AI search kills the click: marketers rethink attribution

Fast Company ↗

The gist

AI-powered zero-click search is breaking the web's click-based measurement system, forcing marketers to rethink how they prove what actually works.

What to know

  • Zero-click Google searches soared from 56% to 69% in a year, with click-through dropping to just 8% when an AI summary appears.
  • Hybrid shopping habits—online discovery, offline purchase, and vice versa—have made it nearly impossible to track conversions from start to finish.
  • Most marketing teams juggle six measurement tools, but only 18% claim a clear view of impact—so brands are swapping platform stats for independent validation and experiments.

AI Overviews Upend Traffic Flows

By late 2026, AI-powered search had erased traditional referral signals, with zero-click answers driving double-digit drops in organic website visits and leaving marketers blind to most user journeys.

By September 2026, the measurement problem was no longer theoretical because AI search had already changed what marketers could observe. DesignRush reported that after Google AI Overviews launched in May 2024, zero-click searches rose from 56% to 69% in a year, meaning “just shy of 7 out of 10 Google searches end before anyone visits a brand’s website,” while around 80% of consumers were getting zero-click answers for at least 40% of searches, contributing to organic traffic declines of roughly 15% to 25% and weakening the referral signals attribution systems depend on.

What surfaced between late September and November 2026 was that these behavior shifts were breaking attribution across both digital and retail journeys, not merely reducing traffic. DesignRush found in a study of 68,000 real queries that click-through fell to 8% when an AI summary appeared versus 15% when it did not, and warned that when Google AI Overviews, Perplexity, or ChatGPT use a brand’s expertise without sending a visit, “GA4 can’t record a visit that never happened”; iTWire captured the broader reset bluntly: “The click used to be the smallest honest unit of marketing measurement… That chain is now bending in the middle.”

Sources

Clicks Vanish, Data Trails Fade

AI search formats are dismantling click-based attribution, as experiments show outbound traffic plummets and cookie trails disappear—leaving analytics tools with only a fraction of the real story.

Zero-click AI search breaks the basic logic of click-based attribution because the interaction that shapes demand often never produces a measurable visit. As Smarter Marketer put it, AI platforms are “very much designed to keep users on the platform they don't want you clicking off,” so “click through rate is super low, it's around 1 to 2%,” meaning “what you see in GA is a very small snippet of that”; that aligns with Pew Research Center’s March 2025 browsing panel, which found users clicked a traditional result on 8% of visits with an AI summary versus 15% without one, and clicked a link inside the summary on only 1% of visits.

The disruption is not just anecdotal: experiments show AI answer formats causally remove the click and cookie trails that older attribution models depend on, while marketers also report that the remaining traffic is hard to classify cleanly. Marketing in the Madness described AI search traffic as still low at “10%” and said Google AI Mode can be “a little murky” because “you’re entering a bot,” while AI Answers Reduce Clicks, but the Size Varies cited Agarwal and Sen’s 2026 experiment showing AI Overviews cut outbound organic clicks by 39.8% and raised zero-click searches by 34.5%, and a separate study of 1,100 active U.S. participants found AI Mode reduced external click-through by 18.8 percentage points versus normal Google search.

Sources
Smarter Marketer#SEOForLunchGrowth MemoMarketing in the Madness

Channel Switching Scrambles Tracking

Consumers now mix AI-driven online research with offline purchases and vice versa, fracturing the conversion path and pushing final sales far outside the reach of standard digital tracking.

The shopper journey now routinely starts on a screen and ends somewhere else. On Bloomberg Surveillance, Tiffany said “most consumers are still using AI more in their initial shopping research and inspiration” and that “about 50% of consumers plan on using AI in this holiday season,” while the study says “Artificial intelligence is becoming a regular part of how consumers research products,” with “48.5%” having “used an AI tool to research a purchase during the past year.” Fast Company also noted that “45% of in-store shoppers use AI in their buying journey,” showing how discovery increasingly happens digitally before consumers ever reach the aisle or register.

What breaks attribution is not just digital research but the scale of channel switching before purchase. Retail Dive, citing eMarketer’s Path to Purchase 2025, reported that “48% of U.S. shoppers discover new brands and products in physical stores, ahead of e-commerce platforms (41%) and social media (39%),” that “when shoppers discover a product online, 40% still visit a physical store before making a purchase,” and that after in-store discovery, 31.5% buy right away—well above retail sites and brand sites—placing conversion in places standard digital tracking cannot follow.

Sources
Bloomberg SurveillanceFast CompanyRetail Dive

Analytics Chaos Hits the Mainstream

Most marketers now juggle six or more tools yet lack a unified view, as measurement fragmentation distorts budgets and leaves even increased spend disconnected from clear growth outcomes.

This is no longer a niche analytics headache but a majority-scale operating problem. MarTech reported that Bitly’s “The Marketing Visibility Report” found the average marketing team now uses six different tools to measure performance, yet only 18% say they have a clear view of what is actually working; the same report found 72% of marketers use organic social media even though it is “the channel where visibility problems are most severe,” showing that measurement gaps already sit inside mainstream channel decisions rather than at the edges.

The commercial consequences are already large enough to distort budgets and strategy. StackAdapt’s October 19, 2026 “State of Programmatic Advertising 2027” found “31% report fragmentation-related budget waste,” while connected measurement remains uneven, with “58% use connected customer data, against 44%” among others, and its companion report said top performers are twice as likely to expand their channel mix; meanwhile, Experian found 76% of marketers have more tools and data than ever, but only 28% can clearly and consistently connect marketing activity to growth, even as 69% of organisations increased budgets without necessarily producing commercial growth.

Sources

Incrementality Becomes the New Gold Standard

Marketers are ditching platform metrics for independent validation and real-world experiments, exposing dramatic gaps between reported and actual impact—and rewriting the rules for budget decisions.

What is replacing platform-reported attribution is not one new metric but a stack: external validation, multi-source evidence and experiments tied to budget decisions. On The WARC Podcast, one speaker put the trust problem bluntly: “independent validation is 100% key” because platforms “are still biased towards their own platform,” while at Goafest 2026, a panel titled “From Platform-Defined to Brand-Aligned: A Reset in Measurement” called for “independent, third-party verification,” with Dhiraj Gupta, CTO of mFilterIt, warning that “the maker cannot be the checker”; that logic is now shaping how marketers separate optimization inside platforms from proof outside them.

That stack is becoming operational. Mobile Dev Memo heard that “huge amounts of people who historically weren't even analysts” are now “really diving into the data,” asking, “wait, if Facebook is actually that performant and we just increase spend on Facebook by 30%, why didn't our sales go up?” MarTech described teams pre-writing decision rules — “If iROAS comes back above 3.0, we’re going to increase spend 50%… Between 1.5 and 2.0, we’ll reduce spend 25%” — because platform and causal readings can diverge sharply: “Meta reports a ROAS of 4.0 during the test, while the incrementality test measures an iROAS of 2.0,” meaning Meta’s reported return is twice the incremental return measured by the test. Marketers are also stitching together broader data sources so experiments reflect business reality rather than a single dashboard. Exchange4Media reported that Truelift “designed a six-week plan in which advertising across Meta and quick commerce was stopped” for Sweet Karam Coffee, finding Meta drove “around 60-70% of new users” and that its efficiency was “understated by 50%” when quick-commerce effects were excluded; Digiday showed the same push toward journey-wide measurement, with The Trade Desk citing a campaign where “both reductions were more than 25 percent,” evidence meant to show incremental households and outcomes, not just credited clicks.

Sources

Get the stories behind the trends

Deep-dive reporting and the weekly brief, in your inbox.