Google Priced Q2 at $63.27B and Rounded Your Clicks to 'Billions'
Google SVP Nick Fox said on LinkedIn on July 17, 2026 that Search sends "billions of clicks to websites every week through AI features in Search alone." Five days later Alphabet reported Q2 Search revenue at $63.27 billion, up 17% year over year, to the exact dollar. The click figure came with no baseline, no methodology, and no site-level breakdown.
"Billions" is a magnitude, not a measurement
In one narrow sense, Fox's post was genuinely news. It was the first time Google attached any number at all to AI-feature referral traffic, after two years of declining to quantify it.
The trouble is that the number has no denominator, no prior-period comparison, and no unit of analysis smaller than "the web." Liz Reid's earlier framing had the same shape, describing total organic click volume as "relatively stable" year over year while publishing no charts and no year-over-year figures you could check it against. Google also added that the clicks are "quality clicks," meaning visitors who stay longer and engage more, which is another claim shipped without an instrument attached.
Compare that to how the same company talks about money. Alphabet put Search and other revenue at $63.27 billion for Q2, up 17% year over year, down from 19% growth in Q1 2026 and up from 15% in Q4 2025. Three quarters, three decimal-precise growth rates, a trend line you can actually model against. You cannot model against "billions."
I don't think that gap is an oversight. Revenue disclosure is legally compelled and audited. Click disclosure is voluntary, and voluntary numbers tend to get published in the form that flatters.
Both statements can be true, which is the part most coverage skips
The reflex is to treat Google's claim and the traffic-decline studies as a straight contradiction where one side has to be lying. In most cases I've seen, they are both describing real things at different levels of aggregation, and reading them as a fight is what keeps people from acting.
AI features expanded the query surface. Sundar Pichai's framing on the earnings call was that Google's "popular AI features are driving Search query growth." If total queries rise enough and per-query click-through rate falls, total clicks can still go up. Aggregate rising, your share falling, same quarter, no contradiction.
The independent measurements are all pointed at the per-query half. Pew, working from the browsing data of roughly 900 US adults, found users clicked a traditional result on 8% of searches that showed an AI summary versus 15% of searches that did not. Ahrefs, across 300,000 keywords of Search Console data, put the position-one CTR drop at 58% when an AI Overview appears, up from a 34.5% drop measured in April 2025. Google disputes those methodologies, which is fair enough as a position, though disputing someone's methodology while declining to publish your own is a weak place to argue from.
We covered the strongest version of this evidence when the first randomized test measured a 39.8% click reduction with paid results untouched. Randomized beats observational, and it landed in the same direction as everything else.
So the practical read on "billions of clicks" is that the pie got bigger. It tells you nothing about your slice.
A number without a denominator is a press release with a decimal point in it.
The report Google shipped has an impressions column and no clicks column
This is where an abstract transparency complaint turns into an operational one, and it is the detail I'd want every SEO lead to know before their next reporting cycle.
Search Console's generative AI performance reports do exist now. They show impressions of your URLs inside AI features, broken out by page, country, device, and date. What they do not show: clicks, click-through rate, average position, or query data. There is no API access either. Google has said click data is planned for a later version without committing to a date.
And there's a wrinkle that quietly breaks the obvious workaround. A click on an external link inside an AI Overview or AI Mode does count as a click, but it lands in your overall Performance report and never appears in the generative AI report. So the one report with "AI" in the name can confirm that you were cited and cannot tell you whether anybody actually came. The clicks are in the building. They're just filed somewhere you can't separate them from everything else.
Which means that when Google says show us your data, and the answer is supposed to be the AI report, the AI report is structurally incapable of answering the question being asked. Publishers have been blunter about this than I'm being.
The 20-minute audit that gets you an actual number
You can approximate the missing metric with what is already sitting in your account. It is not clean attribution and I wouldn't present it as such. It's directionally right, which beats "billions" by a wide margin.
- Open Search Console, go to Performance, then Search results. Set the date range to the last three months and compare it to the same period last year.
- Filter Search type to Web. Then apply a query regex filter for informational intent. Something like
^(what|how|why|when|who|which|is|are|can|does|best|vs)catches most of it. Save it as a view so you're not rebuilding it monthly. - Record impressions and clicks for that segment in both periods and calculate CTR yourself for each. Don't trust the summary tile, pull the raw two numbers.
- Run it again with those question-word queries excluded, which gives you a commercial and navigational control group that AI Overviews trigger on far less often.
What you are looking for is the spread between the two segments. If informational impressions are up while informational clicks are flat or falling, and your control segment holds roughly steady across the same window, you are being cited rather than visited. The size of that spread is your own version of the number Google will not publish.
From what I've seen discussed publicly by people sharing real account data, a 15 to 25 point CTR gap between those two segments over a full year is roughly where the pattern stops being noise. Under that, you more likely have a ranking problem or an intent-mix shift, which is a completely different fix and worth ruling out before you rebuild anything.
Then cross-reference against the generative AI report's impressions by page. Pages carrying heavy AI-feature impressions while their Performance-report clicks decline are your citation-without-traffic pages. Those are the ones to either restructure around something a citation cannot satisfy, like a calculator, a template, a login, or proprietary data, or quietly deprioritize. This is the same argument Avinash Kaushik was making when he told agencies to cut the retainer and own the measurement first.
What I'd stop waiting for
To be fair, none of this is entirely new. Google has always disclosed aggregate framing and let publishers infer the rest, and there was never really a golden age of site-level referral transparency to return to. It just feels a lot less survivable now that the aggregate number and the individual case have visibly come apart.
My guess is that the clicks column in the generative AI report doesn't ship before Q2 2027, and when it does it arrives aggregated at a level that still won't reconcile against your analytics. I'd love to be wrong about that. I'd also plan as though I'm not.
Publishers have spent two years asking Google to prove the claim, and Google has spent two years answering with a magnitude. That exchange doesn't seem likely to resolve on a timeline that helps anyone's Q3 planning. Building your own baseline is a lot less satisfying than being handed one. It's also the version you can start on Monday.
Notice Me Senpai Editorial
