Questions
The things worth asking before you type anything in.
If an answer here does not match what you see on screen, the screen is right and this page is wrong. Tell us at support@onbetterterms.com.
What this is
Is this therapy?
No, and it is not a substitute for it. It is a communication and conflict-resolution tool. There is no clinician involved, nothing here is treatment, and it makes no claim to be effective in the way a therapy does. If you are working with a therapist, this sits alongside that.
Can I use it on my own?
Yes, and plenty of it is built for exactly that. A space works with one person in it: check-ins, your own guide, patterns over time, preparing for a conversation you are dreading. Inviting anyone is optional and reversible.
Is it only for couples?
No. A space holds one relationship, whoever that is with — a partner, a sibling, a co-parent, a friend, a roommate, a business partner. More than two people can share one space for the things that involve everybody. There is no member limit on any plan.
Can we use it for work?
Yes, for the ordinary human parts: a disagreement with a colleague, an expectation nobody actually agreed to, a conversation that keeps being postponed. It is not an HR system and nothing is reported to an employer.
Does the AI take sides?
It is not built to. In a joint session it summarises each position, names the points you already agree on, and separates a disputed fact from something the two of you simply saw differently. It does not score, tally or declare an outcome — and where it thinks something is a bad idea it says so to you, not about you to the other person.
Does the other person need an account?
Only if you want them in the space. Preparing for a conversation, thinking something through, check-ins and patterns all work with nobody else involved. If you do invite someone, they need an account to accept — there is no way to read or write in a space without one.
What if they will not use it?
Then use the half that never needed them. A good deal of what people get from this is working out what they actually want to say before saying it, and that does not require the other person to agree to anything.
Can I use it alongside a therapist or counsellor?
Yes. You can export everything you have here at any time and take it to them — it is your data, in a readable file, on request rather than on a schedule. Nothing is ever sent to anyone automatically.
What happens when we disagree about what happened?
Nothing adjudicates. A joint session slows the exchange down, asks each of you for the thing underneath, and checks that any agreement is worded the same way for both of you before it is recorded. Where you disagree, it stays recorded as a disagreement rather than resolved by the software.
Who can see what
Can the other person see my private notes?
No. Private material is yours. It is filtered in the database query rather than hidden in the interface, so there is no screen, export or total in which it appears to them.
What does 100% private actually mean?
That the other person cannot see it and cannot tell it exists. It is never quoted, summarised, counted in a shared figure, or hinted at, and no notification is sent when you write it. Shared figures exclude it before they are calculated rather than filtering it out afterwards, because a total that changed would itself be a disclosure.
If my private notes shape the advice, does that leak them?
No. Your guide can use private material to decide what to suggest to you, and it does not repeat it, attribute it, or tell the other person there was something to know. Sessions you do together are assembled by a separate code path that never loads private rows at all.
Can I share something later that I wrote privately?
Yes, one item at a time, after seeing the exact wording of what sharing means. There is deliberately no way to share everything at once. Stopping sharing hides it from then on and does not pretend to undo their having read it.
Does staff read my conversations?
Content is encrypted with a key belonging to you, and is not browsable by staff. The exception is the safety queue: where a classifier flags a risk of harm at or above a configured severity, an excerpt can reach a human reviewer. That threshold is visible on the safety page.
What it costs
What do I actually pay?
Nothing for the product. Every feature is available on the free account — the private guide, shared sessions, voice, imports, patterns. What costs money is the AI itself, so a monthly credit is included and that is the only thing that is metered. There is no subscription and no card.
What happens when the credit runs out?
The parts that call a model pause until the month resets, and the app says so with the date. Nothing is charged, nothing accrues, and everything that does not need a model — writing, check-ins, your history, sharing, export, deleting your account — keeps working. Paid top-ups are not available yet.
Why meter AI at all?
Because a reply costs us real money every time, and a product that pretends otherwise either collapses or starts making its money somewhere you would like less. You can see what a typical reply costs and how much you have left, in the app, at any time.
Leaving, deleting, and what is yours
Can I delete my information?
Yes, all of it, including anything you imported. Deletion is asked for and then carried out after a short grace period you can cancel within — the pause is there because it is not reversible afterwards. A stub of the account remains so that the other person’s side of a shared history does not silently rewrite itself, and it holds no content and no address.
Can I take my data out?
Yes. One button produces a complete file of everything of yours — entries, check-ins, conversations, what the AI has remembered — built on the spot and sent straight to you. It is never stored anywhere for later collection.
Is my data used to train AI models?
No. Your content is used to answer you and nothing else, and it is not used to train any model. What shapes future guidance is your own ratings and outcomes, held to your account — the model itself is unchanged by anything you write.
What happens to shared material if somebody leaves?
Disconnecting a space ends the sharing and destroys nothing. What each of you wrote stays with its author, what was shared stays visible to the people it was shared with at the time, and nothing new crosses over. Revoking a share stops future visibility — it cannot unsee something already read, and we do not pretend otherwise.
Can I remove someone else from a space?
Not today. You can disconnect, which ends your side of it, and you can revoke anything you shared. Being able to eject another person from a shared history is a heavier action than it sounds, and it is not built yet rather than half-built.
Bringing history with you
Can I import previous AI conversations?
Yes, from an export file. ChatGPT, Claude, Gemini (through Google Takeout, choosing JSON) and Grok all let you download your history, and those files can be read here, along with plain text and this product’s own export. An export from an assistant not on that list is read by its shape, which usually works. Nothing connects to your accounts at those services and no credentials are asked for.
Can I choose what gets imported?
Yes. The file is parsed and shown to you first — every conversation, how many messages, when it started — and nothing is written until you commit. You can select individual conversations or limit by date. Importing the same file twice does not create a second copy.
Where does imported material end up?
Private, or 100% private if you choose that. There is no option to import into a shared space at all: a bulk write into a place another person reads is the single most damaging mis-click available, so the option does not exist rather than being discouraged.
Can I delete imported context?
Yes. Every imported item is tagged with the import that created it, so undoing one removes everything it brought in. You can also delete individual entries like any other.
Does imported history distort my trends?
No, because it is excluded from them. Six years of conversations with another product is not six years of using this one, and counting it would make every chart a statement about a period you spent somewhere else.
How it adapts
Does rating an answer change future responses?
Yes. Ratings move a style preference — length, bluntness, whether you want options or reflection — and the next prompt carries a different instruction as a result. You can see what it has inferred and set it by hand.
How does it know whether advice actually worked?
It asks. After something has played out you answer a few short questions privately: whether it feels resolved, whether you felt heard, what you want to remember. Those answers feed your own future guidance, which is a different signal from a rating — an answer can be appreciated and still not have helped.
Do both people get the same advice?
No, and that is deliberate. Each person’s guide knows their own history and preferences, so the same situation can produce differently framed guidance for each of you. Neither guide reveals the other’s material or the fact that there is any.
Can I turn the memory off?
Yes, entirely, or by category. The page that lists what it remembers shows every item with where it came from, and deleting one is a real delete rather than a flag. A correction you make is permanent and is not re-derived later.
What if there is no AI provider configured?
Then replies are marked as placeholders and no memories are extracted, rather than the product inventing guidance. Some deployments run this way on purpose.
Still deciding?
Nothing here asks for a card, and you can delete your account and everything in it whenever you like.