Fair questions.

A citation identifies a source using available bibliographic details. A receipt presents a recorded claim outcome, evidence and limitations. It does not guarantee a settled answer, a confidence score or support from every cited source.

Someone mid-argument who knows they read it somewhere and needs the source in ten seconds. Paste a claim in the popup or use a deliberate page selection.

Paste a claim in the extension popup, or deliberately check a selection on supported pages using right-click or your configured shortcut. Direct editing in Google Docs and other rich-text editors is not universally supported; use paste.

The Trust API accepts one claim per request and returns recorded result fields and available sources. Confidence may be unavailable, and checks may produce public receipts. Do not submit confidential or personal information. Read the API page for authentication, limits, current prices and response handling.

Five checks a month are free, receipts and share links included, no card. Bro is $9.99 a month for 400 checks and 100 API calls; Pro is $29.99 for 1,000 of each. Annual is twelve months for the price of ten.

Yes. Bibliographic metadata, quotations and evidence assessments can contain errors. Open the recorded source and check its identity, wording and context. A retrieved source or formatted citation does not by itself establish support for the claim. Read the receipt's limitations before relying on it.

Retraction scans look up recorded DOIs in OpenAlex. They do not cover every source or guarantee an overnight update. A failed or missing lookup is not evidence that a paper is unretracted. Check the publisher's current notice and the receipt's recorded history; a scan is not a fresh judgment of every claim.

The accuracy page shows a historical evaluation report with its recorded date, case count and measurement limits. It does not establish the accuracy of the current release. A source being returned is not proof that it supports a claim; inspect the receipt's evidence and limitations.

Read the stored result and evidence on each side. CONTESTED and CONTESTED PARTIAL distinguish different evidence outcomes; disagreement alone does not predetermine a verdict. A result may also remain insufficient or unpublished.

Obviousness does not guarantee a particular verdict or source count. Some historical receipts use COMMON KNOWLEDGE. Read the actual recorded judgment, its evidence and limitations rather than assuming that label will be returned.

A check can return INCONCLUSIVE or INSUFFICIENT EVIDENCE, or be withheld or fail without a published receipt. Read the returned status and explanation. A missing source is not proof that none exists; a technical failure is not an evidence judgment.

Where a receipt offers a challenge control, sign in and submit a source URL with an optional note. Accepted submissions enter manual review. They do not automatically re-run the engine, add the source to the evidence or change the verdict. A submission does not guarantee a response time or a changed result.

Claim limits depend on the plan and remaining allowance. Flagged claims do not guarantee matching sources or citations. Keep each claim within the stated input limit and inspect the recorded result.

One claim, one question, one citation, one receipt. Read the recorded result and its limitations before relying on it.

Drafts saved to an account by an earlier version are retained while that account exists, and you can delete them from your account page. Accountless batch text and results have a 30-day retention window after completion, failure or cancellation. Public receipts remain after account deletion. See the privacy policy for retained records and removal requests.

Claims can appear in public receipts or the Trust Feed, even if you do not share them. Public pages omit API-key and account fields, but submitted text and sources can identify your users. Do not send personal or confidential information; treat a claim as a public question.

It can. A receipt omits account and device fields, but the claim and sources can contain names or other identifying details. The receipt may already be public before you share it, and recipients may associate you with its claim. Read the receipt before sharing.

Pro gives 1,000 claim checks and 1,000 API calls a month against Bro's 400 and 100, no daily cap, a faster API rate, and the saved claim library.

One request to /v1/verify that returns a judgment, whether the answer came from an existing receipt or a fresh check. Failed checks do not count. Included calls reset on the 1st of each month (UTC) and do not roll over.

Any time, from your Stripe receipt email. No retention maze, no phone call, no three-step confirmation flow.

Bro includes 100 API calls a month and Pro includes 1,000. Past those, or with no plan at all, calls draw on a prepaid wallet: $0.003 for a claim already in the corpus, $0.01 for one nobody has checked yet. Every new account gets $0.10 of wallet credit with no card, top-ups run $5 to $50, and the balance does not expire. The ceiling on what we can charge you is your plan plus the money already in the wallet.

Create an account, generate a key in your wallet, and send it as a bearer token. We show the key once and store only a hash, so a leak of our database yields no working keys and a key you lose cannot be recovered by anyone, including us. Keys are revocable, and five active keys is the limit.

We're cooking the most powerful verification tools on the internet, and a social space where debates get settled with receipts instead of volume. Join us for the ride: leave an email and you get early access to every new feature as it lands.