How Do You Fact-Check a Nonfiction Book Before Publication?

A nonfiction manuscript can contain hundreds of statements that sound completely ordinary until someone asks a difficult question:

How do you know this is true?

The author’s answer might be a government report, an interview recording, a scientific paper, an archived newspaper, or a company filing.

Sometimes, however, the answer is:

“I saw it mentioned in several articles.”

That is where trouble begins.

Professional fact checking a nonfiction book means tracing important claims back to evidence, checking whether the source actually supports the wording, and making sure information has not changed since the chapter was written.

Fact-checking is not limited to statistics. Names, dates, quotations, titles, locations, timelines, case studies, scientific statements, URLs, and even seemingly harmless words such as “first,” “largest,” or “only” can create factual claims.

The goal is not to make the book incapable of containing an error. It is to create a documented process that gives every important claim a defensible source.

Build a Claim Log Before You Start Checking

Reading a manuscript from beginning to end and searching whenever something looks suspicious quickly becomes chaotic.

Instead, create a claim log.

A simple spreadsheet can contain:

For example:

Claim: The company employed more than 40,000 people in 2023.
Source: Company’s 2023 annual report.
Status: Verified.
Note: Figure applies to year-end headcount, not average annual employment.

The note matters.

Without it, an author could later rewrite the sentence as:

“The company employed an average of more than 40,000 people throughout 2023.”

The source may not support that stronger statement.

Prioritize Claims According to Risk

Not every fact deserves the same amount of research time.

A misspelled street name should be corrected, but it generally does not carry the same consequences as incorrect medical advice.

Give greater attention to claims involving:

• Health
• Safety
• Money
• Law
• Identifiable people
• Allegations
• Scientific findings
• Historical controversy
• Political events
• Statistics central to the author’s argument

High-risk claims may require more than one source or a specialist review.

A statement such as:

“This treatment reduces the risk of recurrence by 50 percent.”

needs much stronger evidence than:

“The conference took place in Chicago.”

A useful fact-checking workflow assigns effort according to the consequences of being wrong.

Prefer the Closest Reliable Source

Information gets distorted as it travels.

A research paper is summarized in a press release. The press release becomes a news story. The news story becomes a blog post. The blog post becomes a social media graphic.

By the time the author sees the statistic, its meaning may have changed.

When practical, trace the information toward the original source.

Depending on the claim, this may mean:

• Government database
• Scientific study
• Court filing
• Legislation
• Annual report
• Original interview
• Archival document
• Official transcript
• Survey dataset
• Technical documentation

The Library of Congress defines primary sources as firsthand records or original information, such as letters, diaries, recordings, statistical data, manuscripts, and eyewitness accounts. Secondary sources are created later and interpret or discuss events or information.

Primary sources are often valuable, but they are not automatically correct.

A diary may contain a mistaken memory. A company report may present information from the company’s perspective. An eyewitness can misunderstand what happened.

The fact-checker still needs judgment.

Use Secondary Sources for Context

Strong secondary sources can help explain:

• Why an event matters
• How scholars interpret evidence
• Where experts disagree
• What limitations surround a study
• How an industry has changed

For example, a government document from 1950 may establish what a government department officially stated.

A historian’s later analysis may explain why that document was incomplete or politically motivated.

The two sources answer different questions.

When you have not personally consulted an original source, do not imply that you have. Zotero’s documentation recommends citing the source you actually consulted when a quotation or claim is encountered secondhand.

That keeps the evidence trail honest.

Check Names, Dates, and Titles Against Authoritative Records

Small errors often survive because the author’s memory feels reliable.

Verify:

• Full names
• Middle initials
• Professional titles
• Company names
• Job roles
• Dates
• Locations
• Publication titles
• Organization names
• Awards
• Government offices

Do not assume an old biography remains current.

A person may have changed jobs. An organization may have renamed a department. A company may have merged.

For historical figures, verify which version of the name is appropriate for the period.

Record the accepted form in the book editing style sheet so the copyeditor and proofreader do not introduce new variations later.

Check Direct Quotations Against the Original

Quotations deserve word-for-word verification.

For each one, confirm:

• Exact wording
• Speaker
• Source
• Date
• Context
• Punctuation
• Omitted language
• Translation
• Page number or timestamp

When working from an interview, compare the manuscript against the recording or transcript.

Do not silently convert a paraphrase into quotation marks because the rewritten version sounds cleaner.

If an interviewee said:

“I think we were probably around thirty people then.”

the author should not convert it into:

“We had thirty employees.”

The second version sounds more certain than the evidence.

