Fact Check Prompt

Fact Check Prompt

What This Skill Does

Generates a systematic fact-checking checklist from a draft article, categorizing every verifiable claim by type and priority, with the specific verification question, recommended source type, and appropriate verification strategy for each. Handles six distinct claim categories — each with its own methodology, red flags, and common errors.

When To Use This Skill

• You have finished a draft and want to ensure you have not missed any checkable claim before submitting
• You are fact-checking a piece written by another reporter and need a structured inventory of everything that requires verification
• You are on deadline and need to triage — identify which claims carry the most risk and need to be checked first
• You are training a new researcher or fact-checker and want a model for what a systematic check looks like
• You are reviewing an interview-heavy piece and need to verify quotes, titles, credentials, and biographical claims
• You are checking a data-driven article and need to verify statistical claims, survey methodology, and numerical comparisons
• You are editing a piece with legal or regulatory claims and need to confirm statutory citations, court rulings, and regulatory actions
• You are reviewing a story that references scientific studies and need to verify research findings, sample sizes, and journal citations
• You are working on a historical feature and need to verify dates, sequences of events, causation claims, and archival references

What You Need To Provide

For any fact-check

Required: The draft article text (or the section to be fact-checked); the publication type (news, magazine feature, investigative, profile, opinion with reported elements, data journalism piece). Optional: Which claims you are already confident about (to exclude from the list); any specific high-risk areas you are concerned about; the publication's fact-checking standard (light editorial check vs. full verification with source documentation); whether the piece will appear in print (corrections are permanent) or online (corrections can be appended).

Additional inputs by claim category

For interview-heavy pieces: Whether original recordings or transcripts exist; whether the subject has approved or reviewed quotes; whether the subject holds a public or private role. For statistical pieces: Whether raw data or the underlying dataset is available; whether the writer computed figures independently or is citing a secondary source; the date range of the data. For legal/regulatory pieces: The jurisdiction (federal, state, local, international); whether cases cited are concluded or pending; whether the writer has consulted legal counsel. For scientific/medical pieces: Whether the study cited has been peer-reviewed; whether the writer accessed the full paper or only the abstract/press release; whether the findings have been replicated. For historical pieces: Whether claims come from primary sources (archives, documents, firsthand accounts) or secondary sources (books, articles, encyclopedias); the time period involved.

How the Assistant Approaches This

The assistant selects the appropriate verification mode based on the dominant claim types in the draft. Most articles contain claims from multiple categories — the assistant applies each category's methodology to the relevant claims within the same checklist.

Mode 1: Feature Article Claims

Standard mode for narrative magazine pieces, reported essays, and general-interest features.

1. Reads through the draft and extracts every verifiable claim — statistics, dates, names and titles, institutional descriptions, historical facts, legal statements, causal claims ("X caused Y"), and direct quotes attributed to named sources
2. Categorizes claims by risk level: High (statistics, specific numbers, legal or medical claims, anything that could expose the publication to a defamation claim), Medium (biographical details, historical claims, institutional descriptions), Low (widely known facts, claims easily verified in seconds)
3. For each claim: writes the verification question in plain language, recommends the source type, and flags any claim where the draft text itself is ambiguous or internally inconsistent
4. Identifies structural dependencies — claims that rely on each other, where one being wrong invalidates another
5. Closes with a triage recommendation: which High Priority item to verify first, estimated total checking time, and whether to run house-style-enforcer after fact-checking is complete

Mode 2: Interview and Profile Verification

For profiles, Q&A pieces, and any article built primarily on interviews with named sources.

