An AI citation is what happens when ChatGPT, Claude, Perplexity or Google AI Mode answers a question and points to a source for what it just said. A link, a named publication, a footnote. That is all it is.
I have spent a lot of time looking at them. Across Subscribe PR client campaigns we have counted 27M+ AI citations. My honest conclusion after staring at that number is that a citation is not a ranking and it is not traffic. It is a measure of trust, and specifically trust that other people extended to you before the tool ever got involved.
That changes what you should spend your effort on. Here is the argument.
A citation is borrowed trust, and the tool is doing the borrowing
When you ask a person who to hire, they do not run a calculation. They think of who they have heard good things about, from sources they believe. AI tools are built to imitate that behavior at scale. They read what has been written, decide which sources sound reliable, and repeat what those sources say.
So when a tool cites a publication in an answer about your field, it is telling you where the confidence came from. The tool is not vouching for you directly. It is pointing at whoever vouched first.
I wrote about this in AI tools don't rank you, they vouch for you, and the citation count is the closest thing I have to evidence for it. If citations were driven by who has the slickest website or the most posts, the numbers would follow effort. In my experience they follow independent mentions.
There is outside data pointing the same way. Muck Rack's Generative Pulse research, updated in May 2026, analyzed more than 25 million links from ChatGPT, Claude and Gemini answers across 17 industries. It found that earned media, which they define broadly to include journalism and other third-party sources, accounted for 84% of citations, while paid and advertorial content made up 0.3%. Across its three editions the earned share has stayed in a range of roughly 82% to 89%.
Two cautions, because I would rather you hear them from me. Muck Rack sells to PR teams, so it has an interest in this finding. And their definition of "earned" is broad, so the 84% does not mean 84% of citations come from a news article about a company like yours. It also does not prove that one placement will cause one citation. What I take from it is narrower and, I think, sturdy: the tools lean on what other people have published, and they barely lean on what a company says about itself in paid space.
What I got wrong about volume
When I started watching citation numbers, I assumed bigger was the goal. More mentions, more citations, more of everything.
That is partly true and mostly misleading. One client went from zero to 253.5K AI citations in 60 days. It would be easy to read that as a story about speed or volume. I read it differently. They had been invisible, which means the tools had nothing credible to repeat. Once there were enough consistent, independent sources describing who they were and what they did, the tools had something to say, and they said it a lot.
The citations were a symptom. The cause was that a stranger asking about their category now had material to work with. If I had chased the symptom, by trying to inflate the count without building the underlying material, I would have gotten nowhere. I have seen people try this with thin or repetitive pages, and the pattern is always the same. Lots of activity, little trust.
This is also why I stopped caring about vanity placements. A mention that is flattering but vague does not give a tool anything to cite. A plain, specific sentence in a credible place does.
What I would check this week
You do not need a tracking platform to start. You need about thirty minutes and a willingness to be unimpressed.
Ask AI the questions your buyers ask. Not your name. The problem. If you run a clinic, ask who the best option is for the treatment you perform in your city. If you are a founder, ask who the leading voices are in your niche. Try ChatGPT, Claude, Perplexity and Google AI Mode, because they do not always agree.
Write down which sources get cited. Ignore whether you appear for a moment. Look at what the tools point to. Those publications and pages are the ones that currently carry trust in your category. That list is a map of where your own proof needs to live.
Check what those sources say about you, word for word. If you do appear, read the exact sentence. Does it say what you do, who you serve and where? A mention that skips those facts cannot do much work for you. If you do not appear, you now know what is missing and where it is missing.
Fix one gap. Not ten. Pick the clearest one, such as a name that is spelled three ways across the web, a description that is out of date, or a category where a competitor is cited and you are absent. Then find one credible place that can accurately say the right thing. That is slow, unglamorous work, and it is also the work.
The part I cannot promise
Nobody outside the AI companies can tell you exactly how these tools choose their sources, and they change. A source that is cited heavily this quarter may be quieter next quarter. I would not build a strategy on any single tool's quirks.
What has held up for me is simpler. Be described accurately and consistently by credible people other than you, in places a stranger's question will lead to. That was true when the only search engine was Google, and it appears to be even more true now that the answer arrives as a paragraph with a footnote.
So here is my question for you. If you asked ChatGPT who to trust in your field today, and it cited three sources, would any of them have said your name?
