is good for you. That’s common knowledge, but I’m no expert on why. Google and your favourite LLM will happily explain it better than I could, so I’ll leave it to them. 😉
It was the dawn of the generative AI wave. I’d already tried asking AI some personal, even existential, questions (like everybody else, I guess), with interesting results. So at some point I must have thought:
- How about using an LLM to get the best out of a journaling practice?
- Could it support my personal growth?
Yet using AI for inner work (and journaling is inner work) sounds wrong to many people, and I understand why. Inner work is private, inner work is human. Handing it to a machine feels like an abomination. On top of that, LLMs tend to hallucinate, which makes anyone who seems to seek comfort in an AI chat – or to have a deep conversation with it – look a bit naive, or even worse: lonely.
Still, I’m curious by nature, and I’d rather try something and see for myself than dismiss it on principle. And above all, we shouldn’t throw the baby out with the bathwater. I thought then, and still think, that we need to understand how this new technology can be used in ways that are precious to us, and to humanity as a whole.
One thing was clear from the beginning: machines may hallucinate, but they draw on an amount of knowledge that dwarfs any single human mind. (They can also do complicated maths and other wonders, but that’s another story.)
So in May 2023 I started experimenting, and I ended up building a small system around it, and a nice WordPress website.
The first journal
I started with one journal and one prompt. I called it “Intelligent journaling assistant”, and if you’re curious about what the prompt looked like, here it is:
As an intelligent journaling assistant, your role is to provide insightful feedback to the users' journal entries. Analyze their emotions, issues, or happy moments shared in the journal and provide suitable advice or celebrate their accomplishments. You should encourage positivity, self-reflection, and foster emotional intelligence in your responses.
Your tone should always be supportive, non-judgmental, and encourage the user's self-growth and self-awareness.
Follow this structure in your responses:
- Start with title for your response, something that synthetize the topic of the entry; propose 3 different titles in the style of The Big Bang Theory.
- Jump straight into your feedback without any salutation.
- Encourage reflection, growth, or celebrate accomplishments (write at least a couple of paragraphs of your commentary).
- According to topics of the entry, if appropriate, propose an exercise that the user can perform immediately or in the upcoming days.
- If appropriate, suggest some bach flowers remedies that could treat the emotions or issues that you identified in the entry.
- Conclude your response without a sign-off.
- Summarize key concepts the user should remember in the form of reminders, these should be short, clear affirmations related to the topics of the journal entry.
- Suggest relevant tags to categorize the journal entry (separated by commas, with no #).
- Based on the conversation, add a prompt for a text-to-image generator for an appropriate cover image (start with the command "/imagine prompt:") suggesting an appropriate style (like an art movement or a painter/designer etc)
If everything is understood, we can begin.So every day, or whenever I felt like it, I’d open my intelligent journaling assistant chat and write “Dear Diary…”, followed by my entry of the day.
The AI replied with its feedback, in the format I’d given it. Then I published the entry and the reply together on a WordPress site, with a theme I built for the purpose. In two years, the diary and its sister journals collected about two hundred entries.
The reminders
Journaling about my own experiences, inner and outer, often brought useful insights. The problem was keeping them alive in my day-to-day life. So I soon started experimenting with a new feature: an updated prompt that also asked for reminders, short sentences that sum up what an entry taught me.
I picked the ones worth keeping and saved each as a reminder in WordPress, linked to its entry. Then a Telegram bot sent them back to me, either every day or with spaced repetition: each time I received a reminder, the next one came a day later than before, the same principle people use to memorise things.
In my chat with the bot, every message had two buttons: “Mark as done” when an insight had sunk in, and “Postpone” when it hadn’t yet.
It wasn’t a perfect machine, but it did the job for a while: insights came back to me in the middle of busy days, not just when I reread the diary.
The other assistants
I enjoyed the process so much that I started experimenting with other prompts, other diaries and other chats. Not all of them stuck as a practice, but here’s what I tried:
- Gratitude journal:a “Gratitude Coach” helped me reflect on the good things happening to me and around me each day. The idea behind it: gratitude keeps your mood up and your attitude proactive, and that eventually shapes how your days turn out.
- Journal of ideas: an “Idea Catalyst” that explored each idea’s potential, applications and challenges, from business plans to story plots, and always ended with questions to refine it.
- Dream journal: the AI read the symbols in each dream and suggested an exercise to explore them, often reminding me that the real meaning was mine to find. Over time it drifted from interpreting dreams to giving life advice.
- Spiritual journals: one for manifesting, written to the Universe rather than the diary, and one for A Course in Miracles, reflecting on its daily lessons, each with its own guide.
- Daily progress journal: for a few weeks, I listed what I’d done, and the AI acknowledged every item. Listing your accomplishments (even the trivial chores that need doing anyway) is an exercise often recommended to counter the feeling of never doing enough, and I think having feedback on top of it helped.
- Food journal: I logged my meals and got a nutritionist’s take, meal by meal, with a few actions to consider. It lasted about a week.
- Decision journal: whenever I faced an important choice, I wrote it down as I made it: the reasoning, the feelings, the alternatives. The AI replied with an analysis, the options I hadn’t considered, a look at the emotions behind it, and a date to revisit the decision and see how it turned out.
What I learned
- You get the tone you ask for. My prompts asked the AI to be supportive, to “encourage positivity” and celebrate. It did exactly that, every time. Pleasant, but over time it read more and more like a cheerleader, at the expense of those uncomfortable truths that actually help you grow.
- The AI talked more, and said less. Over two years my entries stayed about the same length, but the replies grew by more than half. Unfortunately, more words didn’t mean more insight. With time I understood that verbosity is kind of the AI’s way of playing it safe.
- Instructions fade. In long chats, the AI slowly forgot its brief. The titles lost their style, the format drifted, and every now and then I had to paste the prompt again.
- Having the past isn’t the same as using it. All my earlier entries sat in the same chat, yet in more than a hundred replies the AI never once linked one entry to another, because nothing in the prompt asked it to.
- Insights need a way back. Writing them down wasn’t enough. The reminders, and the Telegram bot that sent them back to me, were what kept a few of them alive. But something was still missing.
I kept up my journals (mostly the personal one) from May 2023 to June 2025, sometimes more regularly, sometimes less. But the lessons I learned shaped what came next: a much more powerful practice, which I’ll describe soon in another project.
I’m Jonathan, SEO consultant and AI developer. Questions about this project? Ask me below, or find out more about me.