1. Extracts every direct quote and identifies whether it is attributed to a named source, an anonymous source, or paraphrased
2. For each named source: flags their current title, institutional affiliation, credentials, and any description of their expertise — each of these is a separate verifiable claim
3. Checks biographical claims against what can be verified through public records, institutional directories, and published sources
4. Flags "resume inflation" risks — claims about degrees, awards, founding roles, publication credits, or professional affiliations that sources sometimes overstate
5. For direct quotes: notes whether the quote sounds like spoken language or has been cleaned up to the point of misrepresentation; flags composite quotes (combining statements from different parts of an interview) as a separate risk
6. For anecdotes told by sources: identifies which elements are verifiable (dates, locations, other people present) and which rest entirely on the source's account
7. Closes with a recommendation on which sources to re-contact for quote approval and which biographical claims to verify independently

Mode 3: Statistical and Data Claims

For data journalism, trend pieces, economic reporting, and any article that depends on numerical claims.

1. Extracts every number in the draft — percentages, dollar amounts, counts, rates, rankings, growth figures, and dates
2. For each number: identifies whether it comes from a named source (a report, agency, or study), the writer's own calculation, or is unattributed
3. Checks mathematical consistency — do the numbers in the piece add up? If a claim says "32 percent," does the underlying figure support that rounding? If a claim says "tripled," is the actual ratio between 2.5x and 3.5x?
4. Flags comparison traps: numbers from different time periods compared without inflation adjustment, percentages computed from different base populations, rates vs. raw numbers mixed in the same paragraph, cherry-picked start/end dates that exaggerate a trend
5. Identifies methodology questions: was the survey nationally representative? What was the margin of error? Is the sample size large enough to support the conclusion drawn?
6. Checks unit consistency: are all dollar figures in the same year's dollars? Are all geographic comparisons using the same boundaries (city vs. metro area vs. county)?
7. Flags "single-source statistics" — numbers that appear nowhere outside the one report or study cited, which means there is no independent confirmation available

Mode 4: Historical Claims

For historical features, anniversary pieces, retrospectives, and any article that makes claims about past events.

1. Extracts every date, sequence claim ("X happened before Y"), duration claim ("lasted six years"), and causation claim ("X led to Y")
2. Checks dates against multiple independent sources — a single source can contain the same error repeated across secondary works
3. Flags the "first/earliest/only" trap: claims that something was the first, the largest, the only, or otherwise superlative are among the hardest to verify and the easiest to get wrong
4. Identifies anachronism risks — modern terms, titles, or concepts applied to historical periods where they did not exist or meant something different
5. Separates factual claims (what happened, when, where) from interpretive claims (why it happened, what it caused) — both require checking, but by different methods
6. For causation claims: identifies whether the draft asserts direct causation, correlation, or contribution — and whether the evidence cited supports the level of causation claimed
7. Flags "conventional wisdom" claims — widely repeated historical claims that recent scholarship has revised or overturned

Mode 5: Legal and Regulatory Claims

For court reporting, regulatory coverage, policy analysis, and any article citing laws, rulings, or government actions.

1. Extracts every legal citation — statutes, case names, regulatory codes, executive orders, court rulings
2. Verifies the correct citation format: is the statutory section number accurate? Is the case name correctly styled? Is the court identified correctly?
3. Checks whether the law or ruling cited is still current — has it been amended, repealed, overturned on appeal, or superseded?
4. For claims about what a law "does" or "requires": verifies that the draft's characterization matches the actual statutory language, not just a press summary or advocacy group's interpretation
5. Flags pending vs. concluded cases — a case "decided" by a trial court may be on appeal; a law "passed" by one chamber may not have been signed
6. Identifies jurisdictional accuracy: is the correct court, agency, or legislative body cited? Federal vs. state vs. local distinctions matter
7. Flags "legal conclusion" claims — where the draft states a legal outcome as fact ("this violates the law") when it is actually the writer's or a source's interpretation

Mode 6: Scientific and Medical Claims

For health reporting, science journalism, environmental coverage, and any article citing research findings.

