Itâs Tuesday, November 10th: Welcome to another edition of The Byte.
In this piece, Winnie Jin argues that the next big problem for the internet isnât figuring out whether something was made by AI. Itâs figuring out who made it, and whether they had permission to act. As AI-generated content spreads and bots account for a growing share of web traffic, the distinction between human and machine is becoming less useful. A trusted news organization using AI to summarize an article isn't necessarily a problem. Someone using AI to impersonate that organization is.
Winnie sees payments as the place where this identity problem gets solved first. Unlike misinformation, where the costs are difficult to assign, fraudulent purchases leave behind chargebacks, disputes, and clear financial losses. With McKinsey estimating that agents could mediate $3â5 trillion in commerce by 2030, companies like Visa, Mastercard, Stripe, and Google are racing to build the infrastructure that lets AI agents prove who they represent and what they're authorized to do.
That changes how we think about digital trust. Today's fraud systems largely try to distinguish humans from bots using signals like IP addresses, device fingerprints, and browsing behavior. But when a human deliberately sends a bot to make a purchase, those signals stop answering the important question. Winnie points to emerging frameworks that connect a verified identity to an agent's instructions and actions, creating an auditable record of what happened and why.
Her broader argument is that payments could do for identity what online shopping once did for encryption. SSL started by securing checkout pages before becoming standard across the web. Agent identity may follow the same path, with the financial incentives of commerce funding infrastructure that eventually extends to content, information, and the wider internet. The technology is taking shape, but who gets to establish trust, and how that trust travels across platforms, remains unresolved.
Identity and provenance are the missing foundations of the agentic internet
by Winnie Jin

The agentic internet cycle is well underway. With bot traffic surpassing human web traffic for the first time in early June 2026, new paradigms have arrived. Traditional search engine optimization is now being supplemented by âanswer engine optimizationâ, where brands optimize their marketing copy for inclusion in LLM answers rather than solely for ranking in search results. Over 10% of web pages sampled by Pew Research in August 2026 contained AI-generated text.
These developments fuel demand for detection tools like Pangram and provenance standards like C2PA that distinguish AI-generated content from human-generated content. But soon, the conversation is likely to shift from provenance to identity. As AI content becomes more pervasive, both humans and bots may care less about whether content is AI-generated and more about who generated it.Â
For example, if a trusted news source uses AI to generate article summaries, readers (or their AI agents) may still consider the output worth consuming. The unsolved issue is how cheap and easy itâs become to spoof that trusted source. Using AI, bad actors can spin up near-exact replicas of the sourceâs website, social media handles and deepfake videos of its anchors at a negligible cost.
The identity problem is not limited to content. Beyond the internetâs information economy lies its multi-trillion dollar economy powered by online payments, e-commerce and digital advertising. That dollar economy is where agent identity is now coming to a head as key industries build toward agentic commerce.
Why payments are the identity driver
While misinformation and fraudulent content may be costly for consumers, those costs are hard to quantify and are not clearly the responsibility of any single entity. Implementing an identity layer for content and agentic web visits is generally being handled in a low-assurance, binary way â âis this a human or a botâ â and applied unevenly across platforms and use cases.
In contrast, there are payments, where the cost of fraudulent transactions, chargebacks and disputes are clear-cut. With McKinsey estimating agents could mediate $3-5 trillion in agentic commerce transactions by 2030, both the potential upside (capturing agentic commerce revenue) and downside (fraud and dispute losses) are large enough that major infrastructure players are racing to build solutions.
Because agentic commerce requires significant coordination among merchants, banks, payment networks and processors, development began in 2025 primarily through open-source protocols and industry standards. Stripe and OpenAIâs Agentic Commerce Protocol, released last September, and Googleâs Universal Commerce Protocol, released this January, represent competing attempts to build coalitions of influential partners around their respective approaches.
A wave of product launches suggests the industry is moving quickly toward broader consumer adoption. Mastercard has completed live, authenticated Agent Pay transactions with banks and payment processors, and Visa announced live agentic transactions across Europe in July.Â
Combined with dedicated agentic commerce headcount and growing lists of partnerships with major merchants, these announcements suggest that infrastructure providers are positioning for a broader consumer rollout during the 2026 holiday shopping season.
Notably, the financial services industry is already well-acquainted with identity verification through Know Your Customer (KYC) compliance. How that identity integrates into agentic commerce workflows, however, remains an open question.
How AI agents break the current state of digital identity
Today, fraud prevention principles in e-commerce focus on evaluating signals like your logged-in account, IP address, behavior and device fingerprinting. Many signals function primarily to distinguish human traffic from bot traffic.
Now that bot traffic can be intentional and delegated by humans, that paradigm is no longer sufficient. Instead, the problem has moved to distinguishing whose bot is acting, and whether the bot is doing what its owner intended.Â
In agentic commerce, frameworks like Mastercardâs Verifiable Intent aim to solve this by creating a record that links the ownerâs ID, what they asked for and what the bot ultimately did. If someone makes an agentic purchase and disputes the transaction, that record can be audited to trace the complete chain of events from prompt to purchase.
The same principles apply more broadly with AI-generated content. Here, the C2PA standard enables auditing a record that contains the entire edit history of a piece of content, including camera capture, editing software and AI generation. On the identity side, the adjacent Creator Assertions Working Group aims to bind the creatorâs identity to the record.
Across content and commerce, the tech industryâs approaches to digital trust are converging.

Bridging to the new world
In this unfolding era of the agentic web, itâs become clear that agents and AI content must be tied to a verified human or organization. That identity then needs to be linked to a record with a specific authorization scope or provenance trail to produce a complete chain of trust.
The authorization and provenance puzzles have relatively complete proposed solutions, but the identity piece remains unsolved.
Several key players have entered this space. Cloudflare introduced agent wallets in August 2026. Visaâs Trusted Agent Protocol gives approved agents a cryptographic way to prove their identity and associated authorization to merchants. Y Combinatorâs Fall 2026 request for startups highlighted identity as a key component for rebuilding the internetâs trust layer.
There is not yet a clear winner, given the high bar for trust in an identity provider and the many difficult design choices around assurance levels, centralization, privacy and interoperability.
Payments first, what next?
The story of SSL offers a close analogue for the adoption of identity and provenance in the agentic era. Encryption was first implemented on checkout pages to secure online shopping, then spread across the wider internet.
The underlying principle is that e-commerceâs economic incentives are what produce enough pull for digital infrastructure to be developed before scaling more broadly.
LIke SSL, identity and authorization solutions for agentic commerce may also become the identity and provenance solutions for the open web. That is why it is so important to get them right â and why the next six to 12 months could represent a critical inflection point for the next phase of the internet.
The AI Collective is built by volunteers across 180+ chapters in 40 countries.
đ Find an Event
đˇ Join the Team
đ Start a Chapter
Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.
đ§âđť About the Author & the Editorial Team

About Winnie Jin
Winnie Jin is the creator of Trust Compass, where she explores digital identity, online trust, and the infrastructure shaping the agentic internet through writing and video. She brings 10 years of experience in digital media and currently works in go-to-market at Proof, an identity authorization platform. Her work makes the technologies and tradeoffs behind digital trust accessible to a wider audience.

About Josh Evans
Josh is a Managing Editor at The AI Collective Newsletter and leads content for The Byte. Outside of AIC, Josh works in Content Protection at Spotify.


