Writing
List the claims in a draft that need a source
Helps editors identify claims in a draft that need sources, suggest source types, and flag vague, conflicting, or potentially outdated statements.
by Appndo
Tested on GPT-6.1 Sol, Claude Opus 5.5, Gemini 3.8 Flash · Verified Nov 2, 2026
The prompt
Go through this draft and list every claim a careful editor would want a source for.
Draft:
{{draft}}
Who will read it, and how careful it needs to be:
{{context}}
1. Quote each claim exactly as written, in the order it appears. Include numbers, dates, firsts and superlatives ("the largest", "the first"), quotes attributed to people, cause-and-effect statements, and general statements presented as fact ("most people…", "studies show…").
2. For each claim give the kind of source that would settle it (official statistics, the company's own filing, the original study, the person quoted) and how hard it will be to verify: easy, takes digging, or probably can't be verified as written.
3. Flag claims too vague to verify, and suggest a more precise wording to check instead.
4. Flag claims that contradict each other or may be out of date.
5. Skip opinions clearly marked as the author's, and say which you skipped.
Don't tell me whether a claim is true: you're listing what to check, not checking it. If you know a claim is often misquoted or misattributed, say so, but still mark it as something to verify.How to use it
Paste the text into {{draft}} and describe its audience and required level of scrutiny in {{context}}. Use the resulting list to plan source checks; it does not verify whether claims are true.
Example
Inputs
- draft
- Why remote teams get more done The office is no longer the default. In 2019, only 5% of Americans worked mainly from home; today that number has more than quadrupled. And the shift is paying off. A landmark Stanford study found that remote workers are 13% more productive than their office-based colleagues. It isn't just productivity. When the UK ran the world's largest four-day week trial in 2022, 92% of companies decided to keep the shorter week. Most managers now agree that trust, not presence, drives results. Of course, remote work only works with the right tools. Zoom, founded in 2011, grew from 10 million daily meeting participants to 300 million in a matter of months. Since we launched our async video feature, customers tell us their teams hold 40% fewer meetings. Before the pandemic, a quarter of the workforce already worked from home at least occasionally. As Einstein said, "In the middle of difficulty lies opportunity." I think offices are overrated, and the data backs me up.
- context
- Blog post for a remote-work software startup. Marketing wants it accurate enough that a journalist or a competitor can't call us out.
Output
These are the claims to source, in order. Difficulty refers to verifying the wording as written, not whether the claim is true.
| # | Exact claim | Source needed | Difficulty and editorial flags |
|---|---|---|---|
| 1 | “Why remote teams get more done” | Original comparative research measuring team output, with comparable roles and working conditions. | Probably can't be verified as written. The headline asserts a broad productivity advantage and promises a causal explanation. Specify which teams, compared with whom, and what “more” measures. Suggested wording to check: “In [study], [type of remote team] produced [measure] more than [comparison group].” |
| 2 | “The office is no longer the default.” | Representative official statistics on work arrangements for a defined population and date. | Probably can't be verified as written. “Default” could mean the most common arrangement, employer policy, or employee preference. Suggested wording: “In [country/year], [percentage] of [defined workers] worked primarily outside an employer’s workplace.” |
| 3 | “In 2019, only 5% of Americans worked mainly from home;” | Official 2019 statistics, including the survey question, population, and definition of working mainly from home. | Easy. Clarify whether “Americans” means all residents, employed people, or a narrower group. “Only” adds editorial framing. |
| 4 | “today that number has more than quadrupled.” | The latest available statistics using a definition and population comparable to the 2019 baseline. | Takes digging. Replace “today” with a dated reporting period and give the actual percentage. Check whether “number” means headcount or percentage; the preceding claim gives a percentage. This claim will age. |
| 5 | “And the shift is paying off.” | Research establishing benefits attributable to the shift to remote work, with defined outcomes and comparison groups. | Probably can't be verified as written. It implies a causal benefit without identifying who benefits or how. Suggested wording: “In [study], switching to [defined arrangement] increased [outcome] by [amount].” |
| 6 | “A landmark Stanford study found that remote workers are 13% more productive than their office-based colleagues.” | The original study and its methods, results, author affiliations, and publication details; independent evidence of influence if retaining “landmark.” | Takes digging. Check the 13% figure, productivity measure, study population, work arrangement, and comparison group. “Landmark” needs justification or removal. The sentence presents a particular study’s result as a general, present-tense fact; suggested wording: “A [year] study of [population] found [precisely defined result].” |
| 7 | “It isn't just productivity.” | Evidence for an additional benefit of remote work beyond productivity. | Probably can't be verified as written. In context, this extends the claimed benefits of remote work. Name the benefit and supporting research. The following four-day-week example needs a separate explanation of its relevance to remote work. |