1. Extracts every research claim — study findings, sample sizes, effect sizes, journal names, institutional affiliations of researchers
2. Checks whether the study cited has been published in a peer-reviewed journal (vs. a preprint, conference paper, press release, or institutional report)
3. Verifies the study actually says what the draft claims it says — the most common error in science journalism is overstating a study's conclusions
4. Flags the "single study" problem: one study showing an effect is not the same as scientific consensus; checks whether the draft implies broader certainty than a single finding supports
5. Identifies sample size and population concerns: was the study conducted on humans or animals? On a specific demographic that may not generalize? Was the sample large enough for the conclusion drawn?
6. Checks for retracted or corrected papers — especially in fast-moving fields where preprints are common
7. Flags correlation-as-causation errors: the draft says "causes" when the study found "is associated with"
8. For medical claims: flags any claim that could affect health decisions (dosage, treatment effectiveness, risk factors) as highest priority regardless of other risk factors

Claim Type Taxonomy and Verification Strategies

Every claim in a draft falls into one of the categories below. The verification strategy differs by type — applying the wrong strategy wastes time or misses the error.

Identity Claims

What they are: A person's name, title, institutional affiliation, credentials, age, location. Common errors: Outdated titles (a department chair whose appointment ended), wrong institutional name (university renamed or merged), misspelled names, wrong middle initial, wrong age (calculated from birth year without checking whether the birthday has passed). Verification strategy: Check the institution's current directory or website. For public figures, cross-reference at least two independent sources. For titles that rotate (department chair, board president), confirm the appointment is current as of publication date. For spelling, check the person's own published work or institutional bio — do not rely on previous news coverage, which may repeat the same error. Red flags: The source's LinkedIn profile lists a different title than the one in the draft. The institution's website does not list the person at all. The person's name is spelled differently in different published sources.

Numerical Claims

What they are: Any specific number — percentages, dollar amounts, counts, rates, rankings, dates, durations, distances, measurements. Common errors: Transposed digits, wrong unit (millions vs. billions), wrong year, wrong geographic scope (city vs. metro vs. county), percentage vs. percentage point confusion, nominal vs. inflation-adjusted dollars, rounding that changes the meaning. Verification strategy: Trace every number to its primary source — the original dataset, report, or record. Do not verify a number by finding it repeated in another news article; the second article may have copied the error. Check whether the number is current or from a specific reporting period. Verify the unit and the base. Red flags: A number that appears in no source other than the one cited. A percentage that does not specify what it is a percentage of. A dollar figure with no year attached. A ranking with no stated universe ("the third-largest" — third-largest in the state, the country, or the world?).

Quotation Claims

What they are: Direct quotes attributed to a named person, paraphrased statements, and characterizations of what someone said or believes. Common errors: Quotes cleaned up beyond recognition, composite quotes (combining statements from different moments), quotes taken out of context (the surrounding sentences reverse the meaning), quotes attributed to the wrong person in a multi-source piece, fabricated quotes. Verification strategy: Compare the quote in the draft against the original recording or transcript. If no recording exists, confirm the quote directly with the source. For paraphrases, confirm with the source that the characterization is accurate. For historical quotes, trace to the earliest published source — many famous quotes are misattributed or apocryphal. Red flags: A quote that sounds too polished for spoken language. A quote that perfectly supports the writer's thesis (sources rarely speak in thesis statements). A historical quote that appears only in secondary sources with no citation to the original.

Institutional Claims

What they are: Descriptions of organizations — their size, founding date, mission, structure, actions, policies. Common errors: Wrong founding date, wrong headquarters location, wrong employee count (headcount vs. FTE, or an outdated figure), mischaracterized mission or mandate, wrong parent company or corporate structure, describing a proposed action as completed. Verification strategy: Check the organization's own public filings, annual reports, or official website — but verify independently when possible, since organizations sometimes misrepresent their own history. For government agencies, check the enabling statute or executive order. For nonprofits, check IRS Form 990 filings (publicly available). Red flags: A founding date that differs between the organization's website and independent sources. An employee count with no date attached. A description of an organization's purpose that sounds like it was copied from marketing material.

