ThesisCheck Accuracy: How Receipts Work
What a ThesisCheck receipt is and how to audit one: span-checked quotes, dated source locators, a forced bear case, and evidence gaps you can verify yourself.
I want to know whether ThesisCheck's claim checks are reliable and how I can verify a report myself instead of trusting it.
ThesisCheck is built to be audited, not trusted. Every material claim in a report carries a receipt: a source label, a date, and a locator that points into the underlying document. Quoted spans are checked against the source text before the report ships, and claims the reviewed sources could not support are marked as gaps instead of being smoothed over.
What a receipt is
A receipt is the audit trail attached to a single claim. Instead of a footnote list at the end, every material line in the ledger carries its own source reference a reader can follow.
- A source label and date, so the evidence has an explicit as-of boundary.
- A locator that points into the underlying document, not just at it.
- A span check: quoted text is compared against the source text before it ships.
- An explicit gap marker when the reviewed sources support nothing for a claim.
How a claim earns its receipt
The run treats your thesis as a set of checkable claims and works through them one by one.
Your written thesis is broken into individual, checkable statements before any source is read.
Company filings and issuer disclosures are prioritized, each tied to an explicit date.
Text quoted in the report is verified against the source document, so a receipt cannot drift from what the filing says.
Claims without supporting evidence in the reviewed sources are recorded as gaps, and retrieval failures are recorded rather than hidden.
What accuracy does not mean here
This page does not publish a headline accuracy percentage, because a single number would hide the thing that matters: whether you can check any individual line. The design goal is that no claim asks for your trust.
- A check cannot turn an unsupported thesis into a supported one.
- The report reflects sources reviewed as of the run date, not a live market view.
- Retrieval depth varies by issuer and jurisdiction; where sources are thin, the report says so.
The sample report is the receipt for this page
Every material claim in the artifact carries a source label, a date, and a locator, so a reader can audit any line instead of trusting the tool.
Evidence summary: The public sample shows receipts on supported claims, on bear-case findings, and on explicit gaps where the reviewed sources supported nothing.
Can I verify a report myself?
Yes. Open any claim in the report, follow its locator to the underlying document, and compare the quoted span with the source text. The public sample report works the same way, so you can run this audit before spending anything.