Skip to content
Join the waitlist

AI Search

AI Citation Rate

AI citation rate is the share of tracked AI answers that cite at least one page of a given site as a source: the answers that cite it divided by all answers recorded for a fixed set of prompts, surfaces and runs. It measures how often a site is used as evidence, which a brand mention rate does not show.

  • Industry term

By QueryRev ResearchPublished Updated5 min read

Why it matters.

A count of citations is not a rate. Bing’s AI Performance report gives the total citations a site receives across Copilot, Bing’s AI summaries and partner integrations, and the average number of its pages cited per day, and says both reflect how often content is referenced, not ranking or placement[1]. A rate needs the answers that did not cite the site as well, and those are seen only through the prompts you track.

Bing’s Citation Share is a different ratio again: a site’s citations as a share of all citations shown for one grounding query, across all sites[2]. It counts citations, not answers, so an answer that lists a site three times weighs three. A citation rate counts each answer once, cited or not; the share against named competitors is AI share of voice.

The unit decides the figure. Counting answers with at least one citation measures reach; counting every cited URL measures weight; counting only the first source measures prominence. Choose one, write it beside the number, and report the others apart if they matter.

Access sets the ceiling. A surface can cite only what it may read: OpenAI says sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers[3], and the same holds for PerplexityBot and Perplexity’s results. A low rate on one surface beside normal rates on the others points to that surface’s crawler first.

From our practice: for British Council, a 288-prompt AI visibility matrix, 216 B2C and 72 B2B prompts, was scored for source presence, placement, inclusion and visibility.

Unread is not the same as no. QueryRev’s study read the robots.txt of 6,294 of the 9,973 domains of its panel, 63.1%, and reports the other 3,679, 36.9%, by cause instead of counting them as open to every crawler[4]. Answers that fail to load or are refused belong outside a citation rate’s denominator in the same way, reported beside it.

How it is measured.

Fix the prompts, the surfaces, the schedule and the number of runs. For each answer, record whether any URL of the site is cited, which URLs and in what order. The rate is the answers citing the site over the answers recorded, per surface and topic.

Count the site, not the page: an answer that cites two of your pages counts once in the rate, and the pages are read separately as cited pages. Count mentions apart, since a name without a link is not an AI citation.

Give the rate its sample. Twelve citing answers of 40 and 300 of 1,000 give the same rate, but the first moves by two and a half points on a single answer.

Measured by

  • Citing answers over answers
  • Rate per surface and topic
  • Runs left out

A worked example.

Sample data: an invented site, to show the method.

A travel insurance site tracks 50 prompts on four AI surfaces, three runs each: 600 answers in a week. 162 of them cite at least one of its pages, a citation rate of 27%. On Perplexity the rate is far below the rest: 9 of 150 answers.

Its logs show PerplexityBot refused with 403 by a bot rule. The other three surfaces, 153 of 450 answers, give the figure to compare with next week; Perplexity is an access problem to fix first.

Citing answers by surface

What to do, step by step.

  1. Define the rate before the first run: the prompt set, the surfaces, the runs per prompt and the counting unit.

  2. Record every answer, including those that cite nothing, with its cited URLs in order.

  3. Leave failed or refused runs out of the denominator, and report how many there were.

  4. Report the rate per surface and topic with its counts beside it, never as one number across everything.

  5. Where one surface lags, check its crawler’s access to the pages you expect cited before changing any content.

  6. Compare periods on the same set and unit, and restate the past when either changes.

Mistakes to avoid.

  • Dividing by prompts instead of answers. Several runs of one prompt are several answers. The denominator is what was recorded.

  • Counting citations, not answers. Three links in one answer make one citing answer. Weight is a separate figure.

  • Leaving failures in the denominator. A refused or empty run is not an answer that passed you over. Count it apart.

  • A rate without its counts. The same rate can rest on twelve answers or a thousand. Show both numbers beside it.

How QueryRev treats it.

QueryRev computes the citation rate from the answer snapshots of each workspace’s tracked prompts, per surface and topic, with the counts and the failed runs shown beside it, and links each cited URL to the page it names. A surface whose rate falls while its crawler is refused is shown as an access finding first. How each figure is defined is set out in the guide to AI search visibility metrics.

Questions people ask.

What is a good AI citation rate?

There is no general benchmark, because the rate depends on the prompts chosen. Compare it with your own past periods on the same set, and with competitors measured on the same answers.

How is AI citation rate different from share of voice?

The citation rate is your own: the share of answers that cite you. Share of voice is relative: your citations or mentions against those of named competitors in the same answers.

Does Bing report an AI citation rate?

Bing Webmaster Tools reports citation counts, cited pages, grounding queries and Citation Share, the share of all citations for a grounding query that went to your site. A rate per answer comes from your own prompt tracking.

Sources and citation.

Primary sources only: the search engines, the AI vendors, standards bodies and QueryRev’s own research. Each is cited in the text by its number.

  1. Introducing AI Performance in Bing Webmaster Tools Public PreviewMicrosoft Bing ·

  2. New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, CompareMicrosoft Bing ·

  3. Overview of OpenAI CrawlersOpenAI ·

  4. State of AI crawler access, October 2026QueryRev Research ·

Cite this definition

A suggested citation. Its link goes to the definition, whose anchor does not change.

QueryRev Research. “AI Citation Rate.” QueryRev AI glossary, 18 October 2026. https://queryrev.com/glossary/ai-citation-rate#definition

Related terms