Draft — this text has not been reviewed by a lawyer and is not final.
Privacy
What we record about your reading
We record what we showed you and what you did next: which books we recommended (and where they appeared in the list), which ones you saved, started, finished, or set aside — and if you set one aside, roughly how far you got and the reason you picked from our list. The same "what you did next" also covers: turning down a recommendation and telling us why, saying you’d already read a book we suggested, a one-tap reaction after you finish a book (if you leave one), a star rating you imported, and any private note you write in your own log.
We record the answers you give to our either/or book questions, including how long you took to answer — hesitating between two books tells us something a click alone does not. We also write short sentences about your taste and record whether you confirm, correct, or reject them, once we’ve written you one; as of today we haven’t written a single reader one yet, for anyone.
We record this so we can build your own next batch of recommendations, every night — never anyone else’s list. Each of those records is read at a different point in that night’s work. What you read, set aside, and rate is read against the whole catalog. Your either/or answers are read each night by the step that ranks your recommendations, but only for the few books whose character we’ve been able to estimate so far, a small share of the catalog; the rest aren’t reached yet. The sentences you confirm or correct about your taste aren’t read by that step at all — they show up on your own "How well we know you" page.
The nightly step that decides which books are worth studying closely enough to use this way looks at every reader’s diary together, not just yours, to choose what to study next. That part is not personal to you, even though the recommendation it eventually helps build is. For a small number of books chosen this way, we also send their title, author, and catalog description — never anything about you or your reading — to an outside AI service, so it can estimate how the prose reads.
Our application records no IP address, no device, and nothing you type into a search box — with one kind of exception, held in three places. We turn your address into a one-way fingerprint: a scrambled value, not the address itself, and not reversible by anyone without our application’s own signing key. Two forms keep that fingerprint for as long as they keep your message — the contact form and the cover takedown form — and we delete both, your message and the fingerprint together, twelve months after the last activity on it. A third, temporary copy also sits in the safeguard that stops any one address from hammering our forms or the login page too often; that copy clears itself automatically, within a few hours.
Our web server, separately, keeps an access log; today that log does record the address you connect from, the identifying string your browser sends, and the full address of each page you ask for — which includes anything you typed into a search box — for up to twelve months. We’d like to change that: to stop recording your full address and your search terms, and to delete the log after a week instead. That work hasn’t started yet, and this paragraph will change the day it does, not before.
How long we keep it
The detailed record described above — what we showed you and what you did next — has no expiry today. We’d like to reduce it to anonymous monthly totals after eighteen months, the way this section used to say we already do — but that job doesn’t exist yet, and nothing here is deleted on a schedule.
A few smaller technical records don’t have a cleanup schedule either: our internal log of admin actions is set to clean up after a year, but the job that would run it isn’t scheduled; and a failed attempt to deliver you an email, an expired password-reset request, and a line in our server logs noting that we sent you a message all sit without an expiry date today too. None of it is shown to anyone outside our own team.
What stays if you delete your account
Deleting your account removes your reading log, your notes, your taste profile, your confirmed phrases, anything you imported, and every recommendation we ever made you — permanently, with no way to restore them.
Several things stay, with your name and account removed: the tags and content warnings you confirmed or proposed (other readers rely on them); an unnamed record of the reading decisions themselves; your either/or answers; and a term you proposed for our tag vocabulary. None of them carry your name, email, or account id. If you’d claimed an author page, the corrections you sent about your own book’s catalog details or tags stay too.
If you ever paid us for anything, the purchase and its invoice stay as well — they are our own accounting record — with your account, your name and your email address gone from them and the invoice document itself deleted; what is left is the amount, the date, the invoice number, and the payment provider’s own reference for the charge.
An export today covers your reading log, your ratings and notes, anything you imported, and the tags you’ve made. It doesn’t yet include your either/or duel answers, the taste phrases you’ve confirmed, the reasons behind a recommendation, any author claim you’ve filed, your admin activity log entries, or your login sessions — ask and we’ll send you those by hand until the export does.
Nothing reads the reading record on a schedule; one manual catalogue check counts which books were shown in the last 30 days, and that count includes it. The key it’s stored under isn’t destroyed, though: it’s what lets us tell one reader’s rows apart from another’s after the account is gone, and because it also carries the moment your account was made, it could in principle be matched against our web server’s access log, which keeps that same moment for up to twelve months (see above).
What we do with a file you import
If you upload a Goodreads export, we keep the rows of that file so you can see which ones we matched and which we couldn’t. We use the title, author, identifier, your rating, and the shelf you’d filed a book under; we don’t use your reviews or any other social data for anything. Deleting your account deletes the file’s contents along with everything else.
What is private and stays private
How far you got into a book, and the fact that you set one aside, are never shown publicly. There is no public "finish rate" anywhere on this site, and there is no setting that would make one appear.
Content warnings and tags come from readers, never from software
Every trope, heat level, ending type and content warning on this site was entered by a person. We never generate them, guess them, or fill them in from a description. Where nobody has tagged a book yet, we say so — "no content warnings reported" means nobody has reported one, not that there is nothing to report. A separate, clearly labelled note on some book pages — how a book reads, its pace and its tone — is a machine reading of the text; it never counts as a content warning and it can be wrong.
Book covers
We keep a local copy of cover images so pages load reliably and we do not hammer the sources they come from. We do not claim ownership of them.
Authors who have verified a claim to their own books can send us a cover image for those books. When they do, we store the image, and we store the exact permission statement they agreed to, along with the date. A person reviews every such image before it appears anywhere on the site; until then it is not published and it is not reachable.
The takedown route below covers every image we host — the ones we cached and the ones authors sent us.
A rights holder who wants a cover removed can ask us to take it down, and we do — by hand, since there is no moderation queue for it yet. Ask us to remove a cover.
Money never touches recommendations
Our recommendations are computed without any knowledge of whether we can earn anything from a book. Today, we don’t attach purchase links to any book on the site — there’s nothing here to earn from yet. If that ever changes, a link will only be attached after a book has already been chosen on its own merits, and a recommendation with no link will stand exactly the same as one with one.
Who on our side can see this
A small admin panel lets our own staff see the account list (name, email, status, and the date you joined), the moderation queues for tags and author claims — including any evidence someone submitted, like proof of owning a book — and an internal record of admin actions. Nobody outside our team has this access, and it plays no part in building your recommendations.
Who to contact
Questions, corrections, or anything about your own data — our contact form reaches us; it doesn’t need a company to exist yet.
Company name, registration number and registered address print here once the operating entity exists. They are deliberately not filled in with a placeholder.