AI-Recommended Brands Get 2.5x More Website Visits Than Competitors

For the past two years, the pitch for investing in AI visibility has rested almost entirely on a plausible story: if AI assistants are increasingly where people research decisions, then showing up favorably inside those answers should matter. It's a reasonable argument. It's also been, until now, largely unprovable with real behavioral data — measured mostly through citation counts, share-of-voice estimates, and the kind of soft logic that's easy for a CFO to wave away in a budget meeting.

That gap just got a lot harder to justify. Similarweb's new report, The Downstream Impact of AI Visibility, published in June 2026, is the first study to trace what actually happens after someone gets a brand recommendation from ChatGPT — not through survey recall or modeled estimates, but through real clickstream data from an opted-in US desktop panel. The headline finding: brands recommended by ChatGPT were 2.5 times more likely to receive a website visit within seven days than a competitor that wasn't recommended.

That's the number making the rounds in marketing circles right now, and for good reason — it's clean, quotable, and answers the question every founder and CMO eventually gets asked when they propose spending money on AI visibility: does any of this actually drive traffic? For the first time, there's a real, methodologically serious answer.

This piece walks through what Similarweb actually measured, why the study design matters as much as the headline number, what the secondary findings reveal that's arguably more useful than the 2.5x figure itself, what this means industry by industry, and what businesses — especially the people who actually approve GEO budgets — should take from all of it.

What Similarweb Actually Measured

The setup. Similarweb tracked real user journeys across six months of US desktop browsing data — July through December 2025 — supplemented by a January 2026 survey, all drawn from an opted-in panel rather than modeled traffic estimates. The study followed people who asked ChatGPT an industry-relevant question, received a specific brand recommendation in response, and then watched what happened over the following seven days: did they visit the recommended brand's site, a competitor's site, neither, or both?

Crucially, the study excluded anyone who had already visited the recommended brand's site in the prior four weeks, and anyone who'd named the brand directly in their own prompt. That's a meaningful methodological choice — it's specifically designed to isolate the effect of the AI's recommendation itself, rather than just confirming that people who already knew a brand kept visiting it. Similarweb was trying to measure genuine influence, not confirmation bias dressed up as a finding.

The categories. The research covered three verticals chosen for their consumer research intensity — finance, travel, and beauty — with head-to-head brand pairs in each: American Express versus Capital One in finance, Skyscanner versus Kayak in travel, and Sephora versus Ulta in beauty. In the finance category specifically, 7.2% of users who got an Amex recommendation visited its site afterward, compared to 3.1% for Capital One in the same conditions — a gap that shows up consistently across all three verticals, not just as an average smoothing over noisy underlying data.

The headline result. Across all three categories, users who received an AI recommendation were, on average, 2.5 times more likely to visit that brand's site within the following week than they were to visit a non-recommended competitor's site under otherwise similar conditions. That's not a projection or a survey-based estimate of stated intent — it's actual observed browsing behavior from a real panel of real people doing real research.

Why the methodology matters as much as the number. It's worth pausing on why this study is being taken more seriously than the usual run of AI-visibility claims circulating on LinkedIn. Most of what passes for "AI search data" in the current environment is either a single practitioner's anecdotal experiment, a vendor's self-reported case study, or a survey asking people to recall and self-report their own behavior — all of which carry real bias risks. Similarweb's panel-based clickstream approach sidesteps most of that: it's observing what people actually did, not what they say they did or what one company's own customers experienced. That's a meaningfully higher evidentiary bar, and it's a large part of why this particular study has traveled so far, so fast, through the SEO and marketing press.

The Data Most Coverage Is Skipping Past

The 2.5x figure is the headline because it's the easiest number to repeat in a slide deck. But two secondary findings in Similarweb's report arguably matter more for how businesses should actually act on this.

Most of the resulting traffic is invisible to standard analytics. Of the visits that followed an AI recommendation, 55.9% arrived via branded search — meaning the person didn't click straight from ChatGPT to the site, they went and searched for the brand by name afterward. Compare that to a 40.4% branded-search share among the general visitor population, with direct traffic dropping correspondingly (19.9% for AI-influenced visitors versus 38.8% for everyone else). In plain terms: a huge share of the traffic an AI recommendation generates shows up in a company's dashboard looking exactly like an ordinary organic branded search — with zero indication that a ChatGPT conversation happened first. If your analytics team is only crediting AI when a click comes directly from an AI platform's referral, you're almost certainly undercounting the real effect by a wide margin. The AI recommendation is doing real work; your attribution model just isn't built to see it.

