(Reuters) -U.S. online sales growth during the 2025 holiday season is expected to slow from last year, as price-conscious shoppers stay picky with their spending amid rising living costs, according to a Salesforce forecast released on Tuesday.

Salesforce projects online spending between November 1 and December 31 to rise 2.1% to $288 billion, lower than a 4% increase to $282 billion in the same period last year.

Retailers, meanwhile, are expected to be more cautious with promotions. The number of orders using promotional codes is forecast to dip slightly, as brands grapple with higher supply chain costs and become more selective with discounts.

The outlook echoes recent reports from Deloitte and PwC, underscoring a subdued holiday season that will see shoppers prioritize essentials, hunt for deeper discounts and cut back on discretionary purchases as economic uncertainty weighs on sentiment.

In recent weeks, major retailers have issued mixed forecasts heading into the crucial holiday season. While Walmart and Macy's have raised their outlook, toymaker Mattel cut its forecast. Target maintained its annual expectations.

"One of the things that we are ... potentially concerned about is if more consumers get more surprises from import fees than they do now from carriers, that could potentially have an impact on e-commerce," said Caila Schwartz, director of Strategy and Consumer Insights at Salesforce.

Average discount rates in the U.S. during Cyber Week - the five-day stretch from Thanksgiving through Cyber Monday - are expected to be 29%, with top deals in categories such as general apparel, health & beauty and home furnishing, Salesforce said.

Artificial intelligence-powered recommendations and agent-assisted shopping are expected to boost purchases, with Salesforce estimating that these technologies will drive $51 billion in U.S. online sales, 18% of overall projected sales.

Salesforce analyzed data from more than 1.5 billion global shoppers across 89 countries using its cloud platforms, while also blending research from consumer sentiment surveys of 5,500 people to generate its forecasts, it said.