Causal Claims

What they are: Assertions that X caused Y, X led to Y, X resulted in Y, or X is responsible for Y. Common errors: Confusing correlation with causation, omitting confounding variables, asserting direct causation when the relationship is indirect or partial, reversing the direction of causation, implying intent where none is documented. Verification strategy: Identify what evidence the draft cites for the causal relationship. Is it a study that controlled for confounding variables? An expert's opinion? A temporal sequence (X happened, then Y happened)? A temporal sequence alone does not establish causation. Check whether the source cited actually claims causation or merely correlation — the draft may overstate the source. Red flags: "Because of," "as a result of," or "led to" without a cited source or mechanism. A causal chain with missing links. A claim that a single factor caused an outcome that experts describe as multi-causal.

Legal/Regulatory Claims

What they are: Citations to laws, court decisions, regulations, penalties, legal standards, rights, or obligations. Common errors: Wrong statutory section number, wrong court, case overturned on appeal but cited as good law, law amended since the draft was written, characterizing a bill as a law (passed one chamber but not signed), describing a legal standard incorrectly. Verification strategy: Check every statutory citation against the current version of the statute (not a summary or press release). For case law, check whether the decision has been affirmed, reversed, or vacated on appeal. For regulatory actions, verify the current status — proposed rules are not final rules. Use official legal databases or government websites, not news summaries. Red flags: A statutory citation that does not match the subject matter described. A court ruling cited without specifying which court. A law described in present tense ("the law requires") when it has been repealed or amended.

Scientific/Medical Claims

What they are: Research findings, health claims, treatment effectiveness, risk factors, environmental data, study conclusions. Common errors: Overstating a study's conclusions (the study found a correlation; the draft says "proves"), wrong sample size, confusing relative risk with absolute risk, citing a preprint or press release as a published study, citing a retracted paper, applying animal study results to humans without noting the limitation. Verification strategy: Access the actual paper (not the press release or abstract). Verify the journal name, publication date, and authors. Check the paper's own limitations section — what the researchers themselves say their study does not prove. For medical claims, check whether major medical organizations or systematic reviews support the conclusion. Search retraction databases (Retraction Watch). Red flags: A finding described as "groundbreaking" or "the first to show" — these phrases often signal a single unreplicated study. A health claim with no citation. A study from a journal you have never heard of. A finding that contradicts the consensus of major medical organizations.

Historical Claims

What they are: Dates, sequences, durations, descriptions of past events, characterizations of historical figures' actions or motives. Common errors: Wrong year (especially for events near a year boundary — December vs. January), wrong sequence (event A actually happened after event B), "first" claims that are wrong (someone else did it earlier), modernized language applied to historical context, simplified causation (reducing a complex historical process to a single cause). Verification strategy: Cross-reference dates in at least two independent sources — not two sources that cite the same original. For "first" claims, search for earlier instances. For causation claims, check whether current historiography supports the causal story told in the draft. Be alert to the difference between primary sources (documents, letters, records from the period) and secondary sources (later accounts that may contain errors). Red flags: A specific date cited from memory rather than a source. A "first" claim with no documentation. A historical anecdote that is too perfect — it may be apocryphal. A claim that relies entirely on one secondary source.

Verification Chain Methodology

A verification chain tracks each claim from the draft back to its original source, through every intermediary. This section explains how to build and evaluate a chain.

The Chain Structure

Every claim in a published article has a provenance:

Draft text → Writer's source → Intermediary (if any) → Primary source

Example:

• Draft says: "42 percent of teachers reported burnout in 2024"
• Writer got the number from: a newspaper article
• The newspaper got it from: a press release from a teachers' union
• The union got it from: a survey they commissioned from a polling firm
• The polling firm's methodology: online survey of 1,200 union members (not all teachers)

The claim is only as strong as the weakest link. In this example, the draft implies a finding about all teachers, but the primary source surveyed only union members — a specific subset with different demographics and working conditions.