AI-influenced visitors engage substantially harder once they land. Visitors who arrived after an AI recommendation viewed an average of 12.0 pages per session and spent 11.8 minutes on site, compared to 6.5 pages and 5.6 minutes for visitors who weren't AI-influenced. That's roughly double the engagement on both dimensions. This lines up with the more intuitive read of what's actually happening: someone who's had a real conversation with an AI assistant, gotten a specific, contextualized recommendation, and then gone looking for that brand by name is arriving considerably more informed and more intentional than someone who clicked a generic search result. It's a warmer visitor, not just a more frequent one — and for a lot of businesses, that distinction matters more to the bottom line than the raw visit-count multiplier does.

Worth pairing with a caution from adjacent research. Similarweb's report draws on analysis from Rand Fishkin at SparkToro, who separately found that AI tools can return meaningfully different brand recommendations across repeated versions of the same query. Put the two findings together and a more complete picture emerges: AI recommendations genuinely move real traffic and produce more engaged visitors — but which brand gets recommended in the first place can be unstable, shifting from one query run to the next in ways that make AI visibility a genuinely harder thing to plan around than a stable organic ranking. The upside is real. So is the volatility underneath it. Neither finding cancels the other out — together they describe a channel that's both more valuable and less predictable than classic SEO, which has real implications for how a business should budget for it.

Why this specific framing — recommendation, not just mention — matters. It's worth being precise about what Similarweb actually tested, because the distinction between being mentioned and being recommended is doing a lot of work here. The study isolated cases where ChatGPT gave users a specific brand recommendation in response to an industry-relevant question — not simply a passing reference to a brand's name buried in a longer answer. That's a meaningfully higher bar than raw "share of voice" or citation-count tracking, which often treats any mention as equivalent regardless of how prominently or favorably a brand actually appears. The businesses that benefit most from this data are the ones being positioned as a genuine answer to a question, not just one name among several listed in passing.

What This Looks Like by Industry

The three categories Similarweb studied — finance, travel, and beauty — share a common thread worth naming explicitly: they're all categories where people research before they commit. Nobody picks a credit card, a flight booking platform, or a skincare routine on pure impulse the way they might grab a snack at a checkout counter. That's exactly the kind of decision an AI assistant is well-suited to help with, and exactly the kind of decision where a favorable recommendation has real room to shift behavior.

That pattern generalizes well beyond the three verticals actually tested. Any category where customers ask comparative questions before buying — home services, B2B software, healthcare providers, financial advisors, higher education, specialty retail — plausibly sees a similar dynamic, even without its own dedicated Similarweb study yet. The common denominator isn't the industry; it's whether the purchase involves genuine research and comparison rather than habitual or impulse buying. Businesses in low-consideration categories shouldn't assume this data translates directly to their situation. Businesses in genuinely research-driven categories have real reason to treat this as directly relevant, even without a study of their exact sector.

Finance. The Amex-versus-Capital-One comparison is a useful illustration of how AI recommendation dynamics play out in a category already crowded with comparison content, review sites, and affiliate marketing. Even in a space this saturated with existing SEO-optimized comparison content, the AI recommendation still produced a more than two-fold visit gap — suggesting that whatever AI systems are weighing when they recommend one card issuer over another isn't simply reflecting whichever site has the most backlinks or the highest domain authority in classic SEO terms.

Travel. Skyscanner versus Kayak sits in a category defined by genuinely high query volume and habitual repeat use — people book travel repeatedly, and brand loyalty in this space has traditionally been thin, since price and convenience usually win over brand affinity. That makes the AI-recommendation effect in this category particularly interesting: if AI recommendations can meaningfully shift visit behavior even in a price-driven, low-loyalty category, it suggests the effect isn't just about existing brand strength winning out — it's about the recommendation itself carrying real independent weight with the user in the moment.

