The 5 Pillars Every Merchant Needs

Agentic commerce was this pie-in-the-sky concept, something that merchants would have to prepare for at some vague future date. Until it wasn’t.

Muse orders were first detected on the Forter network on September 21, 2026. Agentic orders are now 5.3x their level during the first week of August 2026. This sudden surge happened just in time for our annual IMPACT New York conference, where agentic readiness was the hottest conversation topic among commerce leaders.

Read on to learn the five pillars of agentic readiness, which are all connected by adaptability.

1. Readiness: Optimize what you have today

You’re not starting from scratch. The first step toward agentic readiness is determining what you can improve upon, starting with your product catalog.

The fundamentals of technical SEO still apply to GEO, but structured data is more important. Marking up your products in JSON-LD using schema.org vocabulary gives AI agents a clean, predictable data layer. If your products include precise, factual information — think dimensions and sizes — an AI agent can then accurately query and match them to ambiguous, conversational prompts.

However, agents must be allowed in. Robots.txt files block agents at the site level, which merchants can usually control. Audit your robots.txt file to make sure you’re not blocking all AI crawlers from indexing your website. However, many bot-providers — Cloudflare, Akamai, HUMAN — block all agents at the network level. If that’s the case, ask your providers to let legitimate agents through.

Once an agent can find your website, it has to be able to navigate it. Most website architectures were built for visual human navigation. Hosted off the website’s root directory, an llms.txt file acts as a concierge. It gives AI agents proactive information on where to find your brand information, documentation, and product specifications.

Curious about your starting point? Ask an AI shopping agent like Muse or Instinct to shop your own site.

2. Connectivity: How agents reach you

AI personal assistants can work through the front of your website like a human shopper: loading pages, reading what’s on screen, clicking buttons. This works everywhere bots aren’t automatically banned. But it’s slow and uses a lot of tokens. We recently ran an experiment and found that web scraping is also error-prone, with agents getting stuck or declined.

The alternative is headless checkout: the agent completes the purchase by sending data directly to the merchant’s cart and payment systems. Headless checkout runs on agentic protocols, the shared rules that define how agents request products, place orders, and pay.

In addition to being faster and more efficient than web scraping, headless checkout is a more controlled environment for risk assessment. Because the journey occurs via machine-readable APIs, it opens up deeper data layers. APIs give you the option to send and receive more risk signals that make real-time fraud decisioning more efficient and effective. However, APIs are not a default option and must be added manually. 

You do have leverage. Agents take the path of least resistance. Offering an API steers them away from scraping.

3. Trust: Knowing who you’re dealing with and vice versa

A poll of commerce leaders at IMPACT New York 2026 found that 69% consider trust and fraud to be their top agentic concern. The first challenge is visibility. Agents are on your site, whether or not you realize they’re there. AI personal assistants don’t always identify themselves when they browse your site, mimicking the behavior of a human shopper.

That behavior varies from one agent to the next. The combination of its underlying AI model and the software harness driving the browser tends to produce a distinct algorithmic signature. These signatures reflect how the agents type and navigate, and how long they read a screen. Because they shift every time a model updates, detection has to adapt continuously.

It’s important to work with a provider that can identify agentic traffic and the identity behind it. Knowing where an agent came from doesn’t automatically make it safe because fraudsters can use ChatGPT, too. Once you can differentiate your traffic, you can track fraud rates and velocity for that segment. Trust also runs the other way. Shoppers must be able to trust that your website is yours, and not a mirrored fraudulent website.

4. Payments: New tokens, new authentication

Scraper agents type in the full primary account number the way a human shopper would. Other agents arrive on your site with new kinds of credentials. They use PSP-issued tokens supported by Google’s Universal Commerce Protocol (UCP) specification, or shared payment tokens powered by Agentic Commerce Protocol (ACP), which was co-developed by OpenAI and Stripe. If your payments stack doesn’t recognize and accept them, the agents will move onto another site.

“How does 3DS work when there’s no human at checkout?” was a common question about agentic checkout. Both UCP and ACP support 3DS. Within 3DS, merchant-initiated 3RI lets you trigger authentication when you consider a transaction risky. Decoupled authentication can serve as the challenge, with customers verifying through another channel later. Card networks are also aligning on passkeys at the point of provisioning an agentic network token, which means the customer authenticates up front.

Because agentic is such a nascent sales channel, it’s impossible to forecast its commerce volume. The practical path forward is to build an agentic-friendly payments stack. That will help protect revenue if agentic orders cannibalize orders from existing channels.

5. Loyalty: Closing the guest checkout gap

Agents typically check out as guests. Logging in requires Open Authorization (OAuth), something most merchants haven’t enabled. As a result, the human customer misses out on points, rewards, and tier benefits.

Closing the gap is important, but connecting accounts raises the security stakes. Loyalty programs can be gamified at scale. When points and gift card balances work as payment methods, they become prime targets for account takeovers (ATO) and loyalty fraud. Because loyalty programs are traditionally managed by marketing rather than fraud, protection often lags behind the checkout lane.

Identity underpins agentic loyalty. You need to know who’s behind an account, whether human or bot, before you can set a policy for either.

See live data from the first 60 days of agentic orders

How do AI agents move through checkout? What are typical authorization rates for agentic transactions? Are chargeback rates higher on agentic transactions? On November 12, 2026, at 1:00 PM ET, join Forter VP of Payments Galit Michel for The First 60 Days of Agentic Transactions, where she’ll answer those questions and more with live data from the first 60 days of Muse, Instinct, and Grok Bot transactions.