AI agents as intermediaries: the end of the direct website visitor?

Picture this: a customer wants to buy accounting software. Instead of visiting websites, reading comparison articles and requesting demos, they give their AI assistant a task: “Find accounting software for my one-person business, max $30 a month, that connects to my bank and handles quarterly estimated taxes.”

The AI agent searches, compares, selects and reports back: “I recommend [product X]. It costs $25 a month, connects to Chase and Bank of America, and prepares your quarterly estimates. Shall I create a trial account?”

The customer did not visit a single website. The entire customer journey happened through the AI agent. This is not science fiction; it is the direction the technology is moving in.

What are AI agents?

AI agents are AI systems that carry out tasks on their own on behalf of users. They go further than chatbots that answer questions:

• They can carry out several steps (search, compare, select, act)
• They learn from your preferences and context
• They interact with external systems (book, buy, reserve)
• They report back with a recommendation or result

Examples that exist or are in development:

• Shopping and checkout inside ChatGPT, Gemini, Copilot and Perplexity for US buyers
• Google’s Project Astra (multimodal AI assistant)
• Apple Intelligence (built into iOS)
• Microsoft Copilot in Windows and Office

A fuller explanation with examples for small businesses is in what is an AI agent.

The impact on website traffic

The direct effect

When AI agents search and act on behalf of users, direct website traffic drops. The agent may visit your site (through an API or crawler), but the user does not. That changes basic assumptions:

• Pageviews drop, but conversions do not have to
• Session duration becomes irrelevant, the agent reads in milliseconds
• Bounce rate means nothing, there is no “visit”
• User experience shifts, from human visitors to machine visitors

The indirect effect

At the same time, AI agents create a new type of traffic:

• Branded search rises: the agent names you, the user googles you
• High-quality direct traffic: the visitors who do come have been “approved” by an AI
• API traffic grows: agents interact through your API, not your front end

The impact on conversion

The zero-visit conversion

In the agent world, a conversion can happen without a website visit:

1. The agent finds your product through structured data
2. The agent compares you with competitors
3. The agent recommends you
4. The user approves
5. The agent places the order through an API or checkout integration

Your attribution model has to adapt. The traditional funnel (visit, page view, cart, checkout) is replaced by an agent funnel (fetch data, compare, recommend, transact).

Conversion optimization for agents

Instead of optimizing your website for human visitors, you optimize for machine readability:

• Complete Product schema: price, availability, reviews, specs
• An API, so agents can talk to your system directly
• Offer objects: structured offers agents can process
• Real-time data: prices and availability must be live

Read more in agentic commerce for Shopify, WooCommerce and Magento.

The impact on your brand

The brand as a dataset

In a world of AI agents, your brand is reduced to a dataset: price, features, reviews, availability. The visual brand experience, your logo, your color scheme, your tone of voice, is invisible to an agent.

So brand differentiation shifts:

• From visual identity to data identity (how complete and accurate is your data?)
• From website experience to recommendation experience (how positively does AI talk about your brand?)
• From emotional connection to trust signals (reviews, authority, certifications)

Keep the human layer

Despite the agent layer, human contact stays relevant. The moments when a customer does experience your brand directly become rarer but more valuable:

• The first personal contact after an AI recommendation
• The onboarding experience
• Customer service when there is a problem
• The community around your brand

Invest in these touchpoints. They become your most important brand differentiator.

How to prepare

1. Make your data machine-actionable

The foundation: your product data, prices, availability and specs must be structured and machine-readable. That is the “website” agents visit.

2. Build API readiness

Agents communicate through APIs, not websites. Start by exposing your core data through an API, even a simple read-only one.

3. Invest in reviews and authority

When the visual brand experience falls away, reviews and authority become your primary trust signals. Build both actively.

4. Monitor AI recommendations

Do you know whether AI agents recommend your product? Test relevant prompts regularly and monitor your AI visibility.

5. Redesign your metrics

Traditional web analytics measure the wrong things. Develop new metrics:

• AI citation frequency
• Agent interaction volume (from API logs)
• Branded search as a proxy
• Post-purchase attribution (“how did you find us?”)

The timeline

• 2026: agentic commerce in an early phase. Checkout in AI assistants for US buyers, Perplexity’s merchant program, the first transactions through agents.
• 2027: mainstream adoption among early adopters. AI agents built into operating systems and browsers.
• 2028-2030: agents become a standard intermediary for a significant share of online transactions.

No panic, but action

The end of the direct website visitor is not an apocalypse. It is a shift, comparable to the shift from desktop to mobile. Websites did not disappear; they were optimized for a new device.

The same will happen with agents. Websites will not disappear; they will be optimized for a new intermediary. The businesses that start that optimization now will be ready when agents go mainstream.

See the future of search for the full picture.