When polishing spoken language for readability, establish a clear editorial policy and preserve the speaker’s meaning.

Statistics Need More Than a Matching Number

A statistic can be technically accurate and still be misleading.

Check:

• Who was measured
• When
• Geographic area
• Sample size
• Method
• Denominator
• Unit
• Definition
• Margin of error where relevant
• Whether the data was later revised

Suppose a survey says:

“62 percent of respondents prefer remote work.”

Before using it, ask:

• Respondents from where?
• Which occupations?
• How were they recruited?
• How many people responded?
• When was the survey conducted?
• Did “prefer remote work” mean fully remote or hybrid?

Never remove context simply because the shorter statistic sounds stronger.

Use Citation Software to Keep Sources Organized

Reference-management tools can reduce the risk of losing source information during a long nonfiction project.

Zotero, for example, allows writers to collect and organize books, articles, reports, webpages, manuscripts, recordings, statutes, and other research sources. It also supports notes, tags, collections, in-text citations, footnotes, endnotes, and automatically generated bibliographies.

A useful project structure might include collections such as:

• Chapter 1
• Chapter 2
• Interviews
• Statistics
• Historical sources
• Legal sources
• Needs rechecking

Tags might include:

• verified
• disputed
• current
• outdated
• primary source
• expert review needed

Citation software does not decide whether a source is trustworthy.

It helps prevent a trustworthy source from becoming impossible to find six months later.

Version Control Prevents Corrected Errors From Returning

One of the easiest ways to reintroduce an error is to have too many manuscript versions.

A fact-checker corrects the figure in one document.

The author later opens an older draft, copies several paragraphs, and unknowingly restores the incorrect number.

Use one clearly identified master file.

A naming system might look like:

BookTitle_v08_FactCheck.docx

BookTitle_v09_AuthorCorrections.docx

BookTitle_v10_Copyedit.docx

Also maintain a change record for high-risk corrections.

Avoid:

• final.docx
• FINAL2.docx
• final-new.docx
• FINAL-really-final.docx

When several editors or researchers are involved, agree in advance on who controls the master manuscript.

Memoir Requires a Different Kind of Verification

Memory is evidence of personal experience, but memory and documentary certainty are not always the same thing.

A memoir fact-check may compare recollection with:

• Diaries
• Letters
• Photographs
• Calendars
• Emails
• News coverage
• Medical records
• School records
• Travel documents
• Interviews with relatives

When evidence conflicts with memory, the book does not always need to erase the author’s recollection.

It may need more precise language.

For example:

“I remember the storm beginning that afternoon, although newspaper reports date the worst weather to the following morning.”

Or:

“My brother remembers the conversation differently.”

This distinction can make memoir more credible, not less.

Memoirs involving trauma, cultural identity, disability, or another person’s sensitive experiences may also benefit from a sensitivity reader for books, but sensitivity reading does not replace factual verification.

Business Books Must Recheck Time-Sensitive Claims

Business information ages quickly.

Verify close to publication:

• Executive positions
• Company revenue
• Employee counts
• Market share
• Product features
• Pricing
• Acquisitions
• Platform policies
• Industry statistics

Do not write:

“Company X is the world’s largest provider…”

unless the source defines what “largest” means.

It could mean:

• Revenue
• Customers
• Employees
• Transactions
• Market capitalization

Use the measure the evidence actually supports.

Business case studies should also distinguish between correlation and causation.

A company’s revenue increasing after a marketing change does not automatically prove that the marketing change caused the increase.

Historical Books Need Source Comparison

Historical fact-checking goes far beyond dates.

Verify:

• Laws in force
• Political offices
• Geographic boundaries
• Transportation
• Currency
• Technology
• Clothing
• Military ranks
• Language
• Institutions
• Social customs

Primary sources are especially valuable because they provide evidence created by participants, witnesses, or people living close to the event.

But historical records can conflict.

A newspaper account, government report, private diary, and later academic history may describe the same event differently.

Do not hide genuine disagreement.

Language such as:

“Contemporary accounts differ…”

or:

“Later historians have challenged this interpretation…”

can be more accurate than choosing one version without explanation.

Technical Books Need Hands-On Testing

Technical information can be grammatically perfect and completely unusable.

Test:

• Code
• Commands
• Formulas
• Installation steps
• Recipes
• Measurements
• Software menus
• File paths
• URLs
• Equipment settings

Record the exact environment where appropriate:

• Software version
• Operating system
• Hardware
• Browser
• Programming language version

A screenshot showing last year’s interface may make otherwise accurate instructions impossible to follow.

Technical books should receive another check shortly before publication because software changes quickly.

