Citation analysis is CoreCited’s AI citation tracker. It records the pages AI answers cite for your buyers’ questions, groups them by domain and by kind of source, and picks out the ones that cite your competitors and have never cited you. That list is where AI visibility is won, and most of it is on sites you do not own.
Third-party figures, not CoreCited data: Ahrefs’ Brand Radar sample of US queries[1], and Ahrefs’ study of 863,000 keywords as reported by Search Engine Journal[2].
Analyse the citations in any AI answer
Before the product, the logic, as a free citation analysis tool. Copy the sources from a ChatGPT search, Perplexity or AI Overview answer and paste them below. The analyser groups them by domain, labels each with a source type using the same rules CoreCited applies to every stored citation, and tells you what kind of work each one calls for.
How ChatGPT citations work
ChatGPT citations appear only when ChatGPT searches the web. OpenAI’s help centre says responses that use search contain inline citations, and a Sources button opens the full list of references[3]. When it answers from what the model already knows, there is nothing to cite, which is why the same brand can be recommended in one answer with a link and in another without.
There is also a gap between what ChatGPT reads and what it shows. For its API, OpenAI says inline citations show only the most relevant references, while a separate list holds every URL the model consulted[4]. So a page can shape an answer without being credited in it. OpenAI does not publish how the cited ones are chosen; the guide to how AI chooses citations separates what is known from what is guessed.
| reddit.com | 16.8% |
|---|---|
| wikipedia.org | 7.0% |
| consumerreports.org | 3.7% |
| forbes.com | 3.1% |
| engineerfix.com | 2.8% |
Read that chart as a list of places rather than a ranking to chase. A forum thread and an encyclopedia entry lead it, and neither is a page a brand writes. That is the pattern citation analysis is built around: the page that decides whether you are recommended is usually someone else’s.
Perplexity, AI Overview and ChatGPT citations compared
LLM citations look similar on screen and behave differently underneath. What each engine shows, and what CoreCited can collect from it:
| How it shows sources | Cites every answer | Collected by CoreCited | |
|---|---|---|---|
| Perplexity | Numbered inline citations | ~ | ✓Yes |
| Google AI Overviews | Source cards beside the summary | ~ | ✓Yes |
| Google AI Mode | Inline links and source cards | ~ | ✓Yes |
| ChatGPT | Inline citations and a Sources panel, when it searches | ✕No | ✕No |
| Claude | Citations when web search is used | ✕No | ✕No |
| Gemini | Sources when grounded in Search | ✕No | ✕No |
Perplexity citations are the most complete record: Perplexity says every answer that uses web search carries numbered citations linking to its sources[5]. Google’s AI Overview citations are the most tied to ranking, though less than they were: in Ahrefs’ early-2026 data only 38% of cited pages were in the top 10 for the same query, against 76% in its July 2025 study, a drop Ahrefs partly puts down to better detection[2].
| Top 10 | 38% |
|---|---|
| Positions 11 to 100 | 31.2% |
| Beyond 100 | 31% |
The AI Overviews guide and the AI Mode guide cover how each Google surface picks its sources.
What CoreCited’s citation analysis shows
The source gap: a list of pages, not a score
A visibility score tells you where you stand. It does not tell you what to do on Monday. The source gap does: it names a domain, the competitors it cites, and the questions it was cited for. “A review site cites two of your competitors for ‘best CRM for startups’ and has never named you” is a task with an address on it.
When a gap is on your own content, a content brief turns the missed prompt into a page. When it is a site you do not own, the competitor tracking view shows how much share of voice that source is worth.
What citation analysis is not
What is on which plan
| Capability | From |
|---|---|
| Full AI response viewer, with the cited sources | Starter |
| Engine-specific source strategy in briefs | Starter |
| Competitor source gap | Growth |
Full prices are on the pricing page.
What it cannot tell you
- Why a source was chosen. No engine publishes its selection rules. The analysis shows which sources are cited, how often and alongside whom; the reasons are inference.
- Citations in ChatGPT, Claude or Gemini answers. Tracked through their APIs they return brand names without links, so there are none to record.
- Pages an engine read but did not cite. Only the sources shown with the answer are stored.
Questions people ask
How does ChatGPT choose its citations?
ChatGPT only cites when it searches the web. OpenAI's help centre says answers that use search contain inline citations, with a Sources button for the full list. OpenAI does not publish how sources are chosen. What can be measured is the result: in Ahrefs' September 2026 data, Reddit held 16.8% of ChatGPT's citations and Wikipedia 7.0%.
Why does ChatGPT sometimes give no citations?
Because it answered from what the model already knows rather than searching. Citations only appear on answers that used web search. That is also why a brand can be named without a link: the model remembers the brand, but no page was retrieved for that answer.
What are LLM citations?
The links an AI assistant attaches to its answer to show where a statement came from. Perplexity numbers them inline, ChatGPT search shows them inline and in a Sources panel, and Google's AI Overviews and AI Mode show them as source cards. A model answering from memory has nothing to cite.
Are Perplexity citations the same as the pages it read?
Not necessarily. Perplexity says each answer includes numbered citations to its sources. OpenAI draws the distinction explicitly for its API: inline citations show only the most relevant references, while a separate sources field lists every URL consulted. Being read and being cited are two different outcomes.
Do AI Overview citations come from the top 10 results?
Less than they used to. In Ahrefs' early-2026 study of 863,000 keywords, 38% of cited pages ranked in the top 10 for the same query, with the rest split between positions 11 to 100 and beyond 100. Ahrefs' July 2025 figure was 76%, partly because its detection has since improved.
What does an AI citation tracker do?
It runs the questions your buyers ask through AI engines on a schedule, stores the answers and the sources they cite, and counts which domains come up. The useful part is the comparison: which sources cite your competitors and never you, because those are the pages to work on.
Which engines does CoreCited collect citations from?
Perplexity, Google AI Overviews and Google AI Mode, the ones that return their sources with the answer. CoreCited also tracks ChatGPT, Claude and Gemini, but through their APIs those answer with brand names and no links, so they count towards mentions, not citations. Every citation finding says which engine it came from.
Can I get my site cited by AI?
You can raise the odds, not guarantee it. For engines that search, a page has to be crawlable, rank reasonably and answer the question directly. For the pages you do not own, the forums, reviews and lists that cite your competitors, the work is getting a genuine mention there. No engine sells placement in its citations.