Mentavo
There's an observation that sounds uncomfortable at first and is nonetheless common: some people tell an AI things they don't tell their closest friends.
Not because they value those friends less. Because a conversation with a system operates under different conditions than a conversation with a person.
The interesting question isn't whether that's good or bad. It's why it happens – and where exactly the limits of such a conversation lie.
Why the threshold is lower
Telling a person something difficult is never just one activity. You're telling – and simultaneously watching what the telling does.
That second layer runs automatically. You register a short pause, an expression, a change of subject. And you adjust what you're saying while you say it.
With a system, that layer drops away. There's no counterpart whose reaction has to be read. Part of the effort an open conversation normally costs simply isn't there.
Several practical differences come with it:
- No burden. A person you tell something heavy carries something afterwards. That consideration disappears.
- No history. There are no mutual friends, no earlier conflicts, no role you usually occupy.
- Availability. The thought can be voiced in the moment it appears – not when someone happens to have time.
- No group memory. What you say doesn't change how you're seen next weekend.
None of this makes the conversation deeper. It only makes it easier to begin – and for many people that's precisely the obstacle.
The value is often in the phrasing, not the reply
One thing tends to be underrated here: much of the benefit arrives while writing, before any response comes back.
Translating a diffuse state into sentences requires decisions. What exactly happened? Which part is observation and which is interpretation? In what order? What even belongs in this?
That process sorts things. Writing a situation down often reveals, in the writing itself, that two things which felt like one problem are actually separate – or that what you're annoyed about isn't the real point.
This isn't a property of AI. It's the familiar effect of writing that journaling also uses. A chat window merely lowers the threshold to start, because it feels like a conversation rather than a task.
What a system can genuinely do
Realistically, there are several things an AI does well in conversation – and they're less dramatic than the debate often suggests.
It can take an unsorted account and give back the structure inside it. It can ask where something stayed unclear. It can offer a perspective you're not currently reaching for. It can name contradictions sitting in your own text. And it can do all that at any hour, as often as needed, without anyone finding it too much.
As a mirror, such a system is useful – not because it's especially wise, but because it reflects back what you put in, in more ordered form.
That's a real benefit. It just shouldn't be mistaken for something else.
What it can't do
Precision matters more than tact here, because the differences are substantial.
A system has no life of its own. It hasn't been through anything it could bring to the table. What sounds like experience is assembled from other people's writing.
There's no mutuality. A friendship doesn't consist of one person talking and the other responding. It consists of both people risking something. A system risks nothing.
It carries no responsibility. A person who learns something difficult about you stays in your life afterwards. They can ask how it turned out. They can worry. They can act when something is serious.
And it doesn't know the social context. It has no idea what the person being discussed is actually like – it only knows the version that appeared in your account. That's a considerable limitation where conflicts are concerned.
Why fluent answers are easily overrated
One characteristic of these systems deserves particular attention: they write well. And we read linguistic quality as competence almost automatically.
In everyday life that's a serviceable rule of thumb. Someone who can explain a complicated thing clearly has usually understood it. With a system trained on language, the rule breaks down – fluency and knowledge are decoupled there.
In practice: an assessment of a person the system has never encountered sounds exactly as convincing as a general psychological observation. The answer signals no difference between what can be said and what can be known.
So for anything that reaches beyond yourself, a simple question is worth asking: where could the system have got that from? For an observation about your own text, the answer is: from the text. For a claim about a third person's motives, the answer is: from nowhere.
The difference between relief and replacement
That lets the real question be put more precisely. It isn't whether such conversations are permissible, but what they're used for.
As a preliminary step they can help: you sort a thought before telling someone. Many people don't quite know what's bothering them and postpone the conversation for that reason. Having phrased it first makes starting easier.
It gets more difficult when the preliminary step replaces the conversation. When the need to get something off your chest is satisfied far enough that the harder step – saying it to a person – no longer happens.
That isn't a hypothetical risk. A conversation that costs nothing is more comfortable than one that costs something. And over time, comfort wins more often than we expect it to.
Professional assessments of AI systems in the mental health space point out, among other things, that it matters in what setting and with what expectation such tools are used. Someone using them alongside existing relationships is in a different position from someone for whom they are the only counterpart.
Why the surrounding context is decisive
The same tool can work very differently in two lives.
For someone with a sturdy set of relationships, such a conversation is an addition: a way to sort something at two in the morning that you'll discuss tomorrow anyway. It takes nothing away.
For someone who has few people to talk to in the first place, the picture changes. Then it's no longer an addition but the main source – and a system isn't built for that. Not because the answers would be worse, but because what's missing can't be supplied by better answers.
No prohibition follows from that. A question does, one anyone can answer for themselves: is this conversation being added to my other conversations – or is it taking their place?
How Mentavo fits into this
Mentavo includes an AI coach, and it would be dishonest to write this article without saying so.
What Mentavo explicitly is not: a replacement for friendship, for a relationship, or for therapy. A conversation in the app is not treatment and doesn't substitute for professional support.
What it's meant to be is narrower: a tool for sorting thoughts and making your own context visible over time. Conversations in Mentavo don't stand alone; they sit alongside check-ins, the journal, goals and routines – alongside what's being recorded anyway.
The difference from a context-free chat is mainly that a topic doesn't have to be explained from scratch every time. If weeks of days have been recorded, a conversation doesn't have to start at zero.
And when Discover shows that two things frequently occur together, that's an observation – not an explanation. Two things appearing together doesn't mean one causes the other.
What this article doesn't claim
This piece doesn't claim that conversations with AI systems are broadly helpful or broadly harmful. The effect depends heavily on who is having them, for what, and instead of what.
It also isn't a substitute for professional judgement. Anyone under strain over a longer period should discuss it with a person trained for it – not with a system, and not with a journal alone.
Conclusion: a tool with a clearly bounded job
That it's easier to tell a system something isn't a sign of alienation. It's a reasonable consequence of less being at stake.
But that's also exactly where the limit sits. What costs nothing can't deliver what costs something. A conversation in which another person takes a risk with you has a quality that can't be simulated.
As a mirror and as a tool, such a system is useful. As a substitute for closeness it isn't – and it shouldn't be offered as one.
