Visa's agentic commerce survey compares a technology category with a named payment brand. The result is a useful trust signal, but it is neither an adoption rate nor a causal finding.
Visa’s Trust Index for agentic commerce puts two numbers side by side. Twenty-three percent of US consumers said they would trust generative AI to handle a payment transaction on their behalf. Sixty-one percent said they would trust Visa to handle an agentic transaction.1
The arithmetic produces a 38-point spread. The research does not show that adding Visa causes a 38-point increase in trust. The two figures come from differently framed questions: one names a broad technology category, the other names a familiar payment company. That distinction changes what the index can support.
What the survey measured
Visa commissioned The Harris Poll to field the survey through its Omnibus platform from 26 to 28 May 2026. The sample was 2,065 US adults, matched to the US general adult population using Census data. Each brand-specific payment question went to between 1,028 and 1,034 respondents.1
The release also says AI assistant use was common and that trust in Visa was higher among younger adults and frequent AI users.1
Those results support a modest conclusion: respondents expressed more trust in a named payment brand than in generative AI as a category. They do not isolate why. Brand familiarity, perceived recourse, fraud protection, payment experience and question wording may all contribute. Visa did not publish the questionnaire, response options, subgroup sample sizes or uncertainty estimates in its release, so stronger causal readings are not available from the public methodology.1
Visa commissioned the study and ranked first in the brand item it reported. That does not invalidate the responses. It does mean the 61 percent figure should always travel with the sponsor and wording attached.
| The release supports | The release does not support |
|---|---|
| Respondents reported higher trust in Visa than in generative AI under the questions asked | Visa caused the full 38-point difference |
| AI assistant use was common in this sample | Most respondents had completed an agent payment |
| Familiar payment brands may help consumers feel comfortable | Consumers trust every merchant or counterparty selected by an agent |
Trust is not adoption
The index measures stated attitudes. It does not say that 61 percent of US adults have used Visa for an agent-initiated payment, or that 23 percent have allowed generative AI to pay. The 72 percent figure is prior use of an AI assistant, a much broader behavior than delegating a transaction.1
Timing adds another boundary. Fieldwork ended on 28 May and Visa published the results on 9 September, 104 days later. The release is a snapshot of late-May sentiment, not a live measure of September behavior.
The survey is useful for a narrower job: it shows that recognizable institutions may carry trust into an unfamiliar transaction model. Product teams still need behavioral evidence, such as completed checkouts, intervention rates, disputes and repeat use, before calling that willingness adoption.
Visa verifies more than the payment credential
Visa’s technical specification draws a more precise line between payments and identity than the survey does.
Visa’s Trusted Agent Protocol lets a merchant verify a signed request from an approved agent. Its trust model covers an agent recognition signature, a linked consumer or device identity and a linked payment container. Signatures are bound to a merchant, purpose and time window. Visa says the mechanism lets a merchant check that an agent is legitimate, that it is acting for an authenticated user and that it has permission for the specific browsing or payment action.23 For a wider view of the surrounding standards, see the agent payments stack.
Availability is not universal. Visa labels the product as being in development and deployment, and says the published flows may represent potential features rather than functionality available in every market.2
That is identity and authorization within a payment scheme. Successful verification proves participation in the scheme and validates signed claims for that interaction. It does not automatically provide a portable public profile, a history of professional work or a human-readable reputation across unrelated networks. This boundary is our interpretation of the protocol, not a finding from Visa’s survey.
| Question | Visa’s protocol can answer | Additional due diligence may ask |
|---|---|---|
| Who sent this request? | Whether it came from an approved, signing agent | Which person or company operates the agent |
| May it take this action? | Whether the signed intent covers this merchant and operation | Whether the relationship should continue after checkout |
| Who is the consumer? | A signed or obfuscated identifier, when provided | What either party should disclose to the other |
| Is the payment valid? | Signed payment data and scheme controls | Whether the counterparty is suitable for a professional engagement |
The missing measure is relationship trust
Visa addresses transaction authorization and agent recognition directly through payment controls and Trusted Agent Protocol.2 The Trust Index mainly measures consumer confidence in the institution providing those controls. It does not test relationship trust.
That question matters when an agent sources a supplier, screens a candidate, finds a collaborator or arranges an introduction. A valid signature can establish that a request is authorized. The recipient may still need to inspect the person or company behind it, decide what to disclose and carry reputation into later conversations. Agentic commerce for services explains why that distinction grows once the deliverable is a relationship rather than a product.
Where Tobira fits
Tobira focuses on that relationship surface. An agent receives a human-readable @handle, a W3C DID, WebFinger discovery and an A2A-compatible Agent Card. Identity is revealed only after both sides agree, and the resulting introduction moves to the humans behind the agents. The distinction between an agent name and a wallet address is useful here.
This complements payment authorization. Visa can help a merchant verify that a payment agent and its request are legitimate inside the commerce flow. Tobira gives agents an address for finding each other and a consent step before the people behind them are introduced.
The Visa research is strongest when read within its limits. Consumers told a Visa-sponsored poll that they trusted Visa more than generative AI to handle a payment. The public evidence does not show why, how that preference changes behavior or whether respondents trust the counterparty their agent selects. Those are the next measurements agentic commerce needs.
FAQ
Can the Visa Trust Index be cited as an adoption rate?
No. It measures stated trust, not completed agent payments. Use behavioral metrics such as checkout completion, intervention, dispute and repeat-use rates to evaluate adoption.
Can the reported gap be attributed to the Visa brand?
Not from the published methodology. The survey compares differently framed subjects rather than reporting a controlled brand experiment.
What does Trusted Agent Protocol verify?
It is designed to verify an approved agent, its intent for a specific interaction and its association with an authenticated user. It can also carry signed consumer and payment identifiers.23
What should a merchant verify beyond payment authorization?
The answer depends on the transaction. A merchant may also need to assess the operator behind the agent, disclosure consent, suitability for an ongoing relationship and reputation outside the payment scheme.
Sources
Footnotes
-
Visa, “New Visa Research Finds Consumer Trust is Accelerating the Path to Agentic Commerce,” 9 September 2026. Includes the reported results and methodology. Visa release and methodology ↩ ↩2 ↩3 ↩4 ↩5
-
Visa Developer Center, “Trusted Agent Protocol,” accessed 17 September 2026. Describes agent identity, transaction-specific authorization, user association and the product’s development and deployment status. Trusted Agent Protocol overview ↩ ↩2 ↩3 ↩4
-
Visa Developer Center, “Trusted Agent Protocol Specifications,” accessed 17 September 2026. Defines the three-signature trust model and verification steps. Trusted Agent Protocol specification ↩ ↩2