我们的法律文件仅以英文发布。英文版为权威文本;网站其余部分提供法语和简体中文版本。
Terms of Service
What you can expect from OtterSR, what we expect from you, and exactly how the money works.
Last updated · Version 2026-11-05.3
Contents
- 1.Who you are contracting with
- 2.What the service does
- 3.Accounts and sign-in
- 4.Acceptable use, and no protected health information
- 5.Your references, your PDFs
- 6.What AI output is, and is not
- 7.Price and billing: prepaid, per review
- 8.Retried model calls are billed
- 9.When a review runs out of money
- 10.Ending the agreement, and what happens to your data
- 11.The first-review credit and promo codes
- 12.Refunds
- 13.Taxes
- 14.Availability, and no warranty
- 15.Limitation of liability
- 16.Changes to these terms
- 17.Governing law and disputes
1.Who you are contracting with
OtterSR is operated by 18317208 Canada Corp., operating as OtterSR, a federal corporation incorporated under the Canada Business Corporations Act, with its registered office at 83 Gruhn St, Kitchener, ON N2G 1S5 (“OtterSR”, “we”, “us”). “You” is the person or organisation holding the account.
These terms apply to the OtterSR website and application at ottersr.com. You accept them when you complete sign-up; we record which version you accepted and when. The current version is 2026-11-05.3.
2.What the service does
OtterSR helps you carry out a systematic review: import references, define eligibility questions and extraction fields, have a large language model propose answers with a confidence score and a supporting passage, verify those answers yourself, and export the result.
Every engine tier (Otter-Nano, Otter-Mini, Otter-Pro and Otter-Max) runs on OpenAI’s API in the United States, called with storage turned off. The search index is built there too, by OpenAI’s text-embedding-3-small model. We do not train models, and we do not operate models of our own.
3.Accounts and sign-in
You sign in with an email address and a password. The password is held by Amazon Cognito, in AWS us-east-2. If you make your own account, you confirm your address with a code sent to it; if you forget your password, you reset it with a code sent the same way. Where it is offered, you can sign in with Google or Microsoft instead.
- A Google or Microsoft sign-in is not merged automatically into an account with the same email address. That is deliberate: automatic merging is an account-takeover path.
- Sessions are ours: server-side records referenced by an httpOnly cookie. A session ends when you sign out, after 14 days of inactivity, and after 30 days in all cases, and you can end every session at once with “Sign out everywhere”. Changing or resetting your password ends every other session.
- You are responsible for what happens under your account and for the people you invite to your reviews. Work an invited member does is paid from the balance of the review.
4.Acceptable use, and no protected health information
You must not upload protected health information, individual patient records, or any other directly identifying personal data about patients or research participants. OtterSR is built for published and preprint literature and for the metadata around it. It is not a HIPAA-covered service, there is no business associate agreement available, and nothing here is designed to hold clinical records.
We want to be exact about what this means in practice: we do not scan or classify your uploads, and there is no automated control that would stop you. This is a rule you agree to keep, not a filter we enforce. If we become aware that protected health information has been uploaded we may remove it and suspend the account.
You also agree not to use OtterSR to:
- upload content you do not have the right to upload or process (see §5);
- break the law, infringe someone’s rights, or attack the service;
- circumvent metering, billing, spend limits or rate limits, or resell access to the model gateway as a general-purpose inference API;
- present model output as a completed systematic review that a person has not reviewed.
5.Your references, your PDFs
You keep every right you already have in what you upload. You grant us only the licence we need to run the service for you: to store your content, send the relevant text (and passages of your documents, for the search index) to OpenAI on your instruction, and show the result back to you and the people you have invited.
Uploaded PDFs. You warrant that you have the right to upload each document and to have it processed this way — whether under your institution’s subscription, an open licence, or because you wrote it. We store documents only to provide the service. Each document is written to object storage under a key prefix that belongs to your account alone; there is no shared cache of PDFs across accounts, by DOI or otherwise. Documents are deleted when you delete them, and destroyed when the account closes (§9).
We do not use your references, documents, answers or extracted data to train models or to improve the service for anyone else.
6.What AI output is, and is not
Model answers are suggestions for a person to check. They can be wrong, confidently wrong, and wrong in ways that look plausible. Accepting an AI answer is always an explicit action you take; the software never records a decision on your behalf unless you have configured auto-accept, and auto-accept is off by default.
- The accuracy, sensitivity and calibration figures shown in the app are measured against your own validated decisions on your own data. They are not a claim about performance on any other review, and below a meaningful sample the app says so instead of showing a number.
- The second-opinion (“critic”) agent flags disagreement. It does not make an answer correct, and two agreeing models are not two independent reviewers.
- OtterSR does not replace dual independent screening where your methodology or your journal requires it. You are responsible for your methods, your PRISMA reporting and your conclusions.
We give no warranty that model output is accurate, complete or fit for any particular purpose.
7.Price and billing: prepaid, per review
OtterSR has no subscription and no per-seat fee. You pay for AI model calls, at the per-token price of the engine tier you choose, as published on our pricing page. Prices come from a versioned price book, and the price used is recorded on every usage record, so a past charge never changes because a price changed.
- Money is loaded before it is spent. Each review has its own balance. You load money onto it by card, $10 at a time or more and up to $2000 in one payment, and AI work on that review draws it down. You can import references, build a protocol and screen by hand with nothing loaded.
