The Maze: AI search is not just sending clicks. It may be shaping the next visit before the user ever reaches Google. Similarweb's June 2026 AI visibility study, shared by Tom Critchlow, compared six brand-pair scenarios across finance, travel, and beauty. In every case, the AI-recommended brand won the follow-on visit. Capital One reached 14.2% versus American Express at 3.8%. Kayak reached 12.0% versus Skyscanner at 3.4%. This is not SEO with a new hat. It is brand demand with a recommendation engine in front.
The recommended brand wins every visible match-up. Similarweb's report page identifies the study as work on AI visibility's downstream business impact. The source visual makes the commercial point sharper. American Express beats Capital One when recommended, 7.2% to 3.1%. Capital One beats American Express when recommended, 14.2% to 3.8%. Sephora beats Ulta when recommended, 7.9% to 3.3%. Ulta then beats Sephora when the answer flips, 7.6% to 4.6%. The assistant is not just a discovery surface. It is a demand nudge.
The largest gaps appear where the answer gives the user a shortlist. The biggest spread is Capital One over American Express, a 10.4-point gap. The second is Kayak over Skyscanner, an 8.6-point gap. Those are not direct AI referral numbers. They are follow-on visits within seven days among US desktop users from July to December 2025. That matters because Tom Critchlow's source post also says 55.9% of the traffic arrived through search. The user may ask AI, then validate through search, then visit the brand. The attribution system sees search. The buying journey saw AI first.
Clicks and citations are too narrow for this job. A citation tells you whether the brand appeared in an answer. A click tells you whether the AI platform passed traffic directly. Neither captures whether the answer changed the next query, the brand shortlist, or the user's confidence. The more useful measurement layer is recommendation quality: how the brand is described, which competitor it is compared against, whether the answer names it as the best fit, and whether search demand follows.
The caveat is methodology, not usefulness. The visible LinkedIn comments ask the right questions: why these three industries, what the baseline switching rate is without ChatGPT, and whether brand strength, user location, or sample size were controlled. Those gaps matter. Skyscanner's recommended-brand gap over Kayak is only 1.9 points, while Kayak's over Skyscanner is 8.6 points. That asymmetry could reflect AI influence, existing brand power, trip-planning intent, or all three. Operators should treat the result as a signal to measure, not a universal law to laminate.
For ecommerce teams, this moves AI visibility closer to brand and merchandising. If recommendations create downstream demand, the work is not only technical AEO. It is product data, reviews, category authority, comparison content, pricing clarity, availability, and trust signals that make an assistant comfortable recommending the brand. In old search, you fought for the blue link. In AI search, you fight to become the answer the buyer wants to verify.
Why it matters: Retailers and brands are still tempted to ask the easiest dashboard question: how many visits did AI send? That may miss the larger economic effect. AI can shape consideration, push users back into search, and make one brand feel like the safer next click. The winning team will not only track citations. It will measure recommended-brand share, sentiment, competitor pairings, branded search lift, and conversion quality after AI exposure. The click is becoming the receipt, not the whole transaction.