Beauty. Sephora versus Ulta represents a category where personal recommendation and trust have always mattered more than raw price comparison — beauty purchases are notoriously influenced by word-of-mouth, reviews, and perceived authenticity. An AI assistant standing in for that trusted-recommendation role is a natural fit, and the data bears that out.

It's also worth being honest about what the study didn't test. It looked at established, well-known brand pairs going head-to-head, not smaller or less recognized businesses competing against category leaders. Whether the same 2.5x effect holds for a regional business or a newer entrant competing against an established name is a reasonable open question. The mechanism — AI recommendation shaping subsequent search and visit behavior — has no obvious reason to be exclusive to large brands, but the magnitude could plausibly differ, and businesses should treat the 2.5x figure as directionally meaningful rather than as a guaranteed multiplier for their own specific situation.

What Businesses Should Actually Do With This

Start tracking branded search as an AI-visibility proxy, not just a vanity metric. Given that more than half of AI-influenced traffic shows up as branded search rather than a direct platform referral, a business that isn't already watching branded search trends closely is missing one of the few available signals for whether AI recommendations are actually happening in the background. This doesn't require new tooling — most organizations already have branded search data sitting in existing analytics and search console tools. What changes is the interpretation: a rise in branded search that doesn't correspond to a traditional marketing push is now a reasonable signal worth investigating for an AI-mediated cause.

Test how your brand actually shows up, repeatedly, not once. Given the variability that SparkToro's research points to, a single favorable test query tells you very little. The businesses that get real signal here are running the same category of question across multiple phrasings, multiple times, over weeks — treating AI visibility testing as an ongoing practice rather than a one-time check performed before a board meeting and then forgotten about.

Treat engagement quality as a metric worth reporting alongside traffic volume. If AI-influenced visitors genuinely engage roughly twice as much once they land, that's a number worth surfacing in its own right when making the case for AI visibility investment — not just as a footnote to the traffic multiplier, but as evidence that the visitors this channel produces are qualitatively different, and arguably more valuable, than an average visitor.

Use this data point as an anchor, not a complete case. The 2.5x figure is a genuinely strong opening argument in any internal conversation about funding AI visibility work — but it's an opening argument, not a finished business case. Pairing it with your own branded search trends, your own AI citation testing, and a realistic read of how research-driven your specific category actually is will make for a far more convincing pitch than the stat alone.

The GEO SEO Lab AI Citation ROI Model

Put simply, the chain this data supports looks like: AI visibility → AI recommendation → branded search (often invisible to standard attribution) → website visit → deeper engagement than average → business outcome. Every link in that chain now has at least preliminary empirical support behind it, rather than resting entirely on inference. The practical implication is that measuring AI visibility work purely by direct AI-platform referral traffic will substantially understate its real effect — branded search growth and session engagement quality need to sit alongside AI citation tracking as core measurement inputs, not afterthoughts.

The AI Visibility Investment Case

For a decision-maker weighing whether to fund GEO work, the case this data supports breaks down into three components worth stating plainly. First, the traffic case: a favorable AI recommendation is now shown, in real behavioral data, to produce meaningfully more site visits than not being recommended at all — not a hypothetical, an observed effect. Second, the quality case: those visits aren't just more numerous, they're measurably more engaged, which matters more to revenue in most business models than raw visit count alone. Third, the measurement case: because so much of this effect hides inside branded search rather than direct AI referral, a business that isn't tracking branded search trends and running systematic AI visibility checks is very likely underestimating how much this already matters to its own numbers today, independent of any new investment at all.

The Attribution Problem This Study Actually Exposes

It's worth dwelling on the attribution issue a little longer, because it's arguably the most actionable finding in the entire report, even though it's not the one getting quoted in headlines.

Marketing attribution has always struggled with multi-touch journeys — the classic example being a customer who sees a display ad, later clicks a paid search result, and eventually converts through a branded search weeks afterward, with most attribution models crediting only the last touch. AI recommendations introduce a new version of the same old problem, except with a twist: the "touch" in question happens entirely outside any system a business controls or can directly measure. There's no pixel to fire, no UTM parameter to append, no referral header to capture, when a user has a private conversation with ChatGPT and forms an opinion about which brand to check out.

