Framework How-To Guide aiknowledgesignal.io

How to Fact-Check and Link-Check

One unverified statistic or one dead link is enough to mark a page as careless. Confirm every source exists and says what you claim, and that every link resolves, before and after publish.

Why it matters

Accuracy is a survival trait. Readers, and the systems that rank and quote content, downgrade content with wrong statistics, misquoted research, or broken links — and a single bad citation can taint trust in everything around it. Worse, an unverified "AI usage is 80% globally" can be lifted and amplified by a model, turning your error into a widely-repeated one.

Fact-checking and link-checking is the maintenance discipline that keeps a page citable over time: sources confirmed, dates current, links live, and primary sources preferred over blog-about-a-blog chains.

What good looks like

Every claim and link on the page is verified and current:

Step by step

  1. Verify every source says what you claim Open each source and confirm it actually supports your statement, with the right number and context. Replace any misquoted research with the exact quote; replace "experts say" with a named source + link + year (Phase 4.4).
  2. Check dates and freshness Confirm each statistic is current. Prefer data from the last ~24 months and flag anything older for review or replacement (foundational studies excepted) — stale figures undercut otherwise-strong content.
  3. Test every link Click or crawl every link. Fix 404s, and replace redirect chains (multiple hops) with the direct canonical URL. Broken and chained links signal neglect to crawlers.
  4. Prefer primary sources and add outbound references Swap "blog linking to another blog" for the primary paper, dataset, or official document. Include 2–3 high-quality outbound references per page — outbound links to authoritative sources are a trust signal, not a leak.
  5. Update internal links after restructures When pages move or are renamed, fix the internal links that point to them. Keep the knowledge base (Phase 4.2) and sitemap/llms.txt (Phase 5.5) consistent with the current structure.

Worked examples

1 Before / after: common issues and fixes
IssueFix
"AI usage is 80% globally" (unverified)"72% of organisations surveyed had adopted AI in at least one business function (McKinsey, The state of AI, 2024)" — verified against the report, linked.
Outdated statistic (2019)Recent statistic (last ~24 months), or the old one clearly dated.
Misquoted researchExact quote with correct context.
"Experts say…"Named source + link + year.
Broken link (404) / redirect chainTested working, direct canonical link.
Blog linking to another blogPrimary source (paper, dataset, official doc).
No outbound links2–3 high-quality references per page.
2 A pre-publish verification pass

The routine to run on every page before it ships and on a refresh cycle after.

[ ] Every statistic traced to a named, dated source
[ ] Each source re-read to confirm it says what we claim
[ ] Statistics from the last ~24 months (older ones flagged)
[ ] All links return 200, no redirect chains
[ ] Primary sources linked, not secondary coverage
[ ] 2-3 high-quality outbound references present
[ ] Internal links point to current page URLs

Common mistakes

MistakeFix
Citing a statistic you didn't re-read in sourceOpen the source and confirm the number and context.
Leaving old statistics unflaggedPrefer the last ~24 months; flag and review older figures.
Shipping with 404s or redirect chainsTest all links; use direct canonical URLs.
Linking blog coverage instead of the sourceLink the primary paper/report/dataset.
Checking links by handRun a link checker (e.g. lychee or linkinator) in CI or on a schedule.
Key source disappearsSave an archived copy (e.g. the Wayback Machine) of important sources and link it alongside.
Stale internal links after a restructureUpdate internal links to the current structure.

Limits of this advice

Checklist

  • Every source exists and says what the page claims
  • Statistics are current (last ~24 months; older ones flagged)
  • No 404s and no redirect chains
  • Primary sources linked; 2–3 quality outbound references per page
  • Internal links updated to the current structure