Soula

There is a specific gap that most mental health provision does not reach, and Soula names it precisely: two in the morning, and the ordinary hard days.

Therapy happens on Tuesday at four, and a conversational tool such as Youper answers only while you are in it. A crisis line is for a crisis. The stretch in between — the low week, the thing you keep turning over, the night you cannot sleep and do not want to wake anyone — has historically had nothing in it at all.

Whether an AI companion should fill that gap is a fair question, and this review takes it seriously in both directions. What makes Soula worth examining rather than dismissing is that it draws its own boundary clearly, which most products in this category do not.

What Is Soula, And What Is It Not?

• An emotional support companion for women, stated by the company as explicitly not therapy and not medical treatment.
• Memory is the differentiator — it retains conversation history across sessions, so you are not re-explaining your life every time you open it.
• It looks for patterns over weeks in stress, sleep and mood rather than responding only to what you said today.
• Crisis routing is built in: it is designed to recognise crisis situations and direct you to human help.
• Built by MAMATECH Ltd, backed by Berkeley SkyDeck, Mountfund, Google for Startups and the R&I Foundation, with GDPR compliance, encryption and deletion on request.

Soula: What The Company States

Memory Is The Feature That Changes The Experience

Soula retains conversation history across sessions and adapts to what it has accumulated. That sounds like a technical detail and it is the difference between a usable companion and a frustrating one.

Anyone who has tried a support chatbot knows the failure. You explain your situation, get a reasonable response, come back three days later and explain it all again — because the thing has no idea who you are. Repeating context is exhausting on a good day and impossible on a bad one, which is exactly when you would open it.

Continuity is also what a journalling tool such as Mindsera builds differently, and what makes support feel like support rather than a series of transactions. A friend who remembers what happened last month is doing something a stranger cannot, and that is true whether the friend is a person or not.

The pattern detection built on top is the more interesting half. Noticing that your sleep tends to deteriorate before your mood does, across weeks, is genuinely useful information — and it is the sort of thing nobody sees about themselves from the inside.

The consequence to hold onto is that memory is also accumulation. A product that remembers everything holds a record of your worst weeks, and that record has to be protected by something more durable than good intentions.

Being asked to re-explain your life is what makes most support chatbots useless on the day you actually need one.

Thirty-Second Check-Ins Are A Sensible Design

Soula offers a 30-second daily check-in on mood and sleep alongside longer conversation. The brevity is a deliberate and well-judged choice.

The reason wellbeing apps fail is not that people stop caring. It is that they ask for effort at exactly the moments effort is least available, and the day you skip becomes the week you skip becomes the app you delete.

A half-minute interaction survives a bad day in a way a journalling prompt does not, and consistency is what makes the pattern detection possible in the first place — sparse data produces no pattern worth reading.

It also lowers the threshold for the harder conversation, in a way a structured programme such as Liven does not. Someone who opens the app daily for something trivial is far more likely to use it on the night they need it than someone who has not opened it in three weeks.

Where The Crisis Boundary Sits

Soula states that it recognises crisis situations and directs users to human help, and that it applies a clinical safety framework. This is the most important thing about any product in this category, and it deserves examining rather than accepting.

The company's own framing is admirably clear: an emotional companion, not therapy or medical treatment, offering support in between, and encouraging clinical care where it is needed. Products that blur this line do real harm, and Soula does not blur it.

The gap is that no clinical adviser, clinician or validating body is named anywhere we could find. A clinical safety framework is a meaningful claim, and meaningful claims usually come with a name attached — a psychiatrist, a psychologist, a research partner who designed it.

That absence does not mean the framework is inadequate. It means a user cannot assess it, which for a product handling distress is a real limitation rather than a documentation quibble.

The practical position for a reader is straightforward and worth stating plainly. If you are experiencing thoughts of harming yourself, are in crisis, or have symptoms that are persistent or worsening, contact a doctor or a crisis service — not an app, and not this one. Soula's own material says the same thing, which is to its credit. Used for the ordinary hard days it describes, it is doing something reasonable.