That means the 55.9% branded-search figure isn't just a data point — it's a warning about how much AI-driven demand is likely already sitting inside existing analytics, misclassified as organic or branded search performance with no acknowledgment of the AI conversation that actually triggered it. A marketing team celebrating a quarter-over-quarter lift in branded search, attributing it to a recent PR push or a brand campaign, might be looking at a number that's substantially explained by AI recommendations they have no visibility into at all. Conversely, a team that sees branded search stagnate might be missing a warning sign that their AI visibility is declining, mistaking it for an unrelated dip in traditional marketing performance.

The practical response isn't to abandon existing attribution models — it's to add a new question to the standard marketing reporting cycle: has anything changed in how AI platforms are representing our brand recently, and could that explain movement in branded search that doesn't line up with our traditional marketing calendar? That's a genuinely new muscle for most marketing teams to build, and it's one that becomes more valuable exactly as AI-mediated research keeps growing as a share of the overall customer journey.

Why This Study Changes the Conversation

Every emerging channel goes through a phase where its advocates are arguing from conviction and its skeptics are arguing from the absence of proof. AI visibility has been stuck in exactly that phase for the better part of two years — genuinely compelling logic on one side, genuine uncertainty about hard numbers on the other. This study doesn't end that debate entirely, but it does something more useful than ending it: it gives both sides a real, shared, methodologically serious data point to argue from instead of dueling anecdotes.

That matters practically, not just rhetorically. A founder trying to greenlight GEO spend no longer has to rely purely on directional argument about where research behavior is heading. A skeptical CFO no longer gets to dismiss the whole category as unmeasurable hype. The conversation can move to the more useful question this data actually opens up: given that AI recommendations demonstrably move real traffic and produce more engaged visitors, what's the right level of investment for a business in our specific category, and how do we measure whether it's working? That's a fundamentally better conversation than the one the industry has been having until now, and it's the direct result of one study finally treating AI visibility as something worth measuring with the same rigor as any other marketing channel.

Key Takeaways

• Similarweb's Downstream Impact of AI Visibility study found brands recommended by ChatGPT were 2.5x more likely to get a site visit within 7 days than a non-recommended competitor — based on real clickstream data, not survey estimates.
• 55.9% of that AI-influenced traffic arrived via branded search rather than a direct AI-platform click, meaning most of the effect is currently invisible in standard analytics.
• AI-influenced visitors viewed roughly double the pages and spent roughly double the time on site compared to other visitors — a genuinely more engaged visitor, not just a more frequent one.
• The effect held across finance, travel, and beauty — three categories that share high research intensity before purchase, suggesting the dynamic likely generalizes to other genuinely comparison-driven categories.
• Separate research from SparkToro shows AI brand recommendations can shift across repeated identical queries, meaning consistent, sustained AI visibility work matters more than a single favorable citation.
• This is the strongest available real-world evidence connecting AI citation to measurable business outcomes, and it changes what a budget conversation about GEO investment can actually point to.

About GEO SEO Lab

GEO SEO Lab is a research and strategy group focused on helping businesses understand and improve visibility across AI-assisted search and discovery — Google Search, Google AI Mode, ChatGPT, Gemini, Claude, Perplexity, Grok, and the broader ecosystem reshaping how people research and decide. Our work spans Generative Engine Optimization, AI visibility strategy, entity optimization, and measurement frameworks built to connect AI citation to real business outcomes — not just visibility for its own sake.

AI citations are now shown to drive up to 2.5x more traffic. See where your brand actually stands — get your free AI visibility score at GEOSEOLab

References and Further Reading

• Similarweb. (2026). The Downstream Impact of AI Visibility. similarweb.com/blog
• Search Engine Journal. (2026). AI-Recommended Brands Saw 2.5x More Site Visits: Similarweb. searchenginejournal.com
• Search Engine Land. (2026). ChatGPT Recommendations Drive More Brand Website Visits: Study. searchengineland.com
• PPC Land. (2026). Your Analytics Are Lying: Similarweb Traces AI Recommendations to Real Traffic. ppc.land
• SparkToro. (2026). Research on AI Recommendation Variability Across Repeated Queries. sparktoro.com