Definition
What is AI citation tracking?
AI citation tracking is the practice of collecting the sources AI engines link to in their answers, prompt by prompt, and following them over time. Each citation is a page the engine presents as support for what it wrote. Grouped by domain and by kind of source, the citations show which sites an engine relies on when it talks about your market.
Citations send little traffic on their own. In Pew Research Center's analysis of March 2025 browsing data, users clicked a link inside a Google AI summary on 1% of the visits that showed one[1]. Microsoft draws the same line in the help pages of Bing Webmaster Tools: a citation in its AI Performance report means the content was shown in an AI answer, and it does not represent clicks or traffic[2].
The value of a citation sits upstream. The pages an engine cites are pages it read, and what those pages say about your category feeds the answer. Google says any page that is indexed and eligible to show a snippet can appear as a supporting link in AI Overviews or AI Mode, with no special markup[3]. The pages that win citations are ordinary web pages: reviews, comparisons, documentation, forum threads.
For the prompt “best project management tool for creative agencies”, a citation report might list a review site's roundup, a comparison on a design blog and Cadence's features page. None of the three belongs to Loopwork, and only the roundup mentions it. Those pages are where Loopwork's next mention has to be earned.
Comparison
A citation, a mention and true reach measure 3 different things
A citation is a source an AI engine links to, a mention is a brand named in the answer text, and true reach connects the two: the share of cited pages that also name the brand. A brand can be mentioned without being cited, and cited without being mentioned, so each one needs its own count.
| Concept | What it is | Where it shows | What it tells you |
|---|---|---|---|
| Citation | A page the engine links to as a source | Next to or under the answer | Which sites the engine relies on |
| Mention | The brand named in the answer text | In the answer itself | Whether the engine names the brand |
| Citation rate | Share of answers citing at least one source attributed to the brand | Computed across answers | How often the brand's own pages back an answer |
| True reach | Cited pages that mention the brand ÷ all cited pages | Computed across cited pages | Whether the sources behind the answers talk about the brand |
Each case calls for different work. Loopwork can be named in an answer whose citations all point to other sites: the mention counts, and the citation rate does not move. Loopwork's pricing page can also be cited for a fact about per-seat pricing while the answer recommends Cadence: the citation counts, and visibility does not. The first case is a source problem, the second a positioning problem, and a dashboard that blends them hides both.
Citations also deserve some caution as evidence. A 2023 study of four generative search engines found that, on average, only 51.5% of generated sentences were fully supported by their citations, and 74.5% of citations supported the sentence they were attached to[4]. OpenAI's help page for ChatGPT search warns that citations can be incomplete, outdated or wrong[5]. Open the page behind a citation before you act on it, above all when it is a page that keeps you out of the answer.
Engines
Google, ChatGPT and Perplexity show their sources in 3 different ways
AI engines show their sources in different formats. Perplexity gives every answer numbered citations, ChatGPT search adds links to the answer and a Sources panel, and Google shows supporting links inside AI Overviews and AI Mode. A citation tracker has to read each format and bring it down to the same record: URL, domain and position.
Perplexity's help center says every answer comes with numbered citations that link to the original sources[6]. OpenAI says ChatGPT search answers with links to sources such as news articles and blog posts, plus a Sources button that opens the full list[7]. ChatGPT also decides from the question whether to search the web at all, so some answers carry no citation. In a 2026 study of four engines, 57.8% of ChatGPT's runs had none[8].
Google reports clicks and impressions from AI Overviews and AI Mode inside the Web search type of Search Console's Performance report, mixed with the rest of search traffic[3], and counts an AI Overview as a single position shared by all the links inside it[9]. Bing Webmaster Tools has a dedicated report for citations in Copilot and Bing's AI summaries, built on a sample of citation activity and grouped by the phrases the AI searched for[2].
Neither report shows the answers themselves or the competitors cited next to you, and each covers one company's engines. Tracking at prompt level, on your own prompt set, is the way to see both.
True reach
True reach turns the cited pages into 1 number: the share that names your brand
True reach is the share of the pages an AI engine cited that also mention your brand. A true reach of 10% means nine in ten of the sources behind those answers leave the brand out. Raising it means getting the brand onto the pages engines already cite, through reviews, comparisons, partner content and corrections.
Formula
true reach = cited pages that mention the brand ÷ all cited pages × 100 True reach connects the source layer to the answer layer. When it is low, the engine is reading about your category on pages that leave you out, and its answers have less material about you to draw on.
Worked example: 6 of 24 cited pages name Loopwork, a true reach of 25%
Sample data on fictional brands: 8 prompts on 3 engines, US / en.
Sort the cited pages that leave the brand out by how often engines cite them. The top of that list is an outreach plan: a roundup that skipped you, a comparison that lists three competitors, a forum thread where nobody brought you up. Pages that already cover two or three of your competitors are a natural first approach, since the author already writes about the category.
Read true reach per engine as well, since engines cite different pages and a brand can be well covered by the sources one engine prefers and absent from another's. Over time, a rising true reach on the same prompts shows outreach landing on the pages engines actually use. Track it on the same prompts as visibility, so a change in one can be read against the other.
Method
AI citations can be tracked in 4 steps, starting from the prompts buyers ask
AI citations are tracked by running a fixed prompt set on each engine, saving every cited URL, grouping the URLs by domain and kind of source, and checking which cited pages mention the brand. The first three steps work by hand on a few prompts. The fourth is where tracking at scale needs software, since each cited page has to be read.
-
Step 1
Start from your prompt set
Use the prompts you already track for visibility, so citations and mentions come from the same answers. A citation report on other prompts cannot explain a change in visibility.
-
Step 2
Save every cited URL per engine
Run each prompt on each engine and record every source link with its position in the answer. Keep the full URL, since one domain can host pages that help you and pages that do not.
-
Step 3
Group the sources by domain and kind
Roll the URLs up to domains and label each one: review site, publisher, forum, marketplace or brand site. The groups show where engines look when they answer about your category.
-
Step 4
Check which cited pages mention you
Flag each cited page that names your brand to get true reach, then list the pages engines cite most that leave you out. The Citations API returns every cited page with its domain, its category and whether it mentions the brand. See the Citations API.