Unsourced Claims: Why Citations Matter and How to Fix It

TL;DR The "Unsourced Claims" check fires when a page states facts, statistics, or strong assertions with no citation, link, or attribution behind them. This is a provenance problem, not an age one: the claim may be perfectly current and still show nowhere it came from. Add a credible source next to the claim — a linked study, a named dataset, first-hand data, or an authority — and the flag clears.

Unsourced claims are where a lot of otherwise-good content quietly loses credibility. "Studies show 70% of buyers do X" with no link is a coin flip on whether the reader believes you — and increasingly, on whether an AI engine will cite you.

What this check flags

The crawler and reviewers look for load-bearing statements presented as fact without support: a percentage with no source, "experts agree" with no expert named, a comparison stated as settled with nothing to check it against, or a numeric claim that appears out of thin air. When claims like these carry no citation, link, or attribution, the issue fires. It is flagging where the information came from — a different problem from whether the information is stale, which the outdated-content check handles.

A real example and the fix

A B2B software blog led a flagship post with "73% of companies that adopt marketing automation see ROI within the first year" — no link, no study named, no year. That single unsourced stat undercut an otherwise sharp article, because a skeptical reader (or a fact-checking rater) has no way to verify it, and an AI engine choosing what to cite has nothing to grab. The fix was simple and honest. The team traced the number to a specific vendor report, linked it inline, and named it in the text: "according to [named]'s 2023 automation benchmark." Where a claim could not be sourced, they either removed it or softened it to reflect what they could actually support. They also promoted their own first-hand data — anonymized results from their client base — because original data you can stand behind is the strongest citation of all. The claims went from "trust me" to "here, check it," and the piece started showing up in AI answers that cite sources.

Source types ranked by credibility

Sourcing vs. freshness — and why AI raises the stakes

Do not confuse this with outdated content. Unsourced claims are about provenance: the assertion never showed its origin. Outdated content is about age: the information was sourced fine and has simply gone stale. A page can be brand-new and completely unsourced, or meticulously cited but years out of date. This check is the provenance side; for the age side, see Outdated Content.

Sourcing sits inside E-E-A-T under Trust and Authoritativeness — citing credible sources is one of the clearest ways to signal that claims are verifiable and the author did the work. As always, E-E-A-T is a rater framework, not a direct ranking factor; there is no "citation score" wired into the algorithm. What has changed the calculus is AI search. When Perplexity, ChatGPT, and Google's AI assemble an answer, they lean toward content that is itself well-sourced and toward pages that are citable sources. Unsourced assertions are both less trusted and less likely to be pulled into an AI answer — a real cost in a search world moving toward cited generation. See Citation Optimization for the GEO angle.

How to detect it

1. Read your page and highlight every statistic, percentage, and "studies show / experts agree" statement.
2. For each highlight, ask: can a reader click through or look this up? If not, it is unsourced.
3. Trace secondhand citations upstream — "another blog said" is not a source; find the original study it borrowed from.
4. Flag absolute claims ("the best," "always," "never") that assert a fact without anything to back them.
5. Check whether any of your strongest claims are actually your own data you have not surfaced — that is a citation you already own.

How to fix it

Go claim by claim. For each unsupported statement, link a credible primary source inline and name it in the text so it reads as deliberate, not decorative. Where you cannot find a real source, either remove the claim or soften it to what you can defend. Prefer primary sources over blogs that merely repeat a number — chase the figure to its origin. Surface your own first-hand data where you have it, because original evidence is the strongest citation you can offer. Then recrawl and confirm the flag clears. For how sourcing feeds AI answers, see how AI engines choose sources.

FAQ

Does every sentence need a citation?

No. Common knowledge and your own clearly-labeled opinion do not need sources. Citations are for facts, statistics, and strong claims a reader would reasonably want to verify. Over-citing obvious things just adds clutter.

What if I cannot find a source for a claim I believe is true?

Then remove it or soften it to what you can support. An unverifiable assertion, however true it feels, is a liability. It is better to make a smaller claim you can back than a bold one you cannot.

How is this different from outdated content?

Unsourced claims is about where information came from; outdated content is about whether it has aged. A fresh page can be entirely unsourced, and a well-cited page can be badly out of date. They are separate fixes.

Is linking to another blog good enough?

Usually not, if that blog is itself just repeating a number. Trace the claim to its primary source — the study, dataset, or official figure — and cite that. Secondhand citations inherit secondhand credibility.

Why does sourcing matter more with AI search?

AI answer engines favor content that is well-sourced and that functions as a citable source itself. Unsourced claims are less likely to be trusted or pulled into an AI-generated answer, which is an increasingly real cost to visibility.

Related reading: No Editorial Policy covers the process that produces well-sourced content in the first place, and the featured snippets guide shows how citable, verifiable content earns machine attention.