Building the Chain

1. Start at the draft. Copy the exact claim as written.
2. Identify the cited source. What does the draft say is the source? If no source is cited, that is itself a red flag.
3. Access the cited source. Read what it actually says — not what the draft says it says.
4. Trace backward. Where did the cited source get the information? If it is a news article, what report or data did it cite? If it is a report, what methodology did it use?
5. Reach the primary source. The primary source is the original data, document, record, or firsthand account. Stop here.
6. Evaluate each link. At each step, ask: did the intermediary accurately represent what the source below it said? Did anything get lost, simplified, or distorted?

Common Chain Failures

When the Chain Breaks

If you cannot complete the chain — the cited source does not exist, the primary source contradicts the draft, or the intermediary materially changed the claim — flag the claim as UNVERIFIED in the checklist and note exactly where the chain broke. The writer or editor must decide whether to:

• Find an alternative source that supports the claim
• Rewrite the claim to match what the primary source actually says
• Remove the claim from the draft

Red Flags Reference

These patterns signal claims that are likely wrong or unverifiable. When a claim triggers a red flag, escalate it to High Priority regardless of its initial category.

Language Red Flags

Structural Red Flags

Source Red Flags

Output Format

Standard output

Three-section checklist organized by risk level: High Priority, Medium Priority, Low Priority. Each item formatted as:

• Claim: [exact quote or close paraphrase from draft]
• Type: [Identity / Numerical / Quotation / Institutional / Causal / Legal / Scientific / Historical]
• Check: [the verification question in plain language]
• Source type: [what kind of source resolves this]
• Chain status: [Verified to primary / Traced to intermediary / Untraced / Circular]
• Flag: [any red flag, internal ambiguity, or dependency note — omit if none]

Summary at the end: total item count by risk level, estimated checking time, and any structural concerns (single-source dependencies, comparison traps, chain failures). Output ends with a "Next Step" note: which item to verify first, estimated total checking time, and whether to run house-style-enforcer once fact-checking is complete.

For interview-heavy pieces

Add a separate Quote Verification section before the priority checklist. Each entry:

• Quote: [the direct quote]
• Attributed to: [source name and title]
• Verification status: [Recording exists / Transcript exists / Notes only / Unconfirmed]
• Risk: [High if the quote could be disputed; Medium if routine; Low if trivially verifiable]

For data-heavy pieces

Add a separate Numerical Consistency Check section after High Priority. This section verifies that numbers within the draft are internally consistent — percentages add up, time series are in the same units, and comparisons use the same base population.

Quality Criteria

• Every numerical claim in the draft appears in the checklist
• Every named source's title, institutional affiliation, and credentials appear in the checklist
• Every direct quote is flagged for verification against the original recording, transcript, or source confirmation
• High, Medium, and Low categorizations are accurate — medical, legal, or defamation-risk claims never classified below High
• Verification questions are specific enough to hand to a researcher who has not read the draft
• Each claim is tagged with a type from the taxonomy (Identity, Numerical, Quotation, Institutional, Causal, Legal, Scientific, Historical)
• The verification chain status is noted for every High Priority claim
• Internal inconsistencies in the draft are flagged separately from external fact errors
• No claims omitted because they "seem obviously correct"
• Red flags from the reference section are applied — language red flags, structural red flags, and source red flags
• Structural dependencies between claims are identified (if claim A falls, claims B and C also fall)
• Single-source dependencies are flagged in the summary
• For data-heavy pieces, numerical consistency is checked (percentages add up, units match, bases align)
• For interview-heavy pieces, biographical claims are checked independently of the source's own account
• Output includes at least one concrete next action the user can take immediately
• Estimated checking time is realistic (not underestimated to make the list look manageable)

Example 1: Magazine Feature — Remote Work and Commercial Real Estate

Input

Draft type: Magazine feature Fact-checking standard: Full verification with source documentation Excerpt (450 words):

In 2019, roughly 3.6 percent of full-time US workers were working primarily from home, according to the Bureau of Labor Statistics. By 2023, that figure had risen to nearly 28 percent. The shift has reshaped commercial real estate markets, particularly in cities with large technology and financial sectors.

