Source Attribution

What correct attribution looks like, and what fails

Source Attribution is a quality check that sits one layer beneath Citation Rate. Citation Rate asks whether any citation to your domain appeared at all. Source Attribution asks whether that citation actually points to the page the information came from, or whether it credits the wrong page while still getting the facts roughly right. Both failure directions happen in practice. An assistant can state something true about your product and link to an old, unrelated page on your site, giving the reader a citation that does not actually support the claim next to it. It can also link to the correct page while stating something that page does not actually say, which shades into the separate problem of hallucination rather than attribution. A third pattern, no citation at all attached to a specific claim about your brand, is common on assistants that do not display sources by default, such as ChatGPT and Claude outside of browsing-enabled modes. None of these are visible from a simple citation count; each one needs someone to actually read the answer against the page it points to and check whether the two match.

How it works

Checking Source Attribution means going past the yes-or-no question a Citation Rate check stops at, and opening the specific page a citation points to. The check compares the claim made in the AI-generated answer against the actual content of that page: does the page say what the answer claims it says, is the information current, and is it the page most relevant to the specific claim rather than a tangentially related one. This is slower than a simple mention or citation count, since it requires reading rather than pattern matching, which is part of why attribution checks tend to run on a sample of high-value prompts rather than the full prompt set. A recurring pattern worth logging separately is which of your pages keep getting cited incorrectly or not at all for claims that should point to a specific, current page, since that often points to a structural fix: a pricing claim that should cite your pricing page but instead cites a three-year-old blog post usually means the pricing page itself is not written clearly enough, or not accessible enough, for the assistant's retrieval step to find and prefer it over the older content. Fixing that page directly, rather than trying to influence the assistant's behavior indirectly, is usually the more reliable path to a correct citation next time.

Why it matters for AI visibility

Wrong attribution costs a brand twice. It shows the reader a citation that looks like a source but does not actually support the claim, which weakens trust in the answer generally, and it fails to send credit or traffic to the page that should have earned it. A team that only tracks whether a citation exists, without checking whether it is the right one, can believe its AI visibility is stronger than it is, since a wrong citation still counts as a citation in the simplest version of the metric. Attribution problems are also an early signal worth acting on directly: a page repeatedly getting skipped in favor of an older or less relevant one on your own site usually means that page needs a clearer, more direct answer near the top of the content, the kind of fix that also tends to improve traditional search performance at the same time.

Good practices

• Spot-check a sample of citations against the actual page they point to, rather than trusting that a citation is automatically correct.
• Flag any claim about your brand that cites the wrong page, and treat it as a content gap on the correct page, not just a tracking anomaly.
• Prioritize attribution checks on high-value prompts, pricing questions and comparison questions, where a wrong citation does the most damage.
• Write the specific page a claim should point to so it states that fact plainly, since assistants tend to prefer pages that answer clearly over ones that require inference.
• When an assistant attributes a claim to an outdated page, update or consolidate that page rather than leaving both versions live.
• Use a tool such as Truffle's source outreach to flag and follow up on attribution gaps once they are found.

Common mistakes

• Counting any citation as correct attribution, without checking whether the cited page actually supports the claim next to it.
• Missing that a claim with no citation at all is also an attribution failure, not just a citation-rate gap.
• Leaving an outdated page live after a newer one replaces it, which keeps giving an assistant a wrong page to cite.
• Treating attribution as a one-time audit instead of a recurring check, since assistants change which page they prefer to cite over time.

Related terms

• Citation Rate: the share of answers that cite a source at all, a broader metric than whether the citation is correct.
• Hallucination Rate: how often an AI assistant states something false about your brand, a related but separate failure from citing the wrong page.
• AI Overview: a Google feature where source attribution is visible directly in the citations shown next to the answer.
• Brand Mention: the underlying event that Source Attribution checks are performed against.

Frequently asked questions