AI agents are changing web traffic. Here's how we're preparing.
AI agents are now browsing, comparing, and buying on behalf of real people, and they're showing up in our waiting rooms. Here's how we're helping our customers prepare for this new category of online traffic—from visibility, to authentication, to fairness and control.
Automated traffic now outnumbers human traffic on the internet. At Queue-it, we're seeing it firsthand. Every day, millions of bots try to enter our customers' waiting rooms to purchase tickets, sneakers, and limited-edition products they can hoard and resell to real fans.
Over the past year, though, we've started seeing a different kind of automated visitor in our waiting rooms: AI agents that browse, compare, and complete actions on behalf of real people.
In this post, I'll cover the important questions this new traffic category raises for organizations providing goods or services online, and the technology we're building at Queue-it to answer them.
For years, automated traffic basically meant bad bots. Scalpers, resellers, and fraudsters used automation to buy scarce items and sell them at inflated prices, at the expense of both the merchant and their customers. Building bots came with a high barrier to entry and, in many countries, meant the person behind the bot was breaking the law.
AI agents have changed that. Today, anyone can set up an agent that browses products, compares prices, and even completes a purchase for them. This agent doesn’t represent a scalper with malicious intent. It represents a real person with real intent to own a product, register for a service, or attend a concert. A person who simply wants help from AI.
If these agents are treated as bad bots and are blocked, merchants turn away real customers. But if they let every agent through, scalpers will simply start disguising their bots as agents. Neither option serves merchants or their customers.
Telling a human from a bad bot was already hard. But telling helpful agents from harmful ones is even harder, so companies need new ways to understand and control their traffic.
As agentic traffic grows in volume, there are three things every company or organization needs to ask themselves:
- How much of my traffic comes from AI agents, and what are they doing?
- How do I welcome agents without giving them an edge over humans?
- Which agents do I trust, and what should they be allowed to do?
Here’s how we’re approaching each of these questions at Queue-it: giving insights into agentic traffic, keeping the waiting room fair, and deciding which agents can do what.
You can’t set sensible rules for traffic you can’t see. That’s where our new Visitor MCP Server comes in.
MCP (Model Context Protocol) is an open standard that lets AI agents discover and use tools in a structured way. In practice, it’s a service that speaks the agents’ language.
The Visitor MCP Server serves a dual purpose. It enables agents to interact with waiting rooms in a predictable way, while also giving our customers insight into their agentic traffic.
Why do agents need an MCP server to join a waiting room? Partly because agents sometimes struggle with queues, and partly because they risk being mistaken for bots. Waiting your turn is a very human idea. An agent placed in a queue doesn’t necessarily understand why nothing is happening, so it may start firing off request after request to try get through.
By using our MCP server, an agent gets told where it is in line, how long it needs to wait, and what to do when its turn comes. That lets it behave predictably. It gets through the queue without wasting effort, the shopper behind it gets a fair shot at the product, and the retailer is spared a flood of repeat requests.
Routing agentic traffic through the Visitor MCP Server also makes it easier to tell agents from humans and bots. That’s the basis of our agent analytics, which shows what share of your visitors are agents and which providers they come from. And once you can see your agentic traffic, you can decide what to do about it.
The first version of our Visitor MCP Server is open: any agent can use it without validation. That’s deliberate. For now, it's about predictability and visibility, not gatekeeping.
Of course, an open door raises the obvious question: what stops a scalper from walking through it with thousands of bots disguised as agents?
That’s where authentication comes in. We are currently working on a version of the MCP server that can verify agents carrying a cryptographic signature or certificate. By tying agentic access to authentication, we can get closer to ensuring that the agent has been given a mandate by a real human.
We want to move to a place where every agent we let in has a single, verifiable person behind it: one person, one agent, one spot in line.
Several protocols are emerging in this space, and since none has become the standard yet, we won’t bet on just one. We’ll start with:
- Web Bot Auth, an open standard championed by Cloudflare
- TSAI, an open-source protocol initially developed by AWS
More protocols such as TAP by Visa will follow as the market matures.
We handle the verification, so our customers don’t have to. They just decide which agents get into the waiting room, and which are turned away.
RELATED: How to Trust an AI Agent with AWS's Alexander Günsche and Queue-it's Moji Sarooghi
Different organizations will want to handle agents differently. A retailer running a Black Friday sale might be glad to have them, while a public sector body running registrations with limited places might not be.
With Queue-it, organizations already control access for humans, good bots, and bad bots. Our goal is to extend that control to agentic traffic, helping customers to provide a fair and reliable process for every visitor—human or agent.
Agentic traffic isn’t limited to high-demand sales, though. That’s why we’re building a new product: a gateway that lets retailers and organizations control agentic traffic across their whole site, with or without a waiting room.
With our rules engine, you decide exactly what each type of agent is allowed to do. You could, for example, set up rules like these:
- Agents verified with a trusted protocol can go all the way to checkout
- Unverified agents can browse and compare prices, but not buy
- Agents that can’t show they act on behalf of a real shopper are blocked
To add another layer of validation to these rules, we’re also looking at the payment protocols emerging around agentic commerce, such as Mastercard Agent Pay, AP2 (by Google), and ACP (by OpenAI and Stripe). They let an agent prove that a real person authorized a purchase, so retailers could build that proof into Queue-it’s checkout rules.
Agents aren’t going away, and they won’t all look alike. Some will be a shopper’s helpful assistant, others a scalper in disguise, and every organization will need a way to tell them apart and decide what each is allowed to do.
That is a question of access, and access is what Queue-it has always been about: who gets in, in what order, and whether the process is fair. Extending that to agents is a natural next step. Visibility comes first, with authentication, and an agentic gateway to follow in the months ahead.
Whether a shopper visits a website themselves or sends an agent, the goal stays the same: a fair experience for everyone, on terms each organization sets.
We’re building this for our customers and alongside them. So if you want insights into your agentic traffic and want to help shape what comes next, ask your Customer Success Manager or get in touch with us at support@queue-it.com.