Terms

What this product claims, and what it does not.

A draft, not in force. The service is invite-only and nobody has been asked to accept these terms. Six clauses are still blank; each is marked below with what it is waiting on.

Drafted 2026-09-05 · retention and conversation clauses revised 2026-09-19 · operated by Trevor Van Galder · secondopinion@bullshit.doctor

1What the service does

It analyses a text you provide — how it argues, what it rests on, and, with fact-checking on, whether the figures it cites match outside sources.

Output is analysis of a text. It is not advice, not a rating of any person, and not a substitute for reading the source yourself.

2Your content, and what we keep

Text submitted through the site while signed in is kept beside the analysis, so the conversation can answer questions about the source rather than only about our reading of it. Delete the analysis and the text goes with it. Anyone holding a share link can ask that conversation questions, which means it can quote the source back to them. Text that arrives through the API is not kept at all — only a hash, a word count, whatever metadata the page stated, and the analysis itself.

You may submit only text you have the right to use for this purpose. Do not submit text obtained by circumventing a paywall, login or other access control; the service will not do that for you and you may not do it for the service. We do not train models on your content, and our model providers are configured not to either.

Pending: confirmation of the provider setting behind the no-training sentence. Until it is confirmed the sentence states an intention, and this page does not imply it has been verified.

3What you may not do with an analysis

Read it, research with it, write and teach from it, share it with attribution. What you may not do is turn it into a verdict on a person.

You may not present or resell any output as a rating, score or judgment of a person — the subject of every analysis is a text; harass, defame or target any individual; circumvent usage limits, share credentials or scrape the service; remove the “not verified is not false” and source-and-date information from any fact-check you republish; or post to a shared conversation anything you would not publish — one thread, visible to everyone holding the link, and the owner can delete any message in it or close it.

4What a fact-check means

A fact-check reports what a cited source said about a claim, as of the timestamp shown. Not verified means the claim could not be checked. It does not mean the claim is false.

Sources may be wrong, may change, and may be secondary. You are responsible for reading the cited source before relying on a verdict.

5Accounts, plans and payment

An account is required. Each plan carries an allowance measured in model cost; when it is reached, new analyses are refused until the next period. Conversations draw on a separate allowance set per analysis by its owner, so a long conversation cannot eat the allowance for new readings — when it is spent, that conversation stops accepting messages and stays readable.

Paid plans are billed in advance; the price and what it includes are shown at signup. You can cancel at any time and access continues to the end of the paid period.

Pending: billing period, payment processor and refund policy. The draft’s own recommendation is monthly only at launch, with a full refund on request within fourteen days of a first payment — not yet adopted, so it is not printed as a promise.

6Changes to the method and to these terms

The method, the model and the scores will change; that is what it means for the instrument to be maintained. What will not change quietly is the deal.

Material changes to these terms will be announced before they take effect. Corrections that do not change substance — a typo, a clarified sentence — are exempt. The service is provided as is; we do not guarantee any particular availability, and analyses can be wrong.

Pending: notice period and channel. Fourteen days is the draft’s placeholder; thirty by email is the recommendation on record.

7Liability

If a reading is wrong and you acted on it, the decision was yours.

To the extent the law allows, our liability is limited to the amount you paid for the service in the months before the claim arose. We are not liable for decisions you make based on an analysis.

Pending: the limitation window, and the operator. It is currently a sole proprietor, so there is no liability shield behind this clause; the recommendation on record is to form an LLC before the first payment and name it here.

8Closing an account

Close it and the analyses go, including any source text kept with them and the conversations attached.

We may suspend or close an account that violates clause 3. You may close yours at any time; we delete its analyses within the stated period, except where the law requires retention.

9Contact and governing law

secondopinion@bullshit.doctor.

Governing law and venue are not yet named.

Pending: jurisdiction, until residency is established. Naming a state that is not yet true would be a claim about a fact, made in a contract, that is false when written — which is the one thing this product cannot do.

A reading of how a text argues. Never a verdict on a person.

Back to the app