AI search may be moving the website down the funnel

For more than two decades, search marketing followed a familiar sequence. Someone asked a question, a search engine returned links, and much of the researching, comparing and evaluating happened across websites.

AI search can interrupt that sequence. A person can learn about a category, compare brands, challenge claims and narrow a consideration set inside an answer engine before visiting any first-party site. The website remains important, but it may enter the journey later.

New Brainlabs data reported by Digiday provides a useful view of what that shift could look like. Across 54 clients, aggregate organic sessions fell from 140.1 million to 125.4 million, a decline of 10.5%. Sessions attributed to AI platforms increased 163%, while AI-driven key events rose 335%. Visitors referred by ChatGPT, Copilot, Gemini and Perplexity generated key events at 1.5 times the rate of organic-search visitors.[1]

The results do not establish that AI Overviews caused the traffic decline or that AI traffic converts 50% better. They do suggest that AI referrals and conventional organic visits may represent different moments in a consumer journey.

<h2>What Brainlabs measured</h2>

The dataset covers 54 Brainlabs clients: 24 US companies, 24 British brands and six operating in other markets. Twenty-nine employed at least 1,000 people. The group spans 19 sectors and includes 16 retailers.[1]

Digiday describes the data as covering January 2025 to April 2026. It also calls this a 14-month period, although those endpoints span 16 calendar months if counted inclusively. The defensible approach is to retain the reported endpoints rather than infer a reconciled duration.

Brainlabs identified September 2025 as an inflection point, when AI Overviews appeared on at least 30% of US search-results pages. Forty-six of the 54 clients experienced declining organic traffic. Across the complete group, sessions fell 10.5% from 140.1 million to 125.4 million.[1]

At the same time, AI referral sessions grew 163% and reached roughly 200,000 per month. Organic sessions had been running at about 20 million per month before falling below roughly 15 million.[1]

That denominator matters. The percentage growth in AI referrals is substantial, but the underlying channel remains far smaller. In this dataset, AI referrals did not replace the lost organic sessions on a session-for-session basis.

This finding belongs beside the recent <a href="/articles/ai-search-reduces-publisher-traffic-experiment">randomised Google experiment showing that AI search features can reduce external publisher clicking</a>. The experimental study provides causal evidence about clicks under controlled conditions. The Brainlabs analysis shows an observational pattern across advertiser sites. The two studies answer different questions, but both indicate that website traffic is becoming a less complete measure of AI-mediated influence.

<h2>The 1.5-times finding is a key-event rate, not a universal conversion rate</h2>

Brainlabs also examined Google Analytics key events. AI-driven key events increased 335% across the reported period, and visitors referred by four major AI platforms generated key events at 1.5 times the rate of organic-search visitors.[1]

That finding needs precise language. Google defines a key event as an event measuring an action that is particularly important to a business and says any collected event can be marked as a key event.[4] In the Brainlabs sample, the examples ranged from purchases and newsletter signups to scrolling to the bottom of a page.[1]

The result is therefore not a standardised purchase-conversion comparison. Each advertiser can define a different portfolio of meaningful actions. The study supports a narrower statement: within this client dataset, visitors attributed to AI referrals triggered advertiser-defined key events at a materially higher rate than organic-search visitors.

That is still strategically important. It suggests the visible AI-referral audience is not simply a smaller copy of the organic-search audience.

<h2>The website may be receiving a later stage of user</h2>

One explanation is selection. An answer engine can satisfy informational needs without producing a click. The people who eventually leave the AI environment may disproportionately need something the brand-controlled experience must provide, such as a price, quote, appointment, account, inventory check or transaction.

Under that interpretation, the AI system may handle more of the research and comparison while the website receives a smaller, more advanced subset of users. AI traffic would then show stronger engagement not because the channel made people inherently more valuable, but because the click occurred at a different stage.

Brainlabs’ own January analysis supports the broader direction. It describes AI search as producing longer, more informational queries, fewer immediate clicks and continuing influence over discovery and decisions.[2] Digiday also reports Brainlabs’ view that the lost organic traffic may contain more casual users while AI referrals include people further into product discovery.[1]

The mechanism remains a hypothesis. Brainlabs did not observe each person’s complete cross-platform journey, and the dataset cannot show that AI research caused an individual visitor to reach the website later. The pattern is consistent with a later-funnel or selection effect; it does not prove one.

<h2>AI referral traffic may be a handoff metric</h2>

Referral traffic is usually treated as an audience measure. In AI search, it may increasingly measure a handoff.

A person might spend several minutes asking an AI system about a category, comparing alternatives and refining criteria. The eventual click can mark the point at which the answer engine hands the person to the brand for verification, availability, service or transaction.

That is a different construct from total AI influence. An AI system can introduce a brand, explain its positioning and recommend it without generating a directly attributable visit. A later branded search may be recorded as Organic Search. A manually entered address or a new-window visit may appear as Direct. Brainlabs says some AI referrals are likely missed for precisely this reason.[1]

The visible referral is therefore only one transition in a larger chain:

Visibility → Consideration → Recommendation → Referral or navigation → Business outcome

This connects directly to <a href="/articles/behind-aeo-dashboard-ai-search-measurement">the measurement choices behind an AI Search Intelligence dashboard</a>. Visibility, Recommendation, Referral and Business Outcome are related, but they are not interchangeable measures.

