Citations
The URLs AI models return as sources for their answers, and the sites that cite your competitors but have never mentioned you.
A citation is a URL an AI model returns as a source for the answer it just gave. The Citations page collects every citation your tracked prompts produce, groups them by domain, page, query and video, and marks the sources that cite your competitors but never name you.
Open it when you want to know which sites shape AI answers in your category, which of your own pages models actually use as source material, and where the holes are.
Citations is included with every paid plan. Client portal users get a read-only version without the Outreach tab.
What counts as a citation
Ask Perplexity for the best CRM for a small business and it writes a short answer, then lists the links it used underneath. Those links are the citations. ChatGPT does the same in web-search mode, and Google stacks reference pages under every AI Overview.
Three things a citation is not:
- Not a brand mention. A page can be cited without naming you, and a model can name you from training data without citing anything. Answer-text mentions live in Perception.
- Not a Google rank. Position in a citation list is not a SERP position, and small changes in wording reshuffle it. The presence of the source is the signal, not its slot.
- Not a link to you. Most of your citations are pages on other people's sites. Your own domain is one source among many, and rarely the biggest.
The page that ranks for a query is often not the page that gets cited. A model asked about marathon running shoes pulls a Wirecutter review, a Runner's World guide and a Reddit thread, then writes from the lot. AI visibility follows your citation profile, not your rankings.
Where the data comes from
Three AI surfaces return sources with their answers, and Trakkr captures all three.
| Platform | What it returns |
|---|---|
| ChatGPT Search | The URLs OpenAI's web-search answers cite as sources |
| Google AI Overviews | The reference pages attached to an AI Overview |
| Perplexity | Every URL in Perplexity's citation list |
Each platform runs your active prompts once a day. Every URL that comes back has its tracking parameters stripped, gets normalized to a domain, is deduplicated across runs, then fetched, classified, and checked for mentions of your brand and your tracked competitors.
The other tracked models do not show their working. Claude, Gemini and the rest still shape visibility, but they return no source list, so no citations. Claude is marked as coming soon on the platform filter.
How each citation is classified
Every cited domain is tagged with one of eight source types: Owned, Earned Media, Social, Reviews, Institution, Competition, PR Wire, and Other. Source Type in the metrics reference says which ones carry the most weight and which are usually noise.
Each cited page also carries:
- whether the page mentions your brand (citing you, not mentioned, or not checked), which is what turns it into a gap
- sentiment toward your brand: positive, neutral or negative
- the tracked prompts and the search queries that surfaced it
- the competitors named on the same page
- a domain rating, plus the first and last dates the URL was seen
Every page gets one of three results for your brand:
- Citing you: something from the page names you. That can be the page's own text, a quote an AI engine took from it (Perplexity and Google AI Overviews send one with many citations), its title, or its address. Hover the dot to see the exact words and where they came from.
- Not mentioned: Trakkr read the page and your brand isn't on it. The hover shows when it was read and how many words it had. Only these pages count as gaps.
- Not checked: the page hasn't been read yet, or it couldn't be read. Some sites block automated reading (G2, Gartner, Capterra and Reddit do), some pages need a sign-in, and search results or PDFs have no fixed page text. The hover says which, and when Trakkr tries again. A page that isn't checked never counts as a gap. If an AI answer later quotes it naming you, it moves to citing you.
Trakkr reads each cited page once and reuses that check. A page AI keeps citing, 20 or more times in 30 days, is read again when its check is more than 30 days old, so a list that adds your brand later is picked up. Pages that couldn't be read are retried on their own schedule, from a couple of days after a failed load to two weeks for a site that blocks readers. To reread a page sooner, right-click it or open its ⋯ menu and choose Recheck page. The recheck runs in the next daily refresh. Each page can be rechecked once a day, up to 20 pages per brand per day.
Every source also carries its Domain Rating, and Citation Intent shows the buyer intent behind the queries that produced each citation.
The four tabs
| Tab | The question it answers |
|---|---|
| Sources | Which sites and pages are cited, and what changed |
| Queries | Which questions produce these citations |
| Videos | Which videos AI cites about your category |
| Outreach | Which publishers cite rivals but not you, and how to pitch them |
Sources holds three sub-views of the same data (Domains, Pages, Feed), switched with the pills on the right of the tab row. The old Heatmap grid was retired; ?view=heatmap now lands on Sources, where the Gaps filter covers the same job.
Sources: domains, pages and the feed
Domains
The default view. Every citation is grouped by domain, so you read publications rather than raw URLs. Three filter tabs sit above the list: All shown, Citing (domains with a page that names you), and Gaps (domains AI cites where no page mentions you). Inside Gaps, a second control narrows further to Sites citing competitors but not you.
Sort by Most cited, Highest authority, Best sentiment, Most pages, or Most competitors.
Click a domain to open its profile on the right: citations over time, the specific pages being cited, the prompts behind them, sentiment, the competitors present, and a plain-language takeaway. On a gap domain the profile ends with a Suggested block and a Draft outreach button that hands the domain straight to Outreach.
Pages
Pages flattens the same data to individual URLs. One canonical guide on a domain often drives half its citations while a dozen other pages on the same site contribute almost nothing, and only this view shows you that.
