AI Agents Are Visiting Your Website Right Now. You Have No Control Over What They Find.
By Nick Allain
If you read Matt Shumer’s recent essay about AI — the one