A brand built from nothing picked up 42,435 AI citations by August. Its own analytics counted about 100 visitors for the entire stretch.
One brand, 42,000 citations, about 100 visitors
SE Ranking's AI-visibility team ran that test on themselves. In February they built Civoryx, a fraud-tracking site with no history, no reviews and nobody searching the name.
They published 93 pages across the brand site and 11 partner sites, then tracked 165 buyer questions daily across ChatGPT, Perplexity, Gemini and Google's AI surfaces. Publishing stopped on March 31, and the site sat untouched after that.
Citations of the brand kept climbing anyway. The count went from 17,023 in April to 42,435 in August, growth that had nothing to do with new content since none had shipped in months.
Over that same window, GA4 recorded about 100 users. Search Console logged 11 clicks. The server logs found exactly one visit anyone could trace to an AI assistant.
Branded and non-branded questions split hard. Answers to questions naming the brand cited it 84.8% of the time by September, while answers to generic category questions cited it in just 1.8% of cases.
Most of those 42,000 citations backed up a fact for someone who hadn't decided to care yet. Perplexity carried the bulk of the non-branded mentions, at 30.6%. ChatGPT trailed at 8.3%, and Google's AI Overviews, Gemini and AI Mode barely registered.
One caveat is worth sitting with. The 11 external sites doing most of the non-branded citing belonged to the same company, which makes this a controlled test, not proof that earned coverage grows the same way on its own.
A commenter on the Reddit thread made a sharper version of the same point, calling the setup a private link network with better framing. It's a fair challenge. It doesn't change what the traffic numbers showed once the citations existed, though.
ChatGPT also wasn't holding still during the test. On August 9 it started listing far more sources per answer, then cut back from roughly 17 sources to about 5 on August 28.
Part of any brand's citation count that month depended on how many slots ChatGPT happened to be handing out that week. It had nothing to do with anything the brand did differently.
A citation answers the question before you ever get the click
A team running GEO work for clients posted a smaller, independent version of the same finding on Reddit in July. Out of every page their clients got cited on by at least one AI engine, 73% sent zero measurable traffic over a 90-day window, and some of those pages had 15 or more citations across different models.
73% of our AI citations drove zero traffic in 90 days — we were optimizing the wrong thing
Here's something I don't see talked about enough: getting cited by AI and actually getting traffic from those citations are two completely different games. We spent the first 4 months of our GEO work chasing citation counts. Every week we'd...
Their own explanation holds up. A citation that backs a factual answer satisfies the reader on the spot, because the AI already pulled the number or the definition into its response. The page ends up functioning as a fact-checking layer rather than a destination. Nobody needs to open a new tab to get something they already have.
AI search traffic grew 835% this year. Is your content ready?
RankControl generates 26 content formats optimized for ChatGPT, Claude, and Perplexity. Published on your domain, matched to your brand.
Where you sit in the answer decides whether anyone clicks
Being cited at all turns out to matter less than where the citation sits. In that same Reddit thread, pages that consistently landed as the first or second source in an answer pulled in about eleven times more clicks than pages that showed up third or later.
That tracks with how people read a list of sources, which is to say they mostly don't. A source link buried at position six is functionally invisible next to one sitting right under the sentence someone just read.
Chasing raw citation count rewards a page for being mentioned anywhere. The traffic almost entirely depends on cracking the top couple of slots, and a shrinking source list only makes that fight harder.
Patience changes the number too
Short articles got their first citation after a 21-day median wait, while longer how-to guides took a 126-day median to show up at all in the Civoryx test. Judging a new guide's citation-to-click rate a month after publishing penalizes exactly the format that takes the longest to start working.
That gap also explains why a monthly report can look flat for reasons that have nothing to do with quality. A page written in October may simply not be in the system yet, and checking too early just confirms a zero that was never going to mean anything.
The questions that still send a click
Not every citation carries the same ceiling. Informational queries, the "what is" and "how does it work" kind, rarely drive a visit no matter how well a page is cited, because the AI's own summary already closes the loop. Comparison, pricing and best-option queries behave differently, since an AI answer can gesture at an evaluation without ever fully completing it for the reader.
A separate study, run across ChatGPT and Perplexity over two six-month windows, found product and service pages growing fastest as a citation source. They were up 51% on ChatGPT and 23% on Perplexity, now the largest single group on both engines.
Help docs and support pages grew even faster, more than doubling on both tools. Official regulator and government pages, a category with no commercial reason to chase a click, grew from under 1% to about 5% of Perplexity's citations over the same stretch. An engine cites a page because it answers the question well. Click-through is nowhere in that decision.
One commenter on that thread raised a fair objection before anyone took the numbers at face value. There's no agreed list of what people actually ask an AI engine. Any citation-source study only describes the questions someone chose to run, not the market.
Our breakdown of citations vs. referral traffic covers which of the two numbers to lead with when they move in opposite directions.
Built by the team that got cited in 48 hours.
Content generation, backlink building, AI visibility tracking, and Google rankings. One platform, zero guesswork.
The click that never reaches your dashboard
Some of the value shows up in a way no analytics tool will ever credit to AI. On a GEO-focused subreddit, one agency owner described a pattern that kept repeating with new leads.
A buyer asks an AI engine for a shortlist, picks a name off it, then shows up on a sales call already asking about pricing. They never ask what the company does, because the AI answer already told them.
That buyer never clicked an AI-generated link. They searched the brand name directly or typed the URL from memory, so the visit lands in GA4 as direct or branded organic, with nothing connecting it back to the citation that put the name in their head.
The one place this ever surfaces is an open-text "how did you hear about us" field, and most people skip filling it in. The citation did real work. The dashboard just has no column for it.
What to track instead of raw citation counts
Raw citation volume is the easiest number to report and the least useful one to act on. A citation-to-click rate, the share of citations that actually produce a visit, tells you whether a count that's climbing is building an audience or just padding a vanity metric.
Split that rate by query stage. Lumping an informational "what is X" citation in with a "best X for Y" citation hides exactly the pattern that matters.
A monthly check worth running looks like this:
- Hold a fixed set of buyer questions steady, mixing informational and comparison types, so month-to-month change means something.
- Note which source position each citation landed in, instead of only whether it happened.
- Give new pages at least three months before judging their citation-to-click rate.
- Track branded and non-branded mentions separately, since they rarely move together.
Mention tracking vs. citation tracking is the right place to start if a brand mention and a cited source are currently the same line item in your reporting.
RankControl's AI Visibility Tracking records which questions earned a citation each week, instead of only a running total. A team can tell an encyclopedia-entry mention from one sitting on a buyer's shortlist, rather than guessing from the total alone.
Treat informational citations as a slower bet. They build the association an engine draws on later, in the same way a Google AI Overview citation pays off partly in clicks and partly in exposure a click can't capture. The traffic that does land from AI sources converts reasonably well once it arrives, which is one more reason the real problem isn't the quality of an AI-sourced visitor. It's that so few of the citations ever produce one.
Doing it yourself or handing it off
Splitting citation data by query stage and position by hand, every week, across five or more engines, is the kind of task that gets skipped the first time a launch gets busy. You can track it yourself with a spreadsheet and a fixed question list, updated whenever someone remembers to run it.
Or RankControl can run that same weekly check across all seven engines automatically, so the citation-to-click question has an answer whether or not anyone had time to pull it this month. That's the difference between catching a dropping rate in a weekly dashboard and explaining it to a founder a quarter later, after the pattern has already set in.
See your first AI citation report in under 5 minutes.
No setup calls. No onboarding meetings. Connect your domain and see where AI mentions your brand right now.