• For crisis, self-harm thoughts or worsening symptoms, contact a doctor or crisis service — not an app.
• Treat it as company between appointments rather than as a substitute for them.
• Notice if it becomes a reason not to talk to a person; that is the risk worth watching for in yourself.
• Read the crisis resources linked in the footer before you need them, not during.
• If you are already in care, tell your clinician you are using it — the pattern data may genuinely help them.

Who Is Behind Soula, And What Do They Promise?

Soula is built by MAMATECH Ltd, founded by Natallia Miranchuk, with backing from Berkeley SkyDeck, Mountfund, Google for Startups and the R&I Foundation.

That is more transparency than most consumer wellbeing apps offer, and it matters here more than in most categories. A named legal entity, a named founder and institutional backers are all things that can be checked, and they mean there is somebody accountable for the data.

On privacy the company states GDPR compliance, end-to-end encryption, deletion available at any time, and that it never sells user data. For a product holding months of intimate conversation, those are the right four commitments, and the deletion promise is the one that matters most — an emotional record you cannot remove is a liability you carry indefinitely.

Two things to weigh alongside that. Stated commitments are not audited controls, and no third-party attestation is offered. And end-to-end encryption is a term used loosely across the industry; if the service analyses your conversations to detect patterns, something is reading them, so ask what exactly is encrypted from whom.

None of this is a reason for alarm. It is a reason to read the privacy policy properly before the first real conversation rather than after fifty of them, because by then the cost of leaving is months of your interior life.

Giving It A Fair Two Weeks

There is a free tier, so this can be tested properly without committing. The point of a trial here is not whether the responses sound good — they will — but whether continuity produces anything you did not already know.

The bar is step five. Pattern detection is the claim that distinguishes this from any general assistant, and it is checkable.

1. Read the privacy policy and the crisis resources first. For this category that order genuinely matters.
2. Do the 30-second check-in daily rather than only on bad days, because sparse data produces no pattern worth reading.
3. Have one longer conversation in the first week about something that is actually bothering you, not a test topic.
4. In week two, reference something from week one without re-explaining it, and see whether the memory holds.
5. At day fourteen, ask it what patterns it has noticed. If it surfaces something you had not seen about your own sleep or mood, the product is working; if it restates what you told it, it is a well-written chatbot.
6. Then decide about paying — and check how deletion works before you do, not after.

What Soula Charges

The model is freemium, with free check-ins and a paid tier above them; amounts are not published on the pages we read. There are iOS and Android apps.

The structure is sensible for the category. Daily check-ins are the habit-forming part and cost the vendor little, so giving them away builds the pattern data that makes the paid experience worth having.

The consideration specific to this category is not price but exit. Before subscribing, establish two things: how you delete your history, and whether you can export it. A record of months of conversation is not something to leave inside a product you cannot get out of.

The other honest note is that wellbeing subscriptions are bought during a difficult period and cancelled long after it passes. Set a reminder before renewal and ask whether you are still opening it.

We do not publish amounts, since they change. Check whether the free check-in tier remains usable if you cancel, since that is the part most worth keeping.

Who Might Get Something From Soula

It may help when

• The hard moments fall outside the hours when support is available — late at night, or on an ordinary difficult day.
• You have abandoned support chatbots because re-explaining your situation each time was exhausting.
• You want to notice patterns in your own sleep and mood that are invisible from inside a single week.
• You are between appointments and want company rather than treatment in the gap.
• You care that a named company with named backers is accountable for what you tell it.

It is the wrong tool when

• You are in crisis or having thoughts of harming yourself — contact a doctor or a crisis service instead, which the company itself says.
• Your symptoms are persistent or worsening, where clinical assessment is what is actually needed.
• You would use it as a reason not to speak to a person, which is the risk this category carries.
• You are unwilling to have months of intimate conversation held by a consumer app, however well intentioned.
• You want a validated clinical intervention, since no clinical adviser or validation is named.