How to Use AgentSignal AI Citation Rate: Check Whether AI Answers Cite You

Published Updated 8 min read
How to Use AgentSignal AI Citation Rate: Check Whether AI Answers Cite You

Learn how to register questions in AgentSignal AI Citation Rate, calculate the credits you need, and read the results table and answer logs. We also explain how it differs from the AI reports in Search Console and Bing.

AgentSignal’s AI Citation Rate asks AIs your registered questions with web search and counts the share of answers that cite your site as a source. Open it from the AI Citation Rate project list (login required).

If someone says, “ChatGPT recommended a competitor, but not us,” start by recording that exact question and testing it. Search Console and Bing now have AI reports too. This article also explains which number to use in reports.

What AI Citation Rate tells you and what it does not

AI Citation Rate tells you whether your site appeared among the sources in answers to the questions you registered. You can choose ChatGPT, Gemini, Perplexity and Claude.

Scope diagram. Inside the box: a registered question, four AIs' answers, and whether your site is among the sources. Outside the box: an individual customer's screen, Google's AI Overviews, visits to your site, and AI training

What it tells you What it does not tell you
The share of answers citing you, per registered question What answer an individual customer saw
Differences between AIs and your rank within answers Appearances in Google’s AI Overviews interface
Which other sites were cited Whether people who saw the answer visited your site
Full answers and cited URLs Whether your information was used in AI training

Gemini results are different answers from Google Search’s AI Overviews. Check AI Overviews appearances in Search Console [3]. Site visits are covered by AI Referrals, explained later.

Questions and credits to prepare before using AI Citation Rate

You need a few questions to test and credits for the analysis. Credits needed are questions × AIs × attempts × 2.

Credit formula diagram. Four question cards, three AIs, one attempt and a factor of 2 multiply to 24 credits in an illustrative example, with a note that attempts default to 1 and can go up to 5

Attempts (試行回数) means how many times the same AI is asked the same question. The default is 1, and you can raise it to 5.

Questions AIs Attempts Credits needed
4 3 1 24
5 3 1 30
5 4 1 40
5 3 2 60

If your initial free credits are 25, 5 questions × 3 AIs is not enough. Reduce to 4 questions. AI Suggestions (AI 提案), which drafts candidate questions, also uses credits, so deduct those too. Claude is available only on paid plans and is marked Paid (有料). Check plan details on the pricing page [5].

Choose questions using words from real inquiries and sales conversations. Questions containing only your company name do not test whether you are recommended to people who do not know you. Illustrative example: for accounting software, “Which accounting software is easy for sole proprietors?”

Follow these steps:

  1. Open AI Citation Rate and enter the URL of the site to test.
  2. Add target questions with AI Suggestions (AI 提案) or Manual Input (手動入力).
  3. Choose the LLMs for analysis (分析に使う LLM), Attempts (試行回数) and Language and region (分析する言語・地域).
  4. Compare “This analysis will use N credits” with your Balance (残高).
  5. If you have enough, press Start Analysis (分析開始). Progress appears, and the results table shows when complete.

Reading the AI Citation Rate results table and ranking

In the results table (想定質問 × LLM 引用率), rows are questions and columns are AIs. Each cell shows the citation rate and your rank within the answer.

Illustrative simplified results table with questions as rows and AIs as columns. A horizontal arrow marks differences between questions, a vertical arrow marks differences between AIs, and a zoomed cell shows citation rate and rank within the answer

Read it in three ways:

  • Rows: show differences between questions. For low rows, check whether your public pages answer that question.
  • Columns: show differences between AIs. ChatGPT may cite you while Gemini does not.
  • Rank: shows where you appeared in the answer. A higher rank does not mean more visits.

The Cited Domain Ranking (引用ドメインランキング) lists cited sites by count and shows your position as “Own rank N.” Open the URL list to see the actual pages cited. Open the top pages and compare their opening answers, comparisons or steps, and evidence with your page. You can export the list with CSV Download.

AI answers vary even for the same question. Do not judge from a single rise or fall; compare runs using the same questions, AIs and attempts.

What to check in the answer log when the citation rate is 0

If the citation rate is 0, use the AI Answer Log (AI 回答ログ) to separate a failed run from a completed run that did not cite you. Open it from the Log link in the results table.