| 8 | “When the UK ran the world's largest four-day week trial in 2022,” | Original trial documentation identifying organizers, location, dates, and scale; comparative records of other trials to support “world's largest.” | Takes digging. Check who “ran” it: the wording could imply government sponsorship. Define “largest” by companies, participants, or another measure, and specify “largest as of [date].” A trial report alone may not settle the worldwide superlative. |
| 9 | “92% of companies decided to keep the shorter week.” | The original trial’s results or follow-up report, including the denominator, response rate, decision date, and definition of continuation. | Easy. Specify “participating companies” if that is the intended population. Distinguish continuing at the end of the trial from permanent adoption; later status requires a dated follow-up. |
| 10 | “Most managers now agree that trust, not presence, drives results.” | A representative, dated survey of a defined manager population, with exact questions and response percentages. A causal study would additionally be needed to establish what actually drives results. | Probably can't be verified as written. “Most,” “now,” “trust,” “presence,” and “results” lack definitions. Suggested wording: “In a [date] survey of [population], [percentage] agreed that [exact survey statement].” Agreement alone would establish a belief, not the causal proposition. |
| 11 | “Of course, remote work only works with the right tools.” | Research defining successful remote work and demonstrating that specified tools are necessary for it. | Probably can't be verified as written. “Only” asserts necessity; “works” and “right tools” are undefined. Suggested wording: “In [study], teams using [tools] achieved [measured outcome].” |
| 12 | “Zoom, founded in 2011,” | Company incorporation records, an official company history, or regulatory filings. | Easy. Confirm whether the date refers to founding, incorporation, or product launch. |
| 13 | “grew from 10 million daily meeting participants to 300 million in a matter of months.” | Zoom’s original dated announcements or filings, including metric definitions and any corrections. | Easy. Give both dates and ensure both figures use the same metric. “Daily meeting participants” should not silently become unique daily users. The metric is a known source of reporting confusion, so verify its wording carefully. |
| 14 | “Since we launched our async video feature, customers tell us their teams hold 40% fewer meetings.” | Internal launch records and the underlying customer survey, interviews, or analytics, including sample size, questions, baseline, comparison period, and calculation. | Takes digging. Establish whether 40% is an average, a selected customer’s result, or another statistic. Separate customer reports from measured usage. The placement implies the feature caused the reduction; that needs additional evidence. Suggested wording: “In a [date] survey of [number] customers using [feature], respondents reported [defined reduction] compared with [baseline].” |
| 15 | “Before the pandemic, a quarter of the workforce already worked from home at least occasionally.” | A representative pre-pandemic survey or official statistics defining geography, workforce, reference period, and frequency. | Takes digging. “Before the pandemic” and “at least occasionally” are too loose for a reproducible check. Suggested wording: “In [year], [percentage] of [country’s defined workforce] worked from home [specified frequency].” |
| 16 | “As Einstein said, "In the middle of difficulty lies opportunity."” | A traceable primary source: Einstein’s published writing, correspondence, speech transcript, or an authenticated archival record, with original wording and translation where relevant. | Probably can't be verified as written. This quotation is often flagged as misattributed to Einstein. Verify the attribution rather than relying on quotation websites; remove the attribution if no reliable source can be found. |
| 17 | “the data backs me up.” | Clearly identified research supporting a defined proposition about offices, with relevant limitations. | Probably can't be verified as written. “Offices are overrated” has no measurable benchmark, and “the data” is unspecified. Replace with a narrow finding, such as “[Study] found [outcome] under [conditions].” |
Cross-claim issues to resolve:
- Rows 3–4 versus row 15: “mainly” and “at least occasionally” describe different groups, so the figures are not necessarily contradictory. Make their populations, dates, and frequency definitions explicit before comparing them.
- Row 2 versus rows 3–4: A rise in home working does not by itself establish that offices have ceased to be the default. Define and source that conclusion separately.
- Rows 8–9 versus the remote-work argument: A shorter working week and a remote working arrangement are different variables. The trial’s relevance to remote-work benefits needs evidence.
- Time-sensitive wording: Replace “today” and “now” with dates. Date the trial continuation decisions and customer meeting-reduction reports as well.
- Generalization and causation: A particular study, managers’ beliefs, and customer reports each support different kinds of claims. Check that the headline and surrounding conclusions do not exceed what those sources can establish.
Skipped as explicitly marked opinion: “I think offices are overrated”. The attached assertion that data supports that opinion remains listed above.
One real run with the example above. Results vary between runs and models.