How your data is handled

By default, nothing you write is saved. If you're not signed in — or you are signed in but haven't turned on saving — your story exists only for this conversation and disappears when you close the tab. If you're signed in, we keep a record of how you use the product — which features you open, the choices you make on screens like the save prompt, and how much AI processing each request used — tied to your account, whatever your saving setting. None of it contains anything you wrote. You can see it in Settings, and it's deleted with your account.

To produce a read, what you write is sent to our AI provider, Anthropic, to generate the response. Anthropic doesn't use it to train its models.

Separately from anything above, and regardless of whether you're signed in or have turned anything on: to keep the product usable and keep costs sane, we count how many requests — new conversations, answers to clarifying questions, follow-up messages, and answers to the optional question below — come from the same connection and the same browser each day. For the connection, your IP address is converted into a one-way cryptographic hash before it touches storage — the hash can't be turned back into your IP, and we never store the IP itself. For the browser, we set one cookie holding a random identifier, kept for up to a year. That same cookie also remembers two things: that this browser has used its one free read without an account (for 30 days, after which it's offered again), and that you confirmed you're 18 or older. We also keep a record of each such confirmation — the browser's random identifier and the time — for 12 months, as evidence it was given. None of these is linked to your account or to anything you wrote. (Daily counters aren't yet actively erased after the day ends; we're finalizing that.)

If you sign in and turn on "Save my reads," here's what gets kept: a short summary of your side of things — what you felt, the themes that came up, your own actions, and what might need to shift. Themes and actions are tags from fixed lists. The "what might need to shift" notes are in plain words and can mention the other person and what they did; if you used a real name in your story, it can appear there too. Those notes, and the theme and action tags, are stored encrypted. A few other tags from fixed lists (which feelings and values came up, how weighty the situation is, the kind of relationship, and a concern level) aren't. If you open a strategy or a reverse mirror on a saved conversation, we note which one you opened (and, for a strategy, the goal you picked) — not what it said.

Each saved conversation also gets a short topic title, so you can tell your conversations apart. It's written by the AI from what you saved (your exact words if you keep them, otherwise the summary), stored encrypted, and deleted with the conversation. You can rename it, or turn titles off in Settings.

Two things are about the other person, and both are saved only when you choose to save. First, who a conversation is about, stored as a label you choose (like "my brother" or "Dana at work"); we never ask for a real name. Second, if you describe what someone is usually like, that description is saved with them, so later reads can compare against it. Both are stored encrypted. If you don't choose to save, both are used for that one conversation and not kept. Anything you saved before turning saving off stays until you delete it. On that person's page you can rename them, replace the description, or delete them (which removes both); Settings deletes everything.

The first time you save something, we'll ask you directly whether to also save your exact words — not buried in a settings menu. Here's the honest tradeoff: it lets you come back and re-read a real conversation later, and it's what lets your history survive future improvements to how Naralytics works — without it, a change to how we analyze things can leave older sessions unable to be reprocessed or connected to new ones. It's a real choice with a real cost (your literal words, stored, encrypted) and a real benefit — we're not going to pretend otherwise. Change it anytime in Settings; delete any single conversation's exact words, or everything — your summaries, your saved words, your whole account including your sign-in — permanently, whenever you want. Deleting your account also cancels any active subscription. Stripe, our payment processor, keeps its own record of past payments, which billing and tax rules require.

Here's what we will never do with anything you save: we will never show your content to another user, or let another user's content reach you. We will never build a profile of the people you describe by combining what different users say about them. We will never use what you write to train or fine-tune any AI model. We will never sell your data, to anyone, for any reason.

Your saved text is encrypted at rest, on top of what our database provider already applies. What we can't honestly claim is end-to-end encryption: because Naralytics has to read your words to generate a read, follow up, or later re-process an entry, the server can decrypt that text when it's actually working with it. If that's a dealbreaker, the honest answer is: don't save your exact words, or don't save at all — both stay real, no-pressure options.

On retention: anything you save is kept until you delete it — there's no automatic expiry yet. We're considering an option to auto-expire saved exact words after a bounded window (kept only as long as it's useful, not indefinitely by default) while keeping your summaries; that's under consideration, not built, and derived summaries would stay permanent either way unless you delete them yourself.

After a read you may see one optional question — whether something just happened, you've been sitting with it, or you're just curious. If you're not signed in, your answer is added to a daily total and nothing else: no account, cookie, IP address, or other identifier is stored with it. If you're signed in, it's recorded like our other behavioral counts — the answer only, never anything you wrote. It doesn't change your read.

This is a test version, not a finished product — our data practices may change as we learn, and we'll tell you directly if something meaningful changes.

This is the plain-language version. The full Privacy Policy is the legal document — if the two ever say something different, the Privacy Policy is right, and we'll fix this page to match.