AI Search Is Taking Traffic—but Sending Better Customers

Traffic has been the organizing metric of digital marketing for so long that a falling session count still looks like an uncomplicated failure. Fewer visits usually mean less reach, fewer opportunities to convert, and an uncomfortable agency report.

AI search is disrupting that logic.

Data from marketing agency Brainlabs, reported by Digiday on August 25, 2026, tracked 54 advertiser clients over 14 months. Organic sessions fell 10.5 percent, from 140.1 million to 125.4 million. Referrals from AI platforms rose 163 percent during the same period, while AI-driven key events increased 335 percent.

The referral volume remained small. Its apparent quality did not.

Visitors arriving from ChatGPT, Copilot, Gemini and Perplexity generated key events at 1.5 times the rate of organic-search visitors. AI search is not replacing the traffic it absorbs. It is sending a narrower group of users who appear more ready to act.

For marketers, that creates a measurement problem with commercial consequences. A channel can now look weak on volume and strong on behavior at the same time.

What the 54-Client Dataset Actually Shows

Brainlabs examined Google Analytics data over a 14-month window beginning in January 2025. The sample included 24 companies in the United States, 24 in the United Kingdom, and six operating in other markets. Among the advertisers, 29 employed at least 1,000 people.

In September 2025, AI Overviews began appearing on at least 30 percent of U.S. Google results pages. Brainlabs treated that point as an inflection in its client data. Average monthly organic sessions subsequently moved from roughly 20 million to below 15 million, while AI referrals reached approximately 200,000 per month.

Of the 54 clients, 46 experienced a traffic decline

Those figures offer something the generative engine optimization market has often lacked: observed client behavior across a varied advertiser portfolio. They remain one agency's sample rather than a representative map of the entire advertising economy. The September timing also cannot establish that AI Overviews caused every decline. Search demand, algorithm changes, seasonality, and client-specific factors may have contributed.

Even with those limits, the pattern is difficult to dismiss. Organic traffic fell across most of the sample, AI referrals grew rapidly, and those referrals produced disproportionately more recorded activity.

Separate Demandbase data points in the same direction on referral growth. Across an enterprise dataset covering more than 11 billion website visits, monthly ChatGPT referrals to measured B2B properties increased from approximately 645,000 in June 2025 to 2.6 million in June 2026. That 303 percent increase does not resolve the conversion question, but it confirms that fast referral growth is appearing beyond a single agency roster.

Traffic Loss Is Following Search Intent

The Brainlabs results varied considerably by sector. Fitness, financial technology, insurance and consumer packaged goods suffered some of the largest organic declines. Retail, beauty and entertainment were among the least affected and recorded some of the strongest increases in AI referrals. Retailers accounted for 16 of the 54 clients, giving that category meaningful weight within the sample.

The differences appear connected to the kinds of questions consumers ask.

Google's AI Overviews have been especially prevalent for educational and discovery searches. Those are queries where a synthesized response can satisfy the user without requiring a visit to the underlying source.

Fitness and financial products invite explanation and comparison. A person researching training methods, insurance coverage or the difference between financial products may receive enough information directly from an AI-generated response to postpone or avoid a click.

Retail searches are often closer to an immediate transaction. A user looking for a particular product, price, size or store still has a practical reason to reach the merchant. Brainlabs found retail comparatively resilient while AI Overviews remained less common for transactional searches.

That protection may be temporary. Brainlabs observed that AI summaries were beginning to move into commercial and transactional queries. If that expansion continues, retailers may encounter the same pressure later in the buying journey.

Why the Remaining Visitors May Be More Valuable

An AI referral often follows a longer interaction than a conventional search query.

A user can describe a problem, add constraints, reject an initial recommendation and request a comparison before opening a cited page. Part of the discovery process has already occurred by the time the visit reaches the advertiser. The click may represent verification, deeper evaluation or a desire to act.

This is a plausible explanation for the higher event rate. It is not a causal finding from the Brainlabs research.

There is also a selection effect. AI systems answer many questions without sending anyone away. The smaller group that chooses to leave the interface may naturally contain more motivated users. In conventional search, a casual click carries little cost. In a conversational answer, opening an external source can signal that the generated summary was no longer sufficient.

