Skip to main content
Citation rate is the percentage of AI answers that mention your brand which also cite a page on your own domain. It answers one question: when an AI talks about you, does it link you? The denominator is what makes it useful. It is not every answer we run — it is only the answers where your brand came up. That turns citation rate into a conversion metric: you have already won the mention, and this tells you how often that mention comes with a link back to you.
This definition changed on 14 August 2026. Citation rate previously meant “answers containing at least one citation to any domain, out of all answers” — a property of the platform and the category rather than of your brand. Numbers from before and after that date are not comparable. See what changed.
How often a mention also earns a link to your site.
Answers citing you ÷ answers mentioning you

​
Why it matters

Citations drive traffic. When an AI answer links your domain, some fraction of readers click through. A mention without a citation is awareness; a mention with a citation is awareness plus a path back to you. Citations also signal authority. When an AI cites your domain instead of a review aggregator or a news article, it is making an editorial judgment that your content is a good source for that prompt. Sustained citation presence on a topic is one of the strongest signals of AEO/GEO success. Because the denominator is mentions, citation rate isolates a problem you can act on. A low number does not mean the AI has never heard of you — it means it knows you and still prefers someone else’s page as the source.

​
Formula

Inputs:
  • numerator — answers that mention your brand and cite a page on a domain you own. Counted once per answer, however many of your pages it links.
  • denominator — answers that mention your brand at all, cited or not.
Both are counted per answer, not per link, so an answer that cites five of your pages counts once. Citation rate is available per platform, per topic and per prompt.

​
Example calculation

A brand tracks 60 prompts. In one week they produce 400 answers. The brand is named in 120 of them, and 42 of those 120 also link a page on the brand’s own site.
Read it as: the AIs mention this brand in 120 answers, and in just over a third of those they also send the reader to the brand’s own site.

​
Citation rate is not a subset of mention rate

It is a rate conditional on a mention, which is a different thing. The two multiply cleanly:
In the example above: 30% mention rate × 35% citation rate = 10.5% of all answers cite the brand’s own domain. This identity holds exactly, so a citation figure measured against all answers can always be recovered. The gap between mention rate and citation rate is diagnostic:
  • High mention rate, low citation rate — the AI knows you from training data and third-party sources but does not treat your own site as a quality source. Focus: on-site content quality, schema markup, technical AI readability.
  • Both high — the AI knows you exist and treats your content as authoritative. Maintain freshness, expand topic coverage.
  • Citation rate at zero across platforms — either your content is not reachable by AI crawlers, or the platform is filtering your domain out of source selection. Check your robots.txt and llms.txt and read Fix your llms.txt and robots.txt.

​
Citation rate vs. citation share

They are easy to confuse. Put the formulas side by side and they stop being confusable: Citation rate reads like a conversion rate and typically lands in the tens of percent. Citation share reads like a market share and typically lands in low single digits. Across tracked brands the two are close to uncorrelated, and the ranking inverts at the extremes — a brand can lead its category on share of voice and still sit near the bottom on citation rate. The scale difference is expected, not a fault. Citation share is small because the denominator is every citation in the category, most of which will never be yours.

​
Which platforms this covers

Citation rate is tracked across all five platforms Cited monitors: Google AI Overviews, Google AI Mode, ChatGPT, Gemini and Perplexity. Citation extraction is mature on all five. Measured over a 30-day window across tracked brands:
Perplexity is not the best platform to judge your citation rate on, despite citing heavily. When Perplexity does cite it cites densely — about 10 sources per answer, the highest of the five — but it attaches citations to under half its answers, the lowest share of the five. It also very rarely picks a brand’s own domain: measured own-site citation rate on Perplexity sits near zero for every brand we track. A brand reading its Perplexity citation rate alone will conclude it has a content problem when it has a platform difference.Earlier versions of this page said Perplexity was the most reliable platform for citation data. That was true of extraction in 2025 and is no longer a useful guide to where you actually get cited.
Because per-platform spreads of 30 to 60 points are normal within a single brand, always read citation rate per platform before reading the cross-platform number. A change in the blended figure can come entirely from a shift in which platforms mention you.

​
What changes this metric

  • AI-readable content structure — clear headings, structured data and clean HTML are cited more than pages behind heavy JavaScript rendering
  • Robots.txt and llms.txt configuration — blocking AI crawlers prevents citation regardless of content quality; allowing them is a prerequisite, not a guarantee
  • Source authority — inbound links, domain age, content depth and topical focus
  • Content freshness — platforms favour recently-updated pages, especially on prompts about pricing, features and comparisons
  • Schema markup — helps platforms parse your content into citation-ready summaries

​
Sample size

Citation rate is a ratio on a denominator that can be small. On a narrow topic or a single prompt, a brand mentioned twice and cited once reads 50%, which is noise rather than a finding. Cited shows the underlying counts alongside the percentage so you can see how much weight a number carries, and reserves an em dash for “no mentions at all” — genuinely different from a measured zero.

​
What changed in August 2026

Until 14 August 2026, citation_rate in the Cited dashboard, API and CSV exports meant:
That number sat at 83.5% platform-wide and spanned only 70–97% across brands — it described how citation-rich a category and platform are, not anything about the brand, while carrying a name that customers and this documentation both read as an own-domain figure. The replacement spans 0–70% on the same brands. If you have exports or reports from before that date, the column labelled “Citation Rate” holds the old formula. Do not compare it with anything produced afterwards.

​
How Cited measures it

Every AI answer is parsed for its cited source URLs. A citation counts toward your numerator when the cited domain is one you own — resolved at read time from your configured domains, so adding a domain corrects history rather than only affecting future runs. Ownership resolution is shared with citation share, so the two metrics can never disagree about which domains are yours. Citation rate is a property of a brand on an answer. It is therefore not available per source: a single cited page has no “citation rate”, because the denominator — answers mentioning your brand — has no source dimension. Use citation share for a per-source view.

​
Frequently asked questions

Because it is a different measurement. The old number counted answers carrying a citation to any domain at all, so it was high for almost everyone — 83.5% platform-wide. The new one counts only answers that cite your domain, out of answers that mention you. Nothing about your visibility changed on the day the definition did.
Different denominators. Citation rate divides by the answers that mention you; citation share divides by every citation in those answers, to any domain. The category’s source pool is large and most of it will never be yours, so share is small by construction. Comparing the two numbers directly is a category error.
Three common causes: your robots.txt or llms.txt is blocking AI crawlers; your content is not structured in a way platforms can parse into citations, for example a fully JavaScript-rendered app; or the AI is citing third-party coverage about you — review sites, press — rather than your own domain. Verify crawler access first, then audit content structure. Check per platform before concluding: near-zero on Perplexity alone is normal and is not a content problem.
Yes, and that is the point of the metric. Mention rate is the denominator, so citation rate measures purely how well existing mentions convert into links to you. Raising it is content and technical work on the pages the AI already has reason to reference.
No, and this is deliberate rather than a gap. Citation rate’s denominator is the answers that mention your brand, which has no source dimension — a page cannot “have” a citation rate. The API returns an error rather than a plausible-looking number if you ask for one. Use citation share, which is defined per source.
Not yet. Citation rate is tracked in customer dashboards across all five platforms; percentile benchmarks in the Cited Index are still to come. The earlier blocker on this page — that only one platform produced usable citation data — no longer applies; see the platform table above.