The same applies later during accessible ebook formatting. Tables, code blocks, links, and diagrams need to remain accurate after conversion.

Health Books Need a Higher Evidence Standard

Personal testimonials are not substitutes for clinical evidence.

The National Center for Complementary and Integrative Health advises readers to look for the evidence supporting health information and clearly distinguish research-based claims from testimonials, unsupported statements, and opinions.

Health manuscripts should therefore record:

• Study
• Population
• Study type
• Publication date
• Limitations
• Expert reviewer
• Whether recommendations have changed
• Whether the conclusion applies to the intended reader

Medical or health claims with real consequences should receive review from an appropriately qualified professional.

A general editor should not be expected to determine whether treatment advice is clinically safe.

The NIH also provides tools specifically for evaluating the trustworthiness of science and health information and for communicating research findings and risks accurately.

Check URLs Near Publication

Online references can change between manuscript completion and publication.

For each important URL:

• Open it
• Confirm the page still exists
• Confirm the content still supports the claim
• Check whether a newer version exists
• Record the access date if the citation style requires it

Do not assume that a link working during research will still work six months later.

For essential resources, consider whether a DOI, permanent government page, archival record, or stable publication citation is preferable to a temporary campaign URL.

Bring Specialists in Where the Book Exceeds the Author’s Expertise

A fact-checker does not need to personally know everything.

They need to recognize when the manuscript requires someone who does.

Possible expert reviewers include:

• Physician
• Attorney
• Historian
• Accountant
• Engineer
• Scientist
• Statistician
• Technical specialist

Ask the expert to review specific claims, not simply “approve the book.”

Document:

• What they reviewed
• Which version
• Date
• Corrections
• Remaining disagreements

An expert’s name should not be used as an endorsement without permission.

Fact-Check Again After Editing and Production

Fact-checking does not end when the Word manuscript is approved.

Later stages can introduce errors.

Check the formatted book for:

• Tables
• Captions
• Dates
• Numbers
• Footnotes
• Endnotes
• Cross-references
• URLs
• Image labels
• Bibliography entries

Authors learning how to index a book should wait until print pagination is stable before finalizing page locators.

A separate edition created through large print book publishing will have different pagination and may require revised cross-references and index locators.

Audio production creates another quality-control stage. Before an audiobook narrator audition, prepare verified pronunciation notes for names, technical terms, organizations, and foreign-language material.

The final audiobook production cost can also increase when late factual corrections require pickups or rerecording.

Marketing should use verified information too. A book trailer script can spread an incorrect statistic just as easily as the manuscript can.

Final Fact-Checking Checklist

Before publication, confirm that:

• Important claims are logged.
• High-risk claims receive additional scrutiny.
• Names and titles are verified.
• Dates are checked.
• Quotations match original sources.
• Statistics retain necessary context.
• Primary and secondary sources are distinguished.
• Sources actually consulted are recorded accurately.
• URLs have been tested.
• Citation records are complete.
• One master manuscript is clearly identified.
• Time-sensitive business information has been rechecked.
• Technical instructions have been tested.
• Health claims have received appropriate expert review.
• Unresolved claims are qualified or removed.
• Tables and captions match the verified manuscript.
• Index and cross-references use final pagination.
• Audio pronunciation notes reflect verified terminology.
• Marketing claims match the final evidence.

Good nonfiction does not ask readers to trust every statement because the author sounds confident.

It gives that confidence somewhere solid to stand.

Frequently Asked Questions

Does a Copyeditor Fact-Check Every Statement?

Not necessarily. Some copyeditors perform limited factual checks, but comprehensive fact-checking should be clearly defined in the editorial scope rather than assumed.

Are Primary Sources Always Better Than Secondary Sources?

No. Primary sources provide original evidence, but they can still contain bias, errors, or incomplete information. Strong nonfiction often uses primary evidence alongside reliable secondary interpretation.

Does Every Fact Need a Visible Citation?

Not necessarily. Citation requirements depend on the genre, publisher, and audience. Important claims should still have a traceable source in the author’s research records.

What Should an Author Do When Sources Disagree?

Investigate why they differ. Use the strongest available evidence, explain genuine uncertainty when relevant, and avoid presenting a disputed point as settled fact.

Should Health Books Be Reviewed by a Medical Professional?

When a manuscript makes claims that could influence health decisions, treatment, diagnosis, medication, or safety, review by an appropriately qualified expert is strongly advisable.

What Happens if a Claim Cannot Be Verified?

Remove it, qualify it, clearly attribute it to a source, or explain the uncertainty. An unsupported claim should not become certain simply because it fits the book’s argument.