AI referrals surge as web traffic falls & costs rise

Contentsquare has released new benchmark figures pointing to a rapid rise in AI-driven referrals, even as overall web traffic declines and the cost of attracting visitors increases.

The 2026 Digital Experience Benchmark Report analyses 99 billion online sessions across 6,500 websites worldwide. Visits referred by AI tools rose 632% year on year, though they still accounted for only a small fraction of total traffic in the final quarter of the period analysed.

The findings come as brands reassess how digital journeys begin. The benchmarks suggest discovery is spreading across more sources, customers are spending less time on websites, and early interactions now carry more weight.

AI referrals

AI-referred sources entered the traffic mix during the year and made up 0.2% of total traffic in Q4. Contentsquare links the shift to broader changes in search, where fewer clicks reach brand sites.

Organic search traffic fell 9% year on year. The report describes this as a structural change, with AI-generated summaries answering queries without the same level of click-through to publisher and retailer pages.

Visitors arriving via AI tools also behaved differently. AI-referred conversion rates rose 55% year on year to 1.3%, a level the analysis frames as closer to higher-intent sources such as traditional search.

Bounce rates improved across several acquisition channels. AI-referred visits bounced 5% less than a year earlier, while paid search bounce rates improved by 3% and organic search by 4%.

Rising costs

The figures show pressure on both traffic volumes and acquisition economics. Overall traffic fell 4% year on year. Over three years, cost per visit rose 30%, including a 9% year-on-year increase.

The benchmarks depict a market in which each session is more expensive to secure and sites have less room for error once customers arrive.

Visitors spent 7% less time on sites during the period analysed. The report attributes this to shorter attention windows and lower tolerance for friction in browsing and checkout flows.

Loyalty signals

Loyalty patterns diverged between new and returning audiences. Returning visitors declined by 2%, while new visitors fell by 6%, suggesting repeat behaviour is holding up better than first-time discovery in a more fragmented acquisition environment.

Retail showed a notable shift in entry points. More than 36% of visits started on a product detail page, and those pages accounted for 40% of page views.

Product pages also had the highest bounce rate in retail. With less time spent on site overall, the report argues the first page a shopper sees is increasingly decisive in whether they continue browsing.

Service interactions

Beyond browsing, the benchmarks examine customer service conversations and sentiment. The dataset includes 22 million customer service interactions.

Humans working alongside AI bots resolved 57% of customer inquiries, compared with 29% using bot-only support. Many conversations began with negative sentiment, particularly in email, where 64% started negatively.

In the dataset, sentiment shifts correlated with outcomes. Conversations that improved sentiment resolved more than twice as often as those that did not (67% versus 28%).

"Brands now operate in a world of dual audiences - humans and AI agents. That changes the rules. AI traffic is small today, but its intent signal is already outsized," said Jean-Christophe Pitié, chief marketing and partnerships officer at Contentsquare.

Benchmark scope

The benchmarks also cover 500 billion page views and compare Q4 2024 with Q4 2025. Conversion benchmarks include only sites that collect conversion events in manufacturing, retail, telecommunications, and travel and hospitality.

Contentsquare positions the findings as evidence that digital performance increasingly depends on quickly identifying friction points and focusing on what happens early in a session, as referral sources diversify and attention becomes harder to buy.

"Success today is defined by how quickly brands can detect experience gaps, friction, and mismatches - and fix them before customers even notice. Great customer experience isn't loud. It's precise. It's relentless. And it's built by removing small irritants until the experience simply works," said Pitié.