The Zero-Click Tax: What AI-Answer Erosion Is Really Adding to the Cost of a Paid Lead

The number that changed the conversation

Start with the figure that is hardest to argue with. SparkToro and Similarweb's analysis of US Google search behavior found that 68.01 percent of searches in the first four months of 2026 ended without the person clicking anything, up from 60.45 percent in 2024. That's a jump of nearly eight percentage points in two years, on a metric that was already the majority outcome before this study started. Zero-click search, meaning the search resolves for the person right there on the results page, is not a fringe behavior anymore. It is what happens most of the time when someone searches on Google in the United States.

This figure alone doesn't tell you why clicks are disappearing, only that they are. The next layer of evidence gets at the why, and it points squarely at the AI Overview.

More than two out of three Google searches in the US now end without a single click, up from just over three in five two years ago.

What happens on the query where the answer already appeared

Four independent research efforts, using four different methods, converge on the same underlying pattern even though their exact numbers differ. Pew Research Center tracked the real browsing of 900 US adults across 68,879 Google searches: when an AI summary appeared, people clicked a standard result in 8 percent of visits, versus 15 percent when it didn't. Seer Interactive's longitudinal panel of 3,119 informational queries across 42 organizations, spanning 25.1 million organic and 1.1 million paid impressions, found organic click-through fell 61 percent and paid click-through fell 68 percent on AI-Overview queries. Ahrefs, analyzing 300,000 keywords, found position-one click-through was 34.5 percent lower on average when an AI Overview appeared. And a randomized field experiment (Agarwal and Sen, an academic working paper, its topic and existence independently confirmed though its exact figures could not be independently re-derived this session) reported outbound publisher clicks falling by roughly 39.8 percent.

These are not the same study wearing different clothes. They come from a consumer-behavior researcher, a marketing-technology panel, an SEO-data vendor, and an academic field experiment, each looking at a different slice of the problem, and they land in the same place: on the query where the answer already showed up, the click mostly doesn't happen. That's the part of this story that is genuinely well established, not a single-source claim you should treat cautiously.

Four separate research teams, using four different methods, arrive at the same shape: when the answer shows up on the page, the click mostly doesn't happen.

Being cited is starting to matter more than being ranked

If the click on a query is shrinking, where it goes still matters enormously, and the evidence suggests it is starting to follow citation inside the answer rather than position on the page beneath it. Pew found that only 1 percent of all visits included a click on a link inside the AI summary itself, which sounds small until you realize it means almost the entire click pool that survives an AI Overview is now concentrated on a tiny fraction of the real estate on that page: the handful of sources the AI Overview chose to cite.

This is a structural change, not a cosmetic one. Under the old model, a business could rank fourth or fifth on a page and still capture a meaningful share of clicks, because people scanned several results before choosing. Under the new one, if you are not one of the sources the AI Overview surfaces, you are competing for a sliver of a shrinking pool of clicks that bypass the summary entirely. Being cited has become closer to a precondition for visibility than an optional bonus on top of ranking well.

The market-level number tells a calmer story

Here is where the evidence gets genuinely more interesting than a simple decline narrative. Alphabet's own Q2 2026 results, reported July 22, 2026, show Google Search and Other revenue up 17 percent year over year, in the same quarter that AI Mode crossed 1 billion monthly active users. And WordStream's 2026 benchmark, drawn from 13,474 US search campaigns over the prior twelve months, found average cost per lead actually fell to $66.69, the first year-over-year decline in five years, even as average cost per click rose to $5.42.

If query-level clicks are collapsing this sharply, how is the market-level picture holding up, or even improving on the cost-per-lead side? That is the real tension in this evidence, and it is the one the existing research has not fully resolved.

Search ad revenue grew 17 percent in the same quarter that AI Mode crossed a billion monthly users. Growth and disruption are happening inside the same business at the same time.

Why both things are true at once

The most plausible reconciliation, and it is a reconciliation, not a proven mechanism traced by any single study, comes from looking at what kind of queries are actually triggering AI Overviews in the first place. Ahrefs' separate analysis of 146 million search results found that about 99.9 percent of keywords that trigger an AI Overview carry informational intent, the kind of query someone types when they want a fact, not when they're ready to buy or call. Those are exactly the queries with the lowest value to an advertiser to begin with, and the ones a paid-search account was least likely to be bidding on aggressively in the first place.