Professor Elena Varga, who chairs the Department of Urban Economics at Northfield University, has been studying downtown vacancy trends since 2018. "We've never seen this kind of structural dislocation in commercial real estate outside of a major war or economic depression," she says. "The difference is that this one was self-inflicted — we had the technology to work remotely before COVID, we just chose not to use it."

The City of Portland, Oregon, saw its downtown office vacancy rate reach 32 percent in the second quarter of 2024, up from 6.5 percent in 2019, according to commercial real estate firm Cushman & Wakefield. City officials have proposed a $45 million adaptive reuse fund to subsidize the conversion of office space to housing. The fund was approved by the Portland City Council in December 2024.

Adaptive reuse is more complicated than it sounds. Most commercial office buildings built in the US between 1970 and 2000 were designed with deep floor plates — sometimes 60 to 80 feet from window to core — that make natural light and ventilation difficult in residential units. A 2022 study from Columbia University's Urban Planning department found that of 1,400 commercial buildings evaluated across 12 US cities, only 17 percent had floor plates suitable for straightforward residential conversion.

The economics are also challenging. The average cost per unit of adaptive reuse conversion in the US was approximately $350,000 in 2023, compared to $220,000 for new construction in a comparable location, according to the National Association of Realtors.

Output

Fact-Check Checklist — Remote Work / Downtown Vacancy Feature

HIGH PRIORITY (verify before publication)

1. Claim: "roughly 3.6 percent of full-time US workers were working primarily from home" (2019) Type: Numerical Check: Is 3.6% the correct BLS figure for full-time remote work in 2019? Confirm methodology — "primarily from home" may have a specific BLS definition that affects comparability with the 2023 figure. Source type: Bureau of Labor Statistics — American Time Use Survey or Current Population Survey; confirm exact series. Chain status: Traced to intermediary — the draft names BLS but does not identify the specific survey. Must trace to the exact dataset. Flag: The 2019 and 2023 figures need to come from the same BLS survey series for the comparison to be valid. If they come from different surveys with different methodologies, the opening comparison is misleading. STRUCTURAL DEPENDENCY: Items 1 and 2 are linked — if one fails, the comparison fails.

2. Claim: "By 2023, that figure had risen to nearly 28 percent" Type: Numerical Check: Is 28% the correct figure? Cross-check against BLS data. "Nearly 28 percent" implies the actual figure is between 26% and 27.9%. If the actual figure is 24%, "nearly 28" is not defensible. Source type: BLS primary data; if from a secondary source, trace to the original. Chain status: Untraced — the draft does not cite a specific report or release for this figure. Flag: Language red flag — "nearly" is imprecise. What is the real number? This must be pinned to an exact figure.

3. Claim: "Professor Elena Varga, who chairs the Department of Urban Economics at Northfield University" Type: Identity Check: Confirm current title ("Professor" — assistant, associate, or full?), confirm she currently chairs the department (chair appointments rotate, often every 3-5 years), confirm the department is called "Urban Economics" (not "Urban Studies" or "Urban Planning and Economics"). Source type: University directory or direct confirmation with source. Do not rely on the source's own description — verify independently. Chain status: Untraced — identity claims require independent verification. Flag: Resume inflation risk. "Chairs" must be current as of publication date.

4. Claim: Portland downtown office vacancy "reach 32 percent in the second quarter of 2024, up from 6.5 percent in 2019, according to Cushman & Wakefield" Type: Numerical Check: Verify both figures from the cited Cushman & Wakefield report. Confirm the Q2 2024 report exists. Confirm 32% is the figure for downtown Portland specifically (not metro-wide or the Portland-Vancouver MSA). Source type: Cushman & Wakefield market report — request the specific report by name and quarter. Chain status: Traced to intermediary — the draft names the firm but not the specific report title or release date. Flag: Geographic scope trap. "Downtown" vacancy and "metro area" vacancy are different figures. Confirm the geographic boundary used in the report matches the draft's claim.

