Fact-Check Any Text

Extract every checkable claim from a text, grade each as supported, contradicted, or unverifiable, and get honest limits plus corrected rewrites.

Fact-check the text below. Your job is triage with honest confidence levels, not a verdict on everything, because you can't verify everything, and pretending otherwise is how fact-checks launder errors.

Text:
"""
[PASTE THE ARTICLE, POST, REPORT, OR CLAIM]
"""

1. EXTRACT THE CLAIMS. Number every specific, checkable assertion: statistics, dates, quotes, names, rankings, causal claims ("X caused Y"), superlatives ("first", "largest"). List opinions and predictions separately as NOT CHECKABLE: mixing them into the fact-check inflates the appearance of rigor.

2. GRADE EACH CLAIM:
- SUPPORTED: consistent with well-established information you're confident in. One line on the basis.
- CONTRADICTED: conflicts with well-established information. State the correct version and how confident you are.
- PARTLY TRUE: a real fact, bent: right number/wrong year, real quote/wrong speaker or missing context, true premise/false conclusion. Specify exactly which part fails.
- UNVERIFIABLE (FOR YOU): too recent, too niche, or too specific for your training data. Default here whenever unsure, and for each one, say what to search and what source would settle it (primary document beats press release beats aggregator).

3. THE PATTERN. Step back: do the errors lean one direction, inflating numbers, stripping caveats, sharpening correlations into causes? A directional pattern means persuasion, not sloppiness, and the reader should know which they're dealing with.

4. CORRECTED REWRITES. For every CONTRADICTED or PARTLY TRUE claim: rewrite the sentence accurately, keeping the author's tone. If the accurate version no longer supports the paragraph's point, say so: that's the finding that matters most.

Rules: a claim's source being reputable doesn't make the claim true, grade the claim, not the outlet. Real quotes can still be misattributed or clipped out of context; check both existence and framing. Do not soften verdicts to seem balanced: if eight of ten claims check out, say so; if the central claim is the one that fails, lead with that.

How to use

Without web access this produces a triage list (what to verify and where) which is genuinely more useful than a confident-sounding but unsourced 'all clear'. On models with search enabled, add: 'Use web search for every UNVERIFIABLE claim and link the best source you find.' The pattern step is the sleeper feature: directional bias across many small errors is the strongest signal a text is arguing, not informing.

More business prompts

Build me a monthly budget and a debt payoff plan from these numbers.

Take-home income per month: [AMOUNT, and note if it varies]
Fixed costs: [RENT/MORTGAGE, UTILITIES, INSURANCE, SUBSCRIPTIONS, TRANSPORT, CHILDCARE, with amounts]
Variable spending, last 3 months if known: [GROCERIES, DINING, SHOPPING, ETC.]
Debts: [FOR EACH: name, balan

Budget and Debt Payoff Plan

Turn your real income, bills, and debts into a monthly budget and a payoff schedule with the math shown and the tradeoffs named.

Businessbeginner

Build a 13-week cash flow forecast.

Cash on hand today: [AMOUNT, and which accounts]
Any restricted or committed cash: [AMOUNT AND WHAT IT IS FOR]

Expected inflows:
- Receivables outstanding: [CUSTOMER, AMOUNT, INVOICE DATE, TERMS, and how reliably each pays]
- Recurring revenue: [AMOUNT, BILLING DATE, CHURN ASSUMPTION]
- New sales expe

13-Week Cash Flow Forecast

Build the rolling weekly cash forecast operators actually run on: opening balance, timed inflows and outflows, and the week you run short.

Businessadvanced

Help me set pricing for [PRODUCT/SERVICE].

What it does and who buys it: [PRODUCT + BUYER]
The alternative if they do not buy: [COMPETITOR, IN-HOUSE, SPREADSHEET, DOING NOTHING]
What that alternative costs them: [MONEY, TIME, RISK. Estimate if needed and say so.]
Value we create, quantified if possible: [TIME SAVED, REVENUE GAINED, COST

Pricing Strategy

Work out what to charge and how to package it: value basis, tier structure, the metric you meter on, and how to test before committing.