Sean Walsh
Sean is a Director at Intelligency heading up our digital marketing and client services operations. Sean has 15+ years experiencing working both in-house and agency with brands including Lloyds, Alstom, Hitachi, Lufthansa, Viaplay, DFDS Seaways and Mercedes-Benz.
There has been a gap at the centre of AI search marketing since it became a topic worth discussing. The question everyone wanted answered was simple: if a user sees your brand recommended in an AI-generated answer but does not click through, does it actually matter? Do they come back and search for you later? Does being visible in AI responses translate into anything commercially measurable?
Similarweb has just published the first study to attempt a direct answer. The research connects AI recommendation data with downstream web traffic behaviour using a consumer panel, rather than just counting referral clicks. The findings are significant and, for anyone thinking seriously about where visibility is actually coming from in 2026, they are worth understanding properly.
What the study found
Brands appearing in ChatGPT recommendations were 2.5 times more likely to receive a site visit within seven days than comparable brands that were not recommended. That multiplier held consistently across six different brand comparison scenarios in the dataset. It is the first time AI recommendation data has been tied to actual downstream web behaviour rather than just citation counts or referral URLs.
The mechanism matters. Of the downstream traffic generated after an AI recommendation, 55.9% arrived via branded search, meaning users who saw a brand mentioned in a ChatGPT response subsequently Googled that brand by name. They did not click a link in the AI answer. They went away, remembered the brand, and searched for it directly. The AI recommendation acted as a brand introduction that triggered traditional search behaviour, not a direct referral.
This is the finding that changes the measurement conversation. Most teams tracking AI search performance are measuring referral clicks from platforms like ChatGPT and Perplexity. Those clicks are real but they are a small fraction of the commercial effect. The larger effect, the one that drives more than half of the downstream visits, is invisible in standard analytics. It shows up as branded search, as direct traffic, or not at all.
The quality of AI-referred traffic is materially higher
The Similarweb data also quantifies what many practitioners had been observing anecdotally: that the visitors who do arrive directly from AI platforms engage more deeply than those arriving from traditional search. Users arriving from ChatGPT spent an average of 15 minutes on site, compared to 8 minutes for Google referrals. They generated 12 page views per visit against 9 from Google. They converted to transactional sites at a 7% rate, compared to 5% from Google referrals.
Volume is low. Intent is high. These are users who have already been through a filtering process inside an AI conversation before they arrive on your site. They are further along in their decision-making than the average organic search visitor, which explains both the deeper engagement and the higher conversion rates.
A separate piece of data reinforces this pattern from a different angle. On 7 May 2026, ChatGPT changed how it surfaces brand links, making brand names clickable callouts placed directly inside responses rather than citing sources in follow-up prompts. In the week after that change, total ChatGPT referral traffic increased by 157.7% across tracked websites. But homepage referrals surged by 354.7%. Before the update, roughly 26 to 32% of ChatGPT referrals landed on a brand homepage. After it, around 60% did. Users are arriving at the front of a brand, not a specific article or product page, which is consistent with AI operating as a brand discovery layer rather than a search engine.
Where AI sits in the purchase journey
The study also includes survey data on where in the purchase journey consumers find AI tools most useful. At the discovery and initial ideas stage, 35% said AI tools are most useful at that point, compared to 13.6% for traditional search engines. At the researching and comparing options stage, the split narrows: 30% for AI tools versus 20% for search. At the finding where to buy stage, the gap narrows further: 24.3% for AI tools versus 22.1% for search.
AI dominates the earliest stage of the funnel. This reinforces what the 2.5x traffic multiplier is actually measuring. AI is not replacing search. It is sitting upstream of it, shaping which brands users subsequently search for, visit, and consider. The case for investing in AI visibility is not that AI referral clicks will replace organic traffic. It is that AI shapes the pool of brands that end up in organic search at all.
The measurement problem this creates
The dominant metric in AI search reporting right now is referral click volume from platforms like ChatGPT and Perplexity. That number is being used to make decisions about whether AI visibility investment is working. Based on Similarweb’s data, that is the wrong metric. Referral volume understates the effect by ignoring the downstream branded search that the recommendation triggers. It also does not capture users who never click through at all but whose subsequent direct visits or brand searches were influenced by the AI response.
Some researchers estimate that up to 70% of AI-influenced traffic arrives in analytics as direct traffic or branded search with no attribution to the AI source. This is sometimes described as dark AI traffic. It is real, it is commercially significant, and it is currently invisible to most reporting stacks.
The practical implication for how to measure AI visibility is clear. Stop measuring it primarily by referral click volume. Start measuring:
- Brand mention share across AI platforms, meaning how frequently your brand appears in relevant AI-generated answers relative to competitors
- Branded search volume trends in Google Search Console, which may be rising as a downstream effect of AI mentions before any direct referral traffic appears
- Direct traffic trends, particularly homepage visits, which the Similarweb data suggests are increasingly a downstream signal of AI visibility rather than traditional direct navigation
- Sentiment context of AI mentions, since being mentioned in a negative context is measurably worse than not being mentioned at all
