Most Americans use AI agents weekly, but half won't let them pay

Most Americans now use AI agents, but many stop short of letting them spend money. A new survey found 68% of US adults use an AI agent at least once a week for a personal task. Yet 50% are uncomfortable with an agent using their card details to complete a purchase. An AI agent is software that carries out a task for you, such as buying something you asked for.

What the survey found

SEON, a company that provides fraud prevention and anti-money laundering compliance, surveyed 2,026 US adults online in October 2026. Results were weighted to match the US population, with a credibility interval of plus or minus 3 percentage points. Discomfort with an agent using card details differs by group:

• Baby Boomers: 65%
• Gen X: 52%
• Gen Z: 47%
• Millennials: 34%
• Women: 56%, men: 43%

What people will share

Willingness to share data depends on the type. Location (38%), purchase history (34%) and calendar access (34%) are the most accepted. Only 15% would share financial information. Younger adults are more open, but fewer than one in four would share financial details: 23% of Millennials and 22% of Gen Z, versus 4% of Boomers. Also, 32% of all respondents would not share any information with an agent.

Who gets the blame

Respondents were asked who should be responsible if an agent completes a fraudulent transaction. Among the 885 who gave an opinion, 65% named the company behind the agent. Only 14% said themselves, 11% said the merchant and 10% said their bank or card provider. The top fraud worries, at 20% each, were a scammer gaining access to the agent or the user's accounts, and the agent sharing personal or financial information without permission.

Confidence in spotting trouble is low. About 43% said they lack confidence judging whether an agent's request is legitimate, and 39% do not trust agents to recognize suspicious situations for them. Among Boomers, 61% said they are hardly or not at all confident, compared with 28% of Millennials.

What would build trust

Human support when something goes wrong was the top answer at 50%. The ability to reverse mistakes followed at 42%, then strong identity verification at 39%. Another 23% said nothing would increase their trust.

What it means for businesses

SEON CEO and co-founder Tamas Kadar said people accept agents for everyday tasks, but trust drops fast when money and identity are involved. He said businesses need to tell a helpful agent from a harmful one before money moves. According to SEON, spotting that an agent is involved is only the first step. Companies also need to check identity, device and behavior signals to learn who is behind it and whether it fits the customer's permissions. SEON sells fraud tools, so keep that in mind when reading its findings.