What the screen shows Meaning Next step
Status Failed (失敗), or an Error message No answer was retrieved Rerun the analysis later
Status Running (実行中) or Queued (待機中) Not finished yet Wait for completion
Status Completed (完了) and each attempt Not cited (引用なし) Your site was not cited for that question Open the full answer and its sources

If you were not cited, check which sites were cited instead. Record whether your site lacks the answer or has it but was not selected.

How Search Console and Bing AI reports differ from AI Citation Rate

The three numbers count different things. Do not add them together or treat them as the same number; record them separately.

Report What it counts Scope What you learn
Search Console generative AI report Times links to your site were shown AI Overviews and AI Mode in real Google searches Which pages were shown and how often [2][3]
Bing Citation Share Your share of all citations Answers in Copilot, Bing and select partners for a specific grounding query Changes in your weight among citations [4]
AgentSignal AI Citation Rate Share of answers citing you Answers to questions you registered Results by question and AI, plus other cited sites

Search Console’s report rolled out to all sites on August 31, 2026 [2]. The metric described in its help article is impressions [3]. Bing’s Citation Share entered preview on June 16, 2026. Microsoft describes it as an observational metric, not a ranking or competitive scoreboard; it does not expose competitor domains or represent traffic share [4]. A grounding query is a search the AI runs to find evidence for its answer. AI Citation Rate divides by answers; Citation Share divides by citations.

In short, Search Console and Bing show what happened in real searches, while AI Citation Rate shows what happened for questions you chose. AI Citation Rate also shows which competitor sites were cited. In reports, always state which number you are using.

Comparison of three numbers. Search Console's generative AI report counts link impressions in real Google searches, Bing Citation Share counts your share of all citations, and AgentSignal AI Citation Rate counts the share of answers to your registered questions that cite you; they should not be added together

Using AI Referrals and URL Diagnostics to choose the next page to fix

After citation rates, check visits in AI Referrals, then choose one page to improve with URL Diagnostics.

2x2 matrix of citation rate and AI referrals. Both high: check whether landing pages lead to inquiries. High citation, few referrals: check rank in answers and cited pages. Low citation, many referrals: draft question ideas from landing pages. Both low: inspect public pages with URL Diagnostics. These are next checks, not causes

AI Referrals shows the share of visits coming from AI services and the daily trend. It requires the tracking tag on your site. See How to Install the Tracking Tag.

Citation rate AI referrals Possible situation What to check next
High Many You appear in answers and people visit Whether landing pages lead to inquiries
High Few You appear in answers but get few visits Your rank in answers and which pages are cited
Low Many Visitors may come from questions you have not registered Draft question ideas from landing pages
Low Few AI has not selected you yet Inspect public pages with URL Diagnostics

This table suggests what to investigate next; it does not establish causes.

URL Diagnostics inspects the settings and content of public pages. The diagnosis itself is free; only generating fix code with AI uses credits. Start with one page related to a low-citation question, and record the change date and what you changed. Then rerun the same questions to compare.

For related terms, see What Is AIO?. For how AI expands searches, see Query Fan-Out.

FAQ

Q. If I have Search Console’s AI report, do I still need citation tracking?
They serve different purposes. Search Console shows impressions in Google’s AI Overviews and AI Mode. AI Citation Rate asks ChatGPT, Gemini and other AIs your registered questions and shows whether you were cited and which other sites were. Record both separately.
Q. Why do Bing’s Citation Share and AI Citation Rate show different numbers?
Bing’s Citation Share is your share of all citations shown for a specific grounding query. AI Citation Rate is the share of answers to your registered questions that cite you. The denominators and the questions differ, so the numbers will not match.
Q. Why can’t I select Claude?
Claude is available only on paid plans. On the free plan you can select ChatGPT, Gemini and Perplexity. Check the Paid (有料) label on screen.

Sources

  1. [1] AI features and your website (Google Search Central) — accessed 2026-09-23
  2. [2] Introducing Search Generative AI performance reports in Search Console (Google Search Central Blog) — accessed 2026-09-23
  3. [3] Generative AI performance report (Search) (Search Console Help) — accessed 2026-09-23
  4. [4] New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare (Microsoft Bing Blogs) — accessed 2026-09-23
  5. [5] AgentSignal Pricing (AgentSignal) — accessed 2026-09-23

About the author

Shogo Mizushima

CEO of kairos Inc. / AgentSignal Developer

Develops AgentSignal, a tool for measuring AI crawler visits and AI-referred traffic, and diagnosing AIO readiness. Writes about measurement and practical improvements for AI search using observed data.

Related articles