- Nothing is owed. There is no invoice, no month-end settlement and no spending ceiling beyond the balance itself. When a review’s balance cannot cover a call, the call is not made: AI pauses, a running job waits, and both carry on by themselves once money is loaded.
- Whose card. A card is saved to you, not to a workspace, and any member of a review may attach their card to it or detach the current one. Other members see that a card is attached and by whom; the card’s details are shown only to its holder.
- Auto-reload. Off unless you switch it on. When on, the attached card is charged an amount you set whenever the balance falls below a line you set. A declined reload is not retried for an hour, and every member of the review is told.
- Receipts. Every top-up produces a receipt. A statement of money in and usage, and a table of what each citation and each question cost, download from the review’s billing page at any time.
- Budgets and caps. A budget on a review warns every member at 50%, 80%, 100% and then refuses further AI work; a review owner can cap an individual member’s monthly spend. Neither ever blocks manual screening, validation or exports.
8.Retried model calls are billed
When a model returns an answer we cannot parse or that fails validation, we ask it again. Those repair attempts are charged, because the provider charges us for them. They appear on your usage page as their own line so you can see exactly what they cost.
- A screening or second-opinion answer is attempted at most 5 times in total.
- A data-extraction answer is attempted at most 3 times, because an extraction prompt carries the whole paper and each retry is expensive.
- Connection failures and provider throttling are not charged to you. Where a provider bills us for a call that produced nothing usable, the record is kept but marked unbillable.
9.When a review runs out of money
Nothing is owed and nothing falls due, so there is no overdue state. A review whose balance cannot cover the next call simply pauses its AI work, and a running job waits rather than failing. Loading money, redeeming a code or moving money from another review restarts it without anyone having to find it.
Your data and your exports are never blocked. Not while a review is empty, not over budget, not while an account is suspended. Every one of these stops AI only. CSV, Excel, RIS, the R dataset, the PRISMA diagram and the full account export stay available throughout. This is a product rule enforced in the code, not a courtesy.
10.Ending the agreement, and what happens to your data
There is nothing to cancel: with no subscription, an account you stop using costs nothing. You can also close the account from settings, which stops AI immediately.
- After closure, your content is destroyed 60 days later. Within that window you can cancel the closure and nothing will have been lost.
- A review you delete is recoverable for 30 days and is then permanently destroyed, including any PDFs stored with it.
- What survives: the financial record — usage events, charges, adjustments and the audit log — because it records money that really moved and accounting law requires keeping it. It contains no titles, no abstracts, no PDFs and no answers.
- Before you go, take everything with you: per-review exports in CSV, Excel, RIS and R, plus a complete machine-readable account export.
We may suspend or close an account that breaks §4 or that we are legally required to close.
11.The first-review credit and promo codes
- The first-review credit. Saving a card charges nothing and puts $50 on a review of your account, once per account and once per card. Prepaid cards do not qualify. We keep a fingerprint of the card number from Stripe to hold it to that. We may change or withdraw the offer, and may remove credit obtained by getting around these limits.
- A promo code adds a fixed credit to the review you redeem it on. Each code can be used once per person, and a code may carry its own expiry and a limit on how many people may use it.
- Credit of either kind applies only to usage on OtterSR. It has no cash value and is not refundable. Like any balance, it can be moved between your own reviews.
- Codes run under a monthly budget across all accounts and can be withdrawn at any time. If the budget for a month is exhausted, a code cannot be redeemed until the next; nothing already credited is taken back.
12.Refunds
Money loaded onto a review is not refunded to the card. Unused balance can be moved to any other review of yours at any time, and that is the normal way to recover it. If something went wrong on our side — a top-up credited twice, a job that charged you and produced nothing usable, a bug that caused work you did not ask for — tell us and we will put it right with a credit, or a refund where a credit cannot. Write to support@ottersr.com with the date and the amount.
Nothing here limits any statutory right you have that cannot be limited by law.
13.Taxes
Prices are in US dollars and exclude tax. Where we are required to collect sales tax, VAT or GST, it is added to the charge and shown on the receipt. You are responsible for any other tax arising from your use of the service.
14.Availability, and no warranty
OtterSR is provided as is and as available. There is no service level agreement and no uptime commitment at launch. We may change, suspend or discontinue features; where a change removes something you depend on, we will give reasonable notice and your data will remain exportable.
To the fullest extent permitted by law we disclaim all implied warranties, including merchantability, fitness for a particular purpose and non-infringement. The software includes open-source components provided under their own licences; those notices are reproduced at /about.
15.Limitation of liability
Neither party is liable for indirect, incidental, special or consequential loss, or for lost profits, revenue or goodwill. Our total liability arising out of or relating to these terms is limited to the amounts you paid us in the 12 months before the claim arose.
Nothing in this section limits liability for death or personal injury caused by negligence, for fraud, or for anything else that cannot be limited by law.
16.Changes to these terms
We version these terms and record which version you accepted and when. When we publish a material change we will ask you to accept the new version the next time you sign in, and we will not apply it retroactively to work already done. The current version is 2026-11-05.3.
17.Governing law and disputes
These terms are governed by the laws of Ontario and the federal laws of Canada that apply there, and the courts of Ontario have exclusive jurisdiction, subject to any mandatory rights you have as a consumer in your country of residence.