A reliable fact-checking workflow helps bloggers and indie publishers catch unsupported claims, broken links, misleading figures, and attribution mistakes before readers do. Use the process below to verify a research-heavy blog post, book manuscript, update, or correction review, then keep the claim log so future revisions are faster.
Overview
Fact-checking is more than opening a few search results and adding links. It is a deliberate review of every statement that could be questioned: facts, statistics, dates, quotations, screenshots, product details, historical context, and descriptions of what a study, person, company, or institution has said.
The goal is not to make an article sound overly academic. The goal is to make its important claims traceable and appropriately worded. A useful fact-checking workflow separates three tasks:
- Identify claims: Mark statements that require evidence rather than opinion, interpretation, or common transition language.
- Verify claims: Locate a suitable source, compare the source with your wording, and check whether the evidence supports the full claim.
- Record decisions: Log the source, access date, notes, and any limits so another editor can understand what was checked.
Start fact-checking during drafting instead of waiting until the final hour. A claim that cannot be verified may require a narrower sentence, a different source, or removal from the article. For a repeatable publishing process, add this review to the same editorial calendar used for drafting, editing, and updating. The blogger’s editorial calendar guide can help you assign ownership and deadlines.
What to track
1. Claims and wording
Read the draft once for meaning, then read it again specifically for verifiable claims. Highlight sentences containing numbers, comparisons, superlatives, causal explanations, predictions, legal or safety implications, named organizations, and statements such as research shows or experts agree.
Check whether the source supports the exact strength of the wording. A source that reports an association does not automatically support a sentence saying one factor causes another. A source describing one population, location, or time period may not support a universal statement. When evidence is limited, qualify the sentence with wording such as may, can, in this example, or according to the cited source.
2. Statistics, calculations, and dates
For every statistic, record the original figure, unit, population, geography, time period, and source. Recalculate percentages, averages, and changes when practical. Confirm that you have not confused a percentage with a percentage-point change, or a projected value with a reported result.
Dates need the same care. Check publication dates, update dates, event dates, version numbers, and the date on which a page was accessed. If an article discusses a changing topic, make the time boundary visible in the copy rather than presenting an old observation as current.
3. Quotations and attribution
Compare quotations with the original transcript, recording, interview, book, report, or official page. Preserve the speaker’s meaning and avoid combining fragments in a way that changes the context. Record the speaker’s name, role at the time of the quotation, source location, and any edits made for clarity.
Attribution should answer a reader’s basic question: who said or published this, and where can I check it? Do not attribute a claim to a broad group when the evidence comes from one person or one document.
4. Links, screenshots, and embedded material
Open every link in the published or preview version. Confirm that it reaches the intended page, that the page supports the nearby sentence, and that the anchor text accurately describes the destination. A link can be live and still be a poor citation if it leads to a homepage, search result, or later revision that no longer contains the relevant information.
For screenshots, record the page address, capture date, visible context, and reason the image is included. Check names, labels, timestamps, units, and cropped-out qualifiers. If a screenshot is central evidence, consider linking to the original page as well.
5. AI-assisted research
AI tools can help organize notes, suggest questions, extract terms, or identify claims for review. They should not be treated as the final authority for factual verification. An AI-generated citation may be incomplete, mismatched, outdated, or nonexistent, so open the underlying source yourself.
Keep a clear boundary between discovery and evidence: use AI to help find possible sources, then verify the claim against the original document. Do not paste confidential interview material, unpublished manuscripts, or private data into a tool unless you have permission and understand how the tool handles that material. For a safer editing process, see AI editing prompts for clearer, more original drafts.
A claim-tracking template
Use a spreadsheet, database, or plain-text table. One row per claim is usually easier to audit than one row per source.
| Claim ID | Draft claim | Source and link | Evidence note | Status | Checked |
|---|---|---|---|---|---|
| C-01 | Exact sentence from the draft | Original or authoritative source | What the source confirms or limits | Verified, revise, or remove | Date and initials |
| C-02 | Statistic, quote, or date | Page, section, or timestamp | Units, context, and wording check | Verified, revise, or remove | Date and initials |
Keep the original wording in the log even after you revise the article. That record makes later updates and corrections easier to explain.
How to evaluate sources
| Source type | Best use | Questions to ask |
|---|---|---|
| Primary source | Original data, statement, document, interview, or announcement | Is it genuinely original? What context, method, date, or limitation accompanies it? |
| Institutional or professional source | Definitions, standards, guidance, and background | Who produced it? Is the scope clear, and is the page maintained? |
| Reputable secondary source | Context, reporting, explanation, and comparison | Does it identify its sources and distinguish reporting from commentary? |
| Search result, social post, or AI output | Discovery and leads for further research | Can the underlying claim be verified elsewhere? Is the original source available? |
Source quality is contextual. A first-party page may be the best source for what an organization says about itself, but not necessarily for an independent assessment of its performance. Prefer the source closest to the claim while acknowledging its perspective.
Cadence and checkpoints
Use three checkpoints rather than one large final review.
- During research: Save the original source, note the relevant passage, and record the date. Do not rely on browser history or memory.
- Before publication: Run the complete claim review. Check quotations, calculations, links, screenshots, headings, captions, and metadata. Confirm that citations appear next to the claims they support.
- After publication: Open important links in the live article, review reader corrections, and record any changes. A correction log should state what changed without disguising a material error.
For stable evergreen pieces, schedule a light review monthly or quarterly according to the topic’s volatility. A monthly check may suit articles with frequently changing tools, platforms, or recommendations. A quarterly review may be adequate for slower-moving background content. Refresh source recommendations, platform guidance, AI-search considerations, and examples at least every six months, and check links and citation guidance quarterly.
Connect this work to your wider self-publishing workflow if the content also appears in a book, guide, newsletter, or product.
How to interpret changes
Not every changed page requires a full rewrite. Classify the change before editing:
- Minor change: A broken link, spelling error in a source name, or changed page layout. Fix the reference and note the date.
- Material change: A revised statistic, updated recommendation, new version, or changed definition. Recheck nearby claims and update the publication note.
- Fundamental change: Evidence no longer supports the central conclusion, or the original source has been withdrawn. Rewrite the relevant section, add context, or remove the claim.
Watch for signs that an article needs more than a link replacement: several claims depend on the same outdated source, a recommendation no longer matches the reader’s task, a screenshot omits new context, or comments repeatedly identify the same ambiguity. Treat repeated reader feedback as a prompt to inspect the wording and evidence, not as proof that every suggested correction is accurate.
After editing, repeat the surrounding checks. Changing one number can affect a comparison, conclusion, headline, image caption, newsletter excerpt, or social post. If the article is part of a series, review related pages and internal links as well.
When to revisit
Keep the claim log with the article and set the next review date before publishing. Revisit the fact-check whenever a source changes, a reader reports a possible error, a new edition or version is released, or you reuse the material in another format.
For a practical pre-publish pass, follow this checklist:
- Highlight every factual, numerical, attributed, dated, and comparative claim.
- Find the closest suitable source and open it directly.
- Record the source, relevant passage, access date, and limitations.
- Check that the wording does not claim more than the evidence supports.
- Recalculate important figures and verify units, dates, names, and quotations.
- Test links and inspect screenshots in context.
- Verify AI-suggested information against original sources.
- Mark each claim as verified, revise, remove, or pending.
- Schedule the next monthly, quarterly, or six-month review.
Fact-checking becomes much easier when it is treated as a publishing utility rather than a last-minute rescue. A maintained claim tracker gives you a practical record of what was verified, what may change, and where the next editor should begin.