<h2>The category pattern reinforces the journey-stage explanation</h2>

The effects were not uniform. Fitness, fintech, insurance and consumer packaged goods clients experienced some of the largest traffic declines. Retail, beauty and entertainment saw some of the smallest organic declines and some of the strongest AI-referral gains.[1]

Brainlabs said AI Overviews were still overwhelmingly triggered by educational and informational discovery searches, although they were beginning to appear against commercial and transactional queries. Retail was comparatively protected because transactional searches were less likely to trigger AI Overviews.[1]

Another useful question is how much of a category’s decision journey can occur before a website becomes necessary. Insurance buyers can ask an AI system about product types, coverage levels, providers and policy considerations. A shopper seeking a particular shoe in a particular size may need live inventory and checkout much sooner.

The 54-client sample cannot establish category-wide benchmarks. It does, however, show why a universal AI traffic expectation is unlikely to be useful. The effect can vary with information intensity, purchase cycle and the point at which first-party functionality becomes necessary.

<h2>Traffic loss does not necessarily mean influence loss</h2>

Conventional analytics observes what happens on the website and the referral information that survives the journey. It cannot fully observe what a person learned, compared or rejected inside an AI interface.

That creates a difficult possibility. A brand can lose informational visits while remaining present in the answers that replaced them. It can also gain branded searches or direct visits after an AI interaction without receiving attribution for the influence that created those actions.

The IAB is now developing an AI advertising measurement framework scheduled for November 12, 2026. Digiday reports that the work is intended to address attribution when referral and UTM signals disappear and will likely distinguish between AI creating awareness or intent and AI participating more directly in a decision.[3] Those categories are still under development, but the effort reflects the same measurement gap.

Traffic remains important. It is simply becoming less sufficient as a stand-alone measure of search influence.

<h2>AEO measurement needs to extend beyond a visibility score</h2>

Early AEO measurement has necessarily focused on observable outputs: prompts, responses, mentions, citations, position, sentiment and Share of Visibility. Brainlabs’ findings point to the next challenge: connecting what an AI system says with what consumers subsequently do.

A useful measurement chain separates the stages rather than collapsing them:

Observed demand → Visibility → Brand representation → Consideration → Recommendation → Handoff → Business outcome

Each stage answers a different question. Visibility shows whether the brand appears. Brand representation describes what the system says. Consideration asks whether the brand remains relevant as constraints accumulate. Recommendation records advocacy. Handoff captures referral or navigation. Business Outcome records what happens next.

The chain also exposes the denominator problem. A high key-event rate among the people who clicked does not estimate how many AI-influenced journeys included the brand, how many ended without a visit or how many reached the site through another channel. For the same reason, <a href="/articles/ai-visibility-no-common-denominator">an AI visibility score cannot be interpreted apart from the prompt universe and Measurement Frame that produced it</a>.

<h2>What the study does not prove</h2>

The analysis is observational and based on one agency’s 54-client roster. It is not a representative sample of every advertiser or website. Many factors changed between January 2025 and April 2026, including search algorithms, websites, competitors, investment levels, consumer behaviour and economic conditions.

The defensible conclusion is that organic sessions fell 10.5% across the Brainlabs dataset during a period in which AI Overviews became more prevalent and AI referrals grew rapidly from a small base. It is not that AI Overviews caused organic traffic to fall 10.5%.

The 1.5-times key-event rate also should not become a universal benchmark. The sample spans different sectors, purchase cycles, objectives and key-event definitions. It shows an important pattern within this dataset, not a standard expected result for AI referrals.

<h2>AEO Updates Takeaway</h2>

The most useful interpretation is not that AI traffic is replacing organic search or that AI visitors convert 50% better. It is that AI may be changing the point at which some people need a website.

The website is not disappearing. It may be becoming more concentrated around verification, service and transaction while answer engines absorb more education, comparison and consideration.

If that continues, marketers will need to ask more than how much traffic AI sent. They will need to understand where AI entered the decision, what it caused the consumer to believe, whether it moved the brand into consideration, whether it recommended the brand and which eventual business outcome that influence helped create.

Those questions are harder than counting clicks. They are also closer to measuring what AEO is supposed to accomplish.

<h3>References</h3>

[1] Digiday, “In Graphic Detail: How AI search has impacted the web traffic of over 50 advertisers,” August 25, 2026: https://digiday.com/marketing/in-graphic-detail-how-ai-search-has-impacted-the-web-traffic-of-over-50-advertisers/

[2] Brainlabs, “AI & Search: Ads, Zero-Click, and What to Measure Now,” January 12, 2026: https://www.brainlabsdigital.com/ai-search-zero-click-what-to-measure-now/

[3] Digiday, “The IAB is developing a framework to tackle AI advertising measurement,” August 24, 2026: https://digiday.com/media/the-iab-is-developing-a-framework-to-tackle-ai-advertising-measurement/

[4] Google Analytics Help, “About key events”: https://support.google.com/analytics/answer/9267568?hl=en