5. Claim: "$45 million adaptive reuse fund" approved by Portland City Council in December 2024 Type: Institutional Check: Confirm fund amount ($45 million — not $45 million over multiple years?), approval date (December 2024 — what day?), and council vote result. Is it a fund, a budget allocation, a bond measure, or a proposed line item? Is it fully approved or contingent on a future vote? Has the amount been revised since December? Source type: Portland City Council meeting minutes, December 2024. Official city press release. Council voting record. Chain status: Untraced — no specific source cited in draft. Flag: The draft says "proposed" in one sentence and "approved" in the next. Which is it? If the fund was proposed by officials and then approved by the council, the draft should say so clearly. If it was only proposed, the draft is wrong to say it was approved.

6. Claim: Columbia University study found "17 percent had floor plates suitable for straightforward residential conversion" — 1,400 buildings across 12 cities, 2022 Type: Scientific Check: Confirm the study exists. Verify: publication year (2022), department (Urban Planning — or is it "Urban Planning and Design"?), sample size (1,400 buildings), geographic scope (12 cities — which 12?), and the 17% finding. Access the actual paper — do not verify from a press release. Source type: Columbia University — identify the specific study title, authors, and publication venue (journal, working paper, or institutional report). Chain status: Untraced — the draft does not name the study title or authors. Flag: Source red flag — the draft references a "department" finding, not a specific published paper. Is this a peer-reviewed study, a working paper, a student thesis, or an institutional report? The evidentiary weight differs.

7. Claim: "average cost per unit of adaptive reuse conversion... approximately $350,000 in 2023... compared to $220,000 for new construction" — attributed to National Association of Realtors Type: Numerical Check: Confirm both figures from NAR. Verify "comparable location" methodology — the comparison is only valid if both figures use the same geographic markets. Confirm whether these are median or mean figures. Confirm whether they include land costs. Source type: NAR research report — confirm the specific publication title, date, and methodology section. Chain status: Traced to intermediary — the draft names NAR but not the specific report. Flag: Comparison trap. "$350,000 vs. $220,000" only means something if both figures are computed the same way, in the same markets, in the same year's dollars. If the adaptive reuse figure includes historic preservation requirements and the new construction figure does not, the comparison is misleading.

MEDIUM PRIORITY

8. Claim: Varga "has been studying downtown vacancy trends since 2018" Type: Identity Check: Confirm with source. Is 2018 accurate, or was it a casual approximation in conversation? Does her published work support this — does she have publications on vacancy trends from 2018 or earlier? Source type: Direct confirmation with source. Cross-reference with her publication record. Chain status: Untraced.

9. Claim: US office buildings "built... between 1970 and 2000 were designed with deep floor plates — sometimes 60 to 80 feet from window to core" Type: Historical / Numerical Check: Confirm this is an accurate architectural generalization. Is "1970 to 2000" the correct period for deep-floor-plate construction? Are the 60-80 foot figures sourced from the Columbia study, from architectural references, or from the writer's own description? Source type: Architecture or real estate reference; or trace to the Columbia study if that is where the figures originated. If the writer paraphrased a general architectural principle, it still needs a source. Chain status: Untraced.

10. Claim: The shift "has reshaped commercial real estate markets, particularly in cities with large technology and financial sectors" Type: Causal Check: The causal claim ("reshaped") is a characterization — is it supported by the evidence presented later in the piece? The narrowing to "technology and financial sectors" implies these cities were disproportionately affected. Is that supported by data, or is it the writer's inference? Source type: Industry reports comparing vacancy rates across city types, or expert attribution. Chain status: Untraced — no source cited for this characterization.

LOW PRIORITY

11. Claim: "The City of Portland, Oregon" Type: Identity Check: Verify city and state — Portland, Maine is a common confusion in copy. The piece discusses commercial real estate; both Portlands have relevant markets. Source type: Internal consistency — the Cushman & Wakefield report will confirm the geography. Chain status: N/A

QUOTE VERIFICATION

…(truncated)