If the clicks disappearing are concentrated on low-intent, mostly-informational queries, while transactional, ready-to-buy queries keep behaving closer to normal, then a market-wide average like cost per lead can hold steady or even improve, even while the click-through collapse on informational queries is severe and completely real. This is a reasonable reading of the evidence, but it has not been directly measured in any single study; no source in this research reconciled the query-level collapse and the market-level resilience inside one dataset. Treat it as the most defensible explanation available today, not as an established fact.

The open question: is this a tax, or a transfer

The most important thing this research could not find is also the most consequential one for a marketing budget: what happens to the dollars and the clicks that leave a business's paid-search account when an AI Overview intercepts a query. If advertising eventually follows the customer into the AI-answer surface itself, whether inside Google's AI Mode or a competing engine, this could look more like a transfer: a reshuffling of where the same dollar goes to reach the same person. If it doesn't, and the advertiser dollar and the click both just disappear, this is a tax: a real net cost with nowhere for the spend to follow the customer to.

No published source, in this research or otherwise, currently quantifies which of those two is happening, or in what proportion. That is not a gap in this study specifically; it is a gap in the public record. It is worth being skeptical, too, of predictions that got ahead of the evidence: Gartner's February 2024 forecast that traditional search engine volume would drop 25 percent by 2026 because of AI chatbots is a real, dated, attributable prediction, but multiple mid-2026 retrospective assessments find it has not materialized at that scale. The lesson isn't that AI isn't changing search, the evidence above proves it clearly is, it's that the size and shape of that change is still being measured in real time, and forecasts made early in that process have a track record of overshooting.

Whether the dollars leaving classic search ads are transferring to a new surface or simply disappearing is the question that matters most for a budget, and no published source has answered it yet.

What this shift is already doing to retail traffic

One place the redistribution of attention is measurable, not theoretical, is retail. Adobe, tracking over a trillion visits to US retail websites, found that traffic referred from generative AI sources grew 1,200 percent in February 2025 compared with July 2024, with a 1,300 percent year-over-year spike during the November-December 2024 holiday shopping season. That's a genuinely new referral channel showing up in the data, on top of, not instead of, the click losses documented elsewhere.

This figure is retail-specific and should be read that way; nothing in this research extends it to other categories like home services, legal, or healthcare, and no source cross-tabulated AI Overview trigger rates against industry-level cost data to show which categories are hit hardest or benefit most. What it does show is that the story isn't only subtraction. Some of the attention leaving the classic results page is landing somewhere new, and a business that isn't watching for that new arrival is only seeing half the picture.

Where the evidence stops, and what that means for you

Most of the click-through and trigger-rate figures in this study, the SparkToro and Similarweb zero-click numbers, the Seer Interactive panel, and the Ahrefs analysis, come from a single research firm's own analytics or crawl panel rather than from Google, an academic body, or a government source. That means their magnitudes vary from vendor to vendor and should be read as directional evidence of a real trend, not as a precise, universally agreed number. The field experiment by Agarwal and Sen is an academic working paper whose direction is confirmed but whose exact figures could not be independently re-verified this session. Pew's field research, Alphabet's own earnings disclosure, and WordStream's aggregate benchmark are the most independently checkable figures in this study. No study has yet isolated a clean, market-wide cost-per-lead premium attributable to AI answers specifically, separated out from ordinary auction competition, general inflation, and bidding-algorithm changes. And no methodologically transparent, small-business-focused survey measuring cost-per-lead change from AI-answer erosion specifically was found anywhere in this research.

What that means practically is this: don't assume every lead in your account now costs more because of AI Overviews, the broadest data available doesn't support that as a blanket rule. Do assume that on the specific, mostly-informational queries where an AI Overview shows up, your click-through is falling sharply whether or not your cost per lead reflects it yet, and that being cited inside that answer, not just ranking beneath it, is becoming the thing that determines whether you get any of the attention that query still has to offer.