Sources AI cites
AI citation tracking that shows which sources AI cites, and where you’re missing
Rankry lists the domains AI engines cite when they answer your prompts, labels each one “Yours”, “Gap” or “Mixed”, and ranks the gaps worth closing. Open any domain to see its top pages and the answers that cited it.
7-day Base trial · No card needed
Sources · UGC Planner
Current report14gap sources
Domains whose answers name competitors, not you
- reddit.com · ForumGap
- g2.com · ReviewMixed
- collabstr.com · CompetitorGap
- ugcplanner.com · YouYours
Sample data
In short
AI citation tracking is the practice of recording which websites AI engines cite when they answer the questions your buyers ask. Rankry’s Sources page shows the domains cited across your tracked prompts in the latest report, sorts them into types like Review, Forum and Competitor, and flags the “Gap” sources that cite rivals but never you.
The short answer
Which sources does ChatGPT cite?
It depends on the question. When ChatGPT searches the web before answering, it reads a handful of pages and may link some of them in its reply. Ask about creator tools and you tend to get software review sites, Reddit threads and vendor pages. Ask about roofers in Philadelphia and you get directories, local news and contractor sites. No single “most cited” list fits every market, which is why the useful answer comes from your own prompts.
Rankry asks ChatGPT your prompts through OpenAI’s API with web search on, and stores two kinds of evidence: the search results the model consulted and the links it cited in the answer. Claude, Gemini, Perplexity, Grok and Meta Muse are asked the same way through their own APIs. Google AI Overviews and Google AI Mode, included on Pro or available as Base add-ons, are captured from Google’s live results with the references they show, and Copilot uses Azure GPT with Bing web search through Microsoft’s official API.
Each engine reports sources a little differently. Grok, for example, returns only the links it cites inline. Chat engines also don’t always search: when a model answers from memory, there is nothing to record for that answer. That is normal, and it is why a prompt can show sources on one engine and none on another.
In the app
What does the Sources page show?
The page answers three questions at once: who AI reads, who those sources recommend, and where you are missing.
Coverage strip
“Total citations” across your answers, “Unique sources” with the mix of source types, “Mention you” with the share that comes from your own properties, and “Gap sources” that cite competitors, not you.
Domains table
Every cited domain with its type, “Usage” (citations per answer), how its answers split between you and competitors, which engines cite it, its status and a priority.
Status: Yours, Gap, Mixed
“Yours” is your own domain, or a source whose answers name only you. “Gap” means the answers citing it name a competitor but not you. “Mixed” means both appear.
Priority
High, Medium or Low. It weighs how often a domain is cited, how many of those answers name competitors, and how realistic it is to get onto it: review sites weigh most, then forums, publishers and reference sites.
Domain detail
Click a domain to see how often answers cite it, how many of those answers are about you or about competitors, which engines cite it, its “Top URLs”, and sample answers that link back to their prompts.
Filters
Narrow by topic (your prompt categories), source type and engine, search for a domain, or switch on “Gap analysis only” to see just the sources that leave you out.
Source types
How are AI sources classified?
Rankry labels each domain with a fixed set of rules rather than a model’s guess, so the same domain always gets the same type.
| Type | What counts | What to do about it |
|---|---|---|
| You | Your own domain. | Keep the cited pages accurate and current. If a page is cited but your brand isn’t named in the answer, make the link between the page and your product clearer. |
| Competitor | The website of any brand named in your answers, including brands you don’t track yet. | Read which of their pages get cited. That usually points to the comparison, pricing or how-to content you are missing. |
| Review | Software review sites: G2, Capterra, TrustRadius, Software Advice, Gartner and GetApp. | Complete your profile and ask happy customers for reviews. Review sites carry the most weight in priority. |
| Forum | Community sites: Reddit, YouTube, Hacker News, Quora, LinkedIn and Stack Overflow. | Take part in the threads that get cited, where you can genuinely help, and say who you are. |
| Reference | Wikipedia. | Check that the facts about your category are right. Wikipedia discourages editing about your own company, so work on the independent coverage it relies on. |
| Editorial | Everything else: publishers, blogs, directories, trade media and niche sites. | Pitch the editors who cite rivals, offer data or expert quotes, and ask to be included in roundups that already get cited. |
The rules use known domain lists, so a niche review site outside the list shows as Editorial. Status and priority still apply.
Gap analysis
What is citation gap analysis?
Citation gap analysis is the process of finding the sources AI cites for your market that mention your competitors but not you. Those sources shape the answer. If a review roundup lists three rivals and leaves you out, every engine that reads it learns the same short list.
In Rankry, a domain is a “Gap” when the answers that cite it name a competitor and never name you. Competitor here means any other brand with a website that the answer names, so the gap view also catches brands you haven’t added to tracking yet. Switch on “Gap analysis only” and the table shows only those domains, ready to work down by priority.
A mention gap and a citation gap are related but different. A mention gap means the answer doesn’t name you. A citation gap means a source the answer relied on doesn’t include you. Closing citation gaps is often how you close mention gaps, because it changes what the engine reads.
