What are AI citations and AI visits?
These are two different things, and you want both.
An AI citation is when an AI assistant uses your page in its answer and links to it or names it. Think of it as the new search ranking. Instead of a blue link, your page becomes a source the AI trusts.
An AI visit is a real person who lands on your site after using an AI tool. They read the answer, clicked through, and arrived.
Citations are about being seen by the AI. Visits are about people showing up. This tool tracks both, on your own website, so you are not guessing.
Why AI traffic is so hard to see
Most analytics tools were built for the old web.
AI traffic arrives in ways those tools miss. Some of it is bots that never run JavaScript. Some of it is people whose source is hidden before they reach you.
| How AI traffic arrives | What it is | Easy to miss because |
|---|---|---|
| AI crawlers | Bots like GPTBot fetch your pages to learn what they say | GA4 runs on JavaScript and never sees them |
| AI referrals | A person clicks your link inside an AI answer | Some arrive as referrers, but many do not |
| Referrer-stripped visits | Claude, Grok, and in-app browsers hide the source | They look like plain Direct traffic |
This is why a page can be cited all over ChatGPT and still look quiet in Google Analytics. The traffic is real. It is just labeled wrong, or invisible.
The AI funnel: crawled, cited, clicked
AI visibility happens in three steps.
First an AI bot crawls your page. Then the AI decides to cite it in an answer. Then a person clicks through and visits.
Each step has a drop-off. If you are crawled but rarely cited, your content may be hard for the AI to trust or quote. If you are cited but rarely clicked, your titles and descriptions may need work. The number that ties this together is your crawl-to-citation rate, the share of crawled pages that turn into a citation.
Which AI assistants we track
We follow the major AI tools and the crawlers they use. Some pass a referrer when they send you a visitor, and some do not, which is why on-site tracking matters.
| Assistant | Crawler | Sends a referrer? |
|---|---|---|
| ChatGPT | GPTBot, OAI-SearchBot | Sometimes |
| Perplexity | PerplexityBot | Often |
| Gemini | Google-Extended | Sometimes |
| Claude | ClaudeBot | Rarely |
| Copilot | Bingbot | Sometimes |
| Grok | Varies | Rarely |
You can read each crawler's rules in the OpenAI, Perplexity, and Anthropic docs. If your robots.txt blocks one of them, that assistant cannot crawl or cite you at all.
How to read referrer-stripped AI visits
A referrer is a small note a browser sends that says where a click came from. For years that note told analytics tools that a visit came from Google, or from a link on another site.
Many AI tools do not send that note. Claude, Grok, and the in-app browsers inside mobile apps often strip it for privacy. The visit still happens, but your old analytics files it under Direct, the same bucket as someone typing your address by hand.
That single gap hides a large and growing slice of AI traffic. To recover it we look at more than the referrer: the landing page, the timing, the path through your site, and known AI patterns. The result is an honest estimate of the AI visits your old tools quietly drop. The W3C explains the raw mechanic in its Referrer Policy spec if you want the technical detail.
Supply side vs demand side
There are two ways to measure AI search, and they answer different questions.
Demand-side tools ask the AI models whether you are named in their answers. That is useful for tracking your share of voice against competitors.
Supply-side tracking, which is what we do, measures what actually reaches your site: the crawls, the citations, and the human visits. It is the ground truth for your own pages.
Most teams want both. We focus on the half that lands on your website, where the numbers cannot be argued with.
How to grow your AI citations
Once you can see the funnel, the fixes get obvious. A few habits move the needle the most.
Start by making sure every AI crawler is allowed in your robots.txt. A single blocked bot is a citation you will never earn. Google's own robots.txt guide shows the exact format.
Then keep your main answer in plain HTML, near the top of the page, so a bot that skips JavaScript still reads it. Add a clear author and dates with schema so the AI can trust the source. Write direct, quotable sentences that an assistant can lift into an answer without rewriting them.
After that, let the data lead. Pages with a low crawl-to-citation rate usually need a sharper answer or stronger proof. Pages that get cited but few clicks usually need a better title. Measuring AI traffic turns these guesses into a short, ranked to-do list.
Frequently asked questions
- What is an AI citation?
- An AI citation is when an AI assistant like ChatGPT, Perplexity, Gemini, or Claude uses your page in its answer and links to it, or names it as a source. It is the AI-era version of a search ranking. Instead of a blue link on a results page, your page becomes one of the sources the AI quotes.
- What is an AI visit?
- An AI visit is a real person who lands on your site after using an AI tool. They read an answer, clicked your link or typed in your name, and arrived. AI visits are the human traffic that AI search sends you, separate from the AI bots that crawl your pages.
- How do you track which AI cited my page?
- We measure it on your own website, not by guessing from prompts. A small snippet records humans who arrive from AI tools, and a server-side feeder records AI crawlers that fetch your pages. Together they show which assistants crawled each URL, which ones cited it, and which sent real visitors.
- Why doesn't Google Analytics show my AI traffic?
- Two reasons. First, GA4 runs on JavaScript, so it cannot see AI crawlers like GPTBot at all. Second, a lot of AI visits arrive with no referrer. Tools like Claude, Grok, and in-app browsers strip the source, so GA4 files them as Direct traffic. The AI slice ends up hidden or mislabeled.
- Can you see AI crawlers like GPTBot and ClaudeBot?
- Yes. We capture them server-side and verify them by IP, so a bot that pretends to be GPTBot gets flagged. You see which AI crawlers visited, how often, and which of your pages they fetched. That is the first half of the funnel, before a citation or a click can happen.
- How is this different from Profound or Promptwatch?
- Those tools watch the AI answers. They ask the AI models whether you are named in a reply, and track your share of voice. We watch your website. We measure what actually reaches you: crawls, citations, and human visits. They cover the demand side, we cover the supply side, and many teams use both.
- Do I need to install anything?
- You add one small snippet to your site, and for full crawler tracking, a tiny server-side feeder. After that your dashboard fills in on its own. There is nothing to configure for each AI tool, since the platform classifies the traffic for you.
- What is the crawl-to-citation rate?
- It is the share of your AI-crawled pages that actually get cited in an answer. If 100 of your pages were fetched by AI bots and 30 of them showed up as a source, your crawl-to-citation rate is 30 percent. A low rate usually means your content is hard for the AI to trust or quote, so it is a good early warning that something needs work.