The word "customers" therefore requires care. Brainlabs measured key events chosen by each advertiser. These included purchases and newsletter registrations, but they also included reaching the bottom of a page. A key event can indicate meaningful engagement without producing revenue.

The headline finding is still commercially relevant. AI visitors took recorded actions at a substantially higher rate. Whether they become more profitable customers must be tested against qualified leads, completed purchases, deal value, and retention.

Traffic Remains Useful, but It Has Lost Its Authority

A 10.5 percent decline in organic sessions cannot be waved away as an outdated metric. Traffic still provides reach, supports advertising revenue, and creates opportunities for brands to establish direct customer relationships. Losing it can impose a real economic cost.

What has changed is the authority of traffic as a stand-alone performance judgment.

An AI platform may expose a brand to a consumer, influence a shortlist, and answer several purchase questions without generating a referral. The eventual customer may arrive through a branded search, a manually entered address, or another channel. Analytics will see the final entrance more clearly than the earlier influence.

Attribution can disappear even when an AI user visits the site.

Brainlabs noted that people sometimes open a destination in another browser window after receiving an AI answer. The resulting session may be recorded as direct traffic because the referral information is missing.

Google Analytics has begun recognizing "AI Assistant" as a default channel for sources such as ChatGPT, Gemini, Copilot and Grok. However, visits from Google's AI Overviews and AI Mode remain classified as organic search. A marketing dashboard can therefore divide similar AI-mediated behavior between two categories while losing other visits inside direct traffic.

Traffic should remain in the report. It should no longer be allowed to deliver the verdict by itself.

Agency Reporting Has to Connect Visits With Value

The familiar search report begins with rankings, impressions, clicks, and sessions. Conversion metrics often appear later, after the agency has explained whether traffic moved in the desired direction.

That order becomes less useful when search platforms complete more of the research inside their own interfaces.

Agency reporting now needs to show how much discoverable demand produced a website visit and what happened after arrival. AI referral volume belongs beside event quality, revenue, and pipeline progression. Branded-search movement and changes in direct traffic can provide additional context when referral data is incomplete, although neither should be presented as proof of AI influence without further analysis.

The underlying event definitions also need scrutiny. A dashboard that combines purchases with page-depth events can make a channel look commercially stronger than it is. Agencies should separate revenue events from lighter engagement signals and compare AI referrals with organic search using equivalent outcomes.

Small percentages require absolute numbers. A 163 percent increase can still describe a minor channel, as the Brainlabs data demonstrates. Executives need to see the growth rate, the actual referral count and the amount of lost organic traffic that the new channel replaces.

SEO Retainers Need a New Definition of Performance

AI search does not eliminate the underlying work of SEO. Technical accessibility, reliable information, and clear site architecture still help search engines and AI systems retrieve useful material. Strong original content also gives models something worth citing.

The retainer changes when traffic can no longer serve as the principal proof of value.

Agencies will increasingly be expected to monitor whether a brand appears in relevant AI answers, which sources are being cited and whether those appearances lead to useful business activity. Content work will need closer coordination with product data, public relations and the independent sources that influence AI responses. Regular updates will become more important where prices, product details or market facts can quickly become stale.

Clients should resist replacing old traffic theater with new visibility theater. A proprietary "AI score" has little value if nobody can explain the prompts behind it, the size of the relevant audience or the connection to commercial results.

Retainers should be judged on a combination of discoverability, qualified demand and business outcomes. The exact balance will differ by sector. A retailer may continue to place considerable weight on sessions and transactions, while a complex B2B company may care more about account quality and pipeline contribution.

Better Visitors Cannot Yet Pay the Entire Bill

The Brainlabs findings offer marketers a more useful interpretation of the AI-search transition than either panic or enthusiasm.

Organic traffic is being lost. AI referrals are growing quickly but remain far too small to fill the gap. The visitors who do arrive appear unusually engaged, although the available data does not yet establish their lifetime value or profitability.

That leaves marketing teams with a demanding reporting task. They have to protect the reach that conventional search still provides while learning to recognize value created inside AI-mediated discovery.

The next quarterly review should be able to explain where search traffic disappeared and whether the remaining visits produced stronger commercial outcomes. Without both answers, a company may cut a valuable program because its audience shrank—or pay for an impressive growth rate that has barely touched the business.