Receipts Radar
Grounding score + a “receipt demand” list.
Score
Signals
Receipt Demand List
Details
—
Why this exists
The internet rewards confidence, not evidence. Marketers gloat, models hallucinate, and we all share it anyway. Receipts Radar is a cheap lie detector for prose: it doesn’t tell you what’s true—it tells you what needs proof.
How to use it (60 seconds)
- Paste text or Fetch any public URL.
- Hit Analyze.
- Fix the Receipt Demand List: add links, quotes, and numbers that a stranger could verify.
What the score means (0–100)
- 70–100 — Grounded enough to ship. Proof is visible.
- 40–69 — Mixed: some claims naked, some dressed. Add receipts.
- 0–39 — Story mode. Either cite it or cut it.
(We also show your signals: external links per 500 words, numbers per 500, named entities per 500, weasel-word tax, and self-link ratio.)
What counts as a “receipt”
- Primary source (paper, dataset, doc, court filing, company report).
- Exact quote or figure with page/time.
- Direct link near the claim (not 14 paragraphs later).
If it’s just “as reported by BigBlog,” expect a ❌.
What we flag
- Claims that sound measurable (“increase,” “faster,” “only,” “best”) with no nearby link.
- Numbers and percents with no source.
- Weasel words (“industry-leading,” “world-class”) with zero receipts.
- Over-reliance on self-links.
What we ignore (on purpose)
We strip style/script/code blocks, widgets, and nav fluff before scoring. Your CSS won’t tank your grade. Your marketing will.
Fixing your draft fast
- Add the source right next to the sentence.
- Quote the exact line you’re relying on.
- Prefer originals over roundups.
- If you can’t cite it, rephrase as opinion or delete it.
Buttons, explained
- Fetch — pull a page (public HTML).
- Analyze — score the current text.
- Copy Demand List — grab a to-do list of claims to fix.
- Export JSON — save the full result for receipts/CI.
Limitations (read this)
- High score ≠ true. It means cited, not correct.
- PDFs, images, and paywalled content: paste the text.
- Non-English may be patchy; heuristics are English-first.
- We don’t parse tables perfectly—quote the exact figure.
- Some embedded apps block fetch; paste instead.
Privacy & caching
- No iframes. No third-party calls from your browser.
- When you Fetch, the server grabs HTML and caches it for 10 minutes to keep things fast.
- We rate-limit to curb abuse. We don’t sell your drafts. Period.
FAQ
Does it store my text?
No. Pasted text lives in your browser.
What sources “count”?
Links to primary material or a credible report that itself cites primary. Blog chains are weak.
Why did my nav/footer get flagged before?
We used to read CSS. We don’t anymore. If you still see junk, yell at us.
Can I use this in CI?
Yes—hit Export JSON and fail builds under a threshold.
How is the score computed?
Weighted mix of: external citations/500 words, numbers/500, named entities/500, fewer weasel words, and fewer self-links. It’s a grounding score, not a lie detector.