Measuring What an AI Overviews Citation Is Actually Worth
By Antonio Caruso, Caruso Martech
Published Sep 24, 2026 · Updated Sep 24, 2026 · AI Search & Experience
Click-through rates on AI Overview searches have fallen by more than half since last year, but cited brands still win over twice the clicks of uncited ones. A framework for measuring that value directly.
Getting cited in a Google AI Overview used to feel like a small win nobody could put a number on. Click-through rates on AI Overview searches have fallen by more than half over the past year, and the citation itself rarely shows up as a clean line in your analytics. Most teams still report on clicks anyway, because that is the metric they know how to defend, even when clicks are the wrong measure for what a citation actually does.
Quick answer: what actually measures the value of an AI Overviews citation?
- Compare clicks per impression for cited pages against uncited pages on the same query in Search Console, since cited pages still convert attention at a meaningfully higher rate.
- Track citation depth as three separate outcomes: a linked source, a text mention with no link, and a product recommendation, since each carries different weight and needs its own column in the report.
- Run a fixed set of buyer-style prompts across ChatGPT, Perplexity, Gemini, and AI Overviews every month, and track mention rate as its own metric alongside citations.
- Watch branded search volume and direct traffic as a secondary signal, since a citation builds recall in the many cases it never sends a click.
- Report AI Overviews citations as a brand visibility asset feeding the top of the funnel, separate from channels judged purely on direct conversions.
Why click-through rates on AI Overview searches keep falling
Click-through rates keep falling because AI Overviews increasingly answer the query directly inside the results page, so fewer searchers ever need to leave Google to get what they came for. The decline has accelerated through 2026 even as Google adds more visible citation links.
Pew Research found that only about 8% of searches with an AI Overview present end in an organic click, compared with roughly 15% when no AI Overview appears, essentially halving the click rate. Ahrefs reached a similar conclusion from a different angle: its February 2026 update measured a 58% lower click-through rate for the page ranking first when an AI Overview is present, up from 34.5% when it ran the same study in April 2025.
That trend line matters more than either single number. A gap that nearly doubled in ten months means whatever click-based reporting worked at the start of last year is already unreliable, and it keeps drifting further from the truth every quarter you keep using it.
What a citation is actually worth, even without a click
A citation is worth roughly twice what an uncited ranking is worth on the same query, measured in clicks per impression, and that gap alone justifies treating citations as a real asset in the report. The number holds up at scale, drawn from billions of tracked impressions.
Seer Interactive's April 2026 analysis of 2.43 billion impressions across 53 brands found pages cited inside an AI Overview earn about 2.1% click-through, against roughly 0.9% for pages ranking on the same results page without a citation. In raw terms, queries without an AI Overview send about 33,500 clicks per million impressions, cited brands receive 20,743, and uncited brands receive only 9,445.
That last figure is the one worth sitting with. Uncited brands are losing more than two-thirds of the traffic that query used to send them, which means the cost of not earning a citation is now larger than the value of the citation itself. This is the piece of the attribution gap that rarely makes it into a board deck, since it shows up as a slow bleed in organic traffic over several quarters.
Build a citation depth score instead of counting clicks alone
A citation depth score tracks three separate outcomes instead of collapsing them into one number: a linked source citation, a text mention with no link, and a product or brand recommendation named directly in the answer. Each one signals a different level of trust from the model, and each converts differently.
Treating a mention and a citation as the same event, as we've covered before, hides exactly the kind of visibility problem worth knowing about. A page that earns three text mentions a month but zero linked citations has a different problem than one earning citations from a single low-authority source repeatedly. Neither shows up if you're only counting whether the brand appeared at all.
Building the underlying visibility that earns any of these three outcomes still comes back to the same on-page fundamentals: clear structure, direct answers near the top of a page, and claims a model can quote without stitching together multiple paragraphs. The depth score tells you which part of that work is paying off.
Set a measurement rhythm you can actually keep up with
A measurement rhythm only works if it survives being handed to whoever owns reporting next quarter, which means monthly is the right cadence for the full prompt set and weekly is enough for a smaller check-in. Running it once and calling it done tells you almost nothing, because AI answers shift week to week even for the same prompt.
Set 20 to 30 buyer-style prompts that mirror how your actual customers search, the same discipline this site has written about before for general AI visibility tracking, and run the full set against ChatGPT, Perplexity, Gemini, and Google AI Overviews once a month. Pull a 10-prompt subset of the highest-intent queries and check those weekly, since one bad week in a fast-moving answer engine only becomes meaningful once you see it repeat.
Log citation depth and mention rate every run in the same spreadsheet or dashboard used for clicks. Keep them next to each other. The comparison only works when both numbers live in the same place instead of separate reports nobody cross-references.
The reporting gap this creates, and how to close it
The reporting gap shows up when a citation earns real value in brand recall and downstream branded search, but the only line item in a monthly report is organic clicks, which makes a genuinely working channel look flat or declining. Closing it means giving citation depth and mention rate their own rows in the report, tracked separately from the click-based channels they don't behave like.
Google's own citation panel keeps changing shape and prominence, which affects how models decide what to surface as much as it affects how visible a citation looks once earned. Treat this measurement as a routine to repeat every time the interface shifts, since a setup that worked last quarter can go stale without warning.
If your reporting still leans on clicks alone to judge whether AI Overviews visibility is working, it's worth adding citation depth and mention rate before the next review. We build that kind of measurement into the broader marketing systems we run for clients, or if you want a read on what your current AI visibility actually looks like, reach out.
Caruso Martech
We write about marketing systems, attribution, and growth operations because these are the problems we work on every day. If something in this post is relevant to what you're building, we're happy to talk through it.
Related insights
AI Search & Experience • Aug 1, 2026
How AI Search Engines Decide Which Sites to Cite
ChatGPT, Perplexity, and Google AI Overviews cite some brands constantly and others never. Here is what the data says actually earns a citation, and how a small marketing team can act on it.
Automation & Intelligence • Sep 16, 2026
Self-Adjusting Campaigns: What Marketing Automation Looks Like Beyond the Schedule
Most small teams still run automation on a calendar. Here is what it takes to move from scheduled sends to campaigns that adjust to real behavior, and when that shift is actually worth making.
Acquisition Systems • Aug 25, 2026
How to Measure the ROI of a Fractional Marketing Engagement
A practical framework for tracking what a fractional marketing engagement actually returns, from the cost baseline to the pipeline signals that show up before revenue does.