The short version
Citations are part of how an AI assistant shows its work, and much of why its answer reads as trustworthy.
- Only 1% of visits to a page with an AI summary end in a click on a source it cited, according to Pew Research Center.
- Read the citation list anyway. It's your cheat sheet for changing an AI assistant's mind about you.
- What's on it differs completely by category: travel gear is decided by review press, coffee grinders by a dozen bloggers, eyewear by competitors' own storefronts.
Do AI citations send you traffic?
Barely, and nowhere near enough to cover what the summary costs you. The Pew Research Center tracked the real browsing of 900 U.S. adults across 68,879 Google searches. When a summary appeared, people clicked a traditional result in 8% of visits, against 15% when none appeared. The citations don't recover that gap: just 1% of visits to a page with a summary ended in a click on a source it cited.
That data is from March 2025, so read it as a floor. SparkToro's 2026 analysis of Similarweb clickstream data found 68.01% of U.S. Google searches now end with no click at all.
Why read the citation list anyway?
Because it's what the assistant went and read to answer your customers' questions. Answering a live buying question is retrieval as much as recall: the assistant fans one question out into several narrower searches, fetches pages, and writes from what comes back.
It's also your best clue to what the model was trained on for questions like yours. OpenAI, Google, and Anthropic don't publish their training data, but the sites an assistant reaches for when it needs an answer in your category are the same ones that got scraped to build it.
Your citation list is long, but a few sites are cited consistently in most answers. Across the ten reports we ran for this post, the median brand's answers cited 684 different sites, but its ten most-cited sites showed up in 89% of those answers.
What does your citation list tell you?
First, that your own site is rarely the source. Across the ten reports we ran, when an assistant raised a brand on its own rather than being handed the name, that brand's own site was among the sources behind the answer just 19.5% of the time. We broke down where the rest comes from in a separate look at the source list behind one brand's answers.
Then it names the sites deciding your category. Four appeared for all ten brands: reddit.com, youtube.com, forbes.com, and businessinsider.com. Below those four, no two lists look alike.
Matador, which makes travel gear, is written about by specialist review press: packhacker.com with 240 citations, outdoorgearlab.com with 150, bettertrail.com with 115. Earning a place there means getting product into the hands of enthusiast reviewers who publish long comparative teardowns, which is a public-relations problem rather than a publishing one.
Timemore, which makes coffee grinders, has a citation list of only 71 sites, and the top of it is enthusiast blogs and small roasters: coffeechronicler.com with 17 citations, kaldiscoffee.com with 13, pullandpourcoffee.com with 8. A dozen dedicated writers set the consensus in that category, which makes it a far smaller problem once you know their names.
Warby Parker's most-cited third-party sources are competing eyewear retailers: framesdirect.com with 86, lenscrafters.com with 76, eyebuydirect.com with 74. When an assistant explains Warby Parker, it is mostly reading its competitors' websites to do it, so the comparison gets written by the brands it's being compared against.
You can't tell from the outside which of those you are, and the list is the only place it's written down.
Want to see yours? The free check starts tracing the sources behind the answers about your brand. The $39 report runs full depth across ChatGPT, Gemini, and Claude, and traces every source behind your answers by type, including any that cite a competitor but never you. It lands in your inbox in under an hour, with no account and no sales call.