The pills across the top double as filters: pages, your pages, citing you, gaps, and not checked when any exist. Your pages is the useful corrective. The list of your own content that models actually use is short for nearly everyone, and rarely the pages you expected. It also appends pages that hold a top-ten Google rank and have never been cited by AI, taken from Search Console.
When a cited URL already has a page record, an open-record icon appears on hover and a journey rail shows its stages. That link opens the full record in Pages, where citation evidence sits beside work and history.
Feed
A chronological log of what changed between snapshots, over a 7, 14 or 30 day range.
| Event | What it means |
|---|---|
| New | A page started citing you or a competitor where it did not before |
| Lost | A page that previously appeared stopped showing up |
| Sentiment changed | A page's framing of your brand crossed a threshold |
| Competitor appeared | A new competitor showed up on a page in your landscape |
Read the Feed weekly on the 7-day range. It is the quickest way to catch lost coverage, a competitor taking a page you used to own, or the first results of a PR push.
Queries: the questions behind the citations
Every citation traces back to the prompts and search queries that triggered it, and Queries reads from that direction. Each query is classified by intent (Discovery, Comparison, Best For, Alternative, or Recommendation), and each intent is satisfied by a different kind of page. Missing from Alternative queries is usually a content gap, not a brand problem.
Pills filter the list to queries, citing you, gaps, winnable, and new, with a Coverage readout for the share of tracked queries that cite you. Sort by Opportunity score, Gaps first, Most frequent, Most competitors, Most citations, or Newest. Switch between a flat List and Group by intent, and narrow to an Audience when the brand has personas.
Two handoffs live on each row. Create content opens the composer in Content prefilled with the query, its intent and the competitors winning it. Open prompt page appears when the query maps to a prompt you already track, and goes to that prompt's detail page.
Videos
A gallery of the videos AI cites when answering questions about your category, mostly YouTube. Pills filter to videos, mention you, yours, competitor, and inaccurate; sort by Most cited, Most prompts, Recently seen, Best sentiment, or Most viewed. Use it to find the channels shaping answers in your category, and the clips that get you wrong.
Whether a video mentions you is judged from its title, channel name and description, so a brand named only out loud in the video is not counted.
A video is marked inaccurate only when a web search confirms that what it says about you is wrong or out of date.
Scope, filters and export
Three controls in the page header scope every tab at once. Location appears when the brand tracks more than one market. The platform chip limits the page to ChatGPT Search, Google AI Overviews or Perplexity, and only shows when at least two platforms have data. Tags limit the page to citations from tagged prompts. Anything active shows as a chip under the tabs, so counts never move without saying why.
Export gives you the current tab's full set in scope, not the filtered view: All sources, All pages, All queries, or the feed. Detailed citations, one row per query by cited page, is available from every tab.
A weekly pass
- 1.Open Sources, filter to Gaps, and refine to sites citing competitors but not you. Read the top three domains and click into the pages being cited.
- 2.Open Queries, filter to gaps, sort by Opportunity score. Pick one or two and either create content for them or work the domains that answer them.
- 3.Open Feed on the 7-day range. Scan for lost coverage and new competitors, and click through to anything that looks wrong.
Common questions
What is the difference between a citation and a mention?
A citation is a URL the model returned as a source. A mention is your brand name appearing in the answer text. A Wirecutter review can be cited without your brand being named in the answer, and a model can name you from training data without citing anything at all. Citations tracks the URLs; for answer-text mentions, read Perception.
Why is one of my pages cited often and another never?
Models pick sources that read as authoritative and on-topic for the specific query. Long, structured, well-linked content tends to win, while short marketing pages without a clear answer tend to lose. Filter Pages to your pages to see which of your own work is doing the lifting.
How fresh is the data?
Citations refresh once a day, aligned with the daily prompt runs, and the Feed always shows changes against the previous snapshot. After a PR push, expect new citations within a day or two, usually faster on Perplexity because it retrieves live.
Whether a cited page mentions your brand is checked when the page is first read. Pages AI cites often are reread every 30 days, pages that couldn't be read are retried automatically, and Recheck page rereads one in the next daily refresh.
Why does a page say "Not checked" when I know we're listed on it?
Some sites block automated reading, so Trakkr can't open the page to look. It won't guess that you're missing, so the page stays out of your gaps. When Perplexity or Google AI Overviews cite the page with a quote that names you, it shows as citing you with that quote.
Why does Perplexity return more citations than the others?
By design. Perplexity attaches a citation list to almost every answer, Google AI Overviews attach a handful of references, and ChatGPT Search varies a lot. Read the platform filter as a coverage check rather than a scoreboard: the question is whether each surface cites you at all, not which produces the biggest number.
Where did the Heatmap view go?
It was retired. The grid showed which brands were cited on which sources, and the Gaps filter in Sources answers the same question more directly, with the competitor refinement on top. Old ?view=heatmap links now open Sources.
Where do I act on a citation gap?
Two places. The Outreach tab turns publisher gaps into a working queue with a researched pitch for each one. Citation gaps also arrive as Suggested actions with the source page, evidence and work type attached, and those carry a measurement window into Results.