Meta, Sierra, Shopify and Walmart are backing a standard for AI agents that shop on customers' behalf. Here's what it means for your site and what to prepare.
On 6 October, Sierra announced the Personal Agent Protocol, an open standard being developed with Meta. It sets out how a customer's personal AI agent should identify itself to a business and what it is allowed to do there. Genesys, Instinct, Rocket, Shopify, Stripe and Walmart are named partners (The Next Web).
If you sell anything online, this is worth reading now. Some of your future visitors will not be people clicking around your site. They will be software acting for a person, asking about stock, return policies and order status. The way your business answers will decide whether those customers keep going.
What the Personal Agent Protocol actually is
It is a proposed rulebook, not a product you switch on. According to coverage of Sierra's announcement, it is built on OAuth and covers how an agent authenticates with a business and how the business sees what that agent does (Channel Insider). The main pieces:
- Guest access first. An agent can connect as a guest to check things like product availability or a return policy.
- Customer-controlled permissions. Once the customer signs in, they choose whether the agent gets read-only or write access.
- Business-controlled actions. The business decides which actions agents may perform.
- One session across channels. A question asked before login and an order change made afterwards can count as a single visit (The Next Web).
A business can let agents in through its website, through APIs built on standards such as MCP and OpenAPI, or through an agent of its own (The Next Web).
What it does not cover yet
Be realistic about the timing. A v0.1 specification is due later in October, so nothing is final. Payments, push notifications and more detailed permissions are listed as possible future extensions, with no dates given (Channel Insider).
It also is not the only effort. Visa's Trusted Agent Protocol, reported in June, lets agents buy from merchants through ChatGPT, and Stripe and Shopify now sit in both camps (The Next Web). Amazon has blocked Meta's Muse agent over scraping concerns, so the biggest retailers are not all on board. One analyst at Constellation Research called the spec "really early" but suggested it could move quickly, as Model Context Protocol did (Constellation Research).
Why this matters even if you never adopt it
Meta's own agent, Muse, launched in September. A Meta executive told CNBC it already had millions of US users (Channel Insider). Whether or not this particular protocol wins, agents are already sitting between customers and websites.
Sierra co-founder Bret Taylor described the current situation as "kind of chaos until such a standard exists" (Channel Insider). That chaos is what small and mid-sized businesses feel first: an agent that cannot read your return policy will simply tell the customer it could not find one, and the customer moves on to a competitor whose answer was easy to find.
There is a measurement problem too. Constellation Research's analyst questioned how personal agents will affect sales and how disputes will be handled (Constellation Research). If agent visits show up in your analytics as odd traffic, your conversion data gets harder to trust.
A practical readiness checklist
You do not need to build anything for a spec that is not published. You can, though, fix the things any agent, from any vendor, will trip over.
- Put the answers agents ask for in plain text. Stock status, delivery times, return windows and pricing should be on the page as readable text, not locked in images, PDFs or pop-ups.
- Clean up your product data. Accurate titles, variants, prices and availability matter to AI shoppers as much as to Google. A Shopify SEO and marketing review is a good place to start.
- Check your structured data and crawl access. Make sure the pages that carry those answers can be read and are not blocked by accident. Our technical SEO work covers this.
- Decide your line between agent and human. Which actions would you allow an agent to take on a customer's behalf, and which need a person? Write it down now. The protocol is built around businesses making exactly this call.
- Watch your analytics. Segment unusual, very fast or repetitive sessions so they do not distort your conversion rate. If your setup is patchy, web analytics setup is the fix.
Visibility in AI answers is the same job
Agents that recommend and compare products draw on the same signals as AI search: clear, accurate, well-structured information that is easy to quote. That is the work behind GEO and AI search optimisation. If a customer's agent cannot work out what you sell and on what terms, it will not suggest you.
Key takeaways
- The Personal Agent Protocol is an early, open standard backed by Meta, Sierra, Shopify, Stripe, Walmart and others. Its v0.1 spec is due later in October.
- It gives customers control over what their agent can do and gives businesses control over which actions are allowed.
- Payments are not in the first version, and a rival effort from Visa already exists, so expect the standards picture to shift.
- The sensible move now is not to build for the spec but to make your product, policy and stock information easy for software to read, and to keep your analytics clean.
If you would like a second pair of eyes on how readable your site is to both people and AI tools, we're happy to help. You can get a free quote and we'll tell you where to start.
John Doe
Supreb Information. Thanks
15 July 2014 View replies . 1