Routine
How do I find sources that cite competitors but not me?
A short routine you can repeat after each report.
- 01
Switch on “Gap analysis only”
The table drops every domain whose answers already name you. What remains is the list of sources talking about your market without you.
- 02
Sort by priority
High-priority gaps are cited often, mostly in answers that name rivals, and are realistic to act on, like review sites and forums.
- 03
Filter by engine and topic
Pick one engine or one prompt category if a single model or product line matters most. Different engines can lean on different domains.
- 04
Open the domain
Read its “Top URLs” and the sample answers. You see the exact page that matters and which prompts it shows up for.
- 05
Choose the move that fits
Update a review profile, answer a forum thread, pitch an editor, or write the missing page on your own site in Content Studio.
- 06
Add it to Tasks
Create a task for each gap you take on, with a priority and notes, so the work doesn’t live in a spreadsheet.
- 07
Check after the next run
Sources reflects the current report. After the next run, look again: a domain moves to “Mixed” once the answers that cite it name you too.
Your own site
Is my site cited even when my brand isn’t named?
It happens. An engine can read your guide to pricing UGC deals, cite it, and still recommend someone else. On the Sources page your domain appears with the type “You”. Open it and the detail panel shows how many of the answers citing it are about you and how many are about competitors.
If many answers cite your pages but name a rival, the page is useful but doesn’t make your brand the obvious answer. Make the product connection explicit near the top of the page: what you offer, who it’s for, and how it compares.
Get cited more
How do I get more of my pages cited by AI?
Citations follow pages that answer the question clearly and that crawlers can read. Start here.
- Lead each page with a one- or two-sentence answer to the question it targets.
- Use question-style headings that match how buyers phrase prompts.
- Add a real table when you compare options, prices or features.
- Show a visible update date and an author.
- Back claims with numbers, and name where they come from.
- Let AI search crawlers in through robots.txt and keep key content in the HTML.
- Earn mentions on the review sites, forums and publishers that already get cited.
- Run the AI Readiness audit to find what blocks citations on your homepage.
- Write missing comparison and FAQ pages in Content Studio.
Explore Rankry
Related features
What is the difference between sources and citations?
A source is a website an AI engine used or linked while answering. A citation is one instance of that: one answer referencing that domain. Rankry groups citations by domain, so the Sources page lists each source once with how often it was cited, while the domain panel shows sample answers and the top pages behind them.
Which AI engines does Rankry capture citations from?
ChatGPT, Claude, Gemini, Perplexity, Grok and Meta Muse, asked through their APIs with web search, plus Google AI Overviews and Google AI Mode on Pro or as Base add-ons, and Copilot, which uses Azure GPT with Bing web search. Coverage differs by engine: some also report the pages they consulted, Grok returns only the links it cites inline, and an answer given without a web search has no sources at all.
Can I see how my citations changed over time?
Not on the Sources page. It reflects your current report, without a date range or trend chart. The simplest way to see progress is to note your gap sources and the “Mention you” figure now, then compare them after the next run.
How does Rankry decide whether a source is a Gap?
A domain is a “Gap” when the answers that cite it name at least one competitor and never name you. Competitor here means any other brand with a website that the answers name, tracked or not. It is “Yours” when it is your domain or only you are named, and “Mixed” when both you and competitors appear.
Is source classification done by AI?
No. Types come from fixed rules: your domain is You, known lists cover Review, Forum and Reference sites, the website of any brand named in your answers is Competitor, and everything else is Editorial. Rules are predictable, though a niche review site outside the list shows as Editorial.
What is a good citation rate?
There isn’t a universal benchmark, and figures from different tools aren’t comparable because they ask different prompts in different ways. Use your own baseline: the share of cited answers that mention you and the number of gap sources. If gaps shrink and “Mention you” grows from one report to the next, you are moving the right way.
Do Reddit and Wikipedia dominate AI citations?
They show up a lot in many markets, but not in every one, and it varies by engine and question. Rankry labels Reddit as Forum and Wikipedia as Reference, so you can filter by type and see how much your own answers lean on them before you spend time there.
Can I see which page on a domain was cited?
Yes. Open a domain to see its “Top URLs”, up to five pages with their counts, plus sample answers that cite the domain, each linking to the prompt it came from.
Can I export the Sources list?
Not yet. There is no export on the Sources page today. You can ask Rankry Brain “Where are competitors cited but I am not?” for a summary, or read your sources from Claude Code, Cursor or another compatible MCP client through Rankry’s MCP server.
How do I turn a citation gap into work?
Decide the move for each high-priority gap, then add it as a task on the Tasks page with a priority and notes. If the fix is a page on your own site, such as a comparison or FAQ, write the first draft in Content Studio and export it as Markdown, CSV or Word (DOCX).
See who AI reads before it recommends
Run your first report and open the Sources page to find the gaps worth closing.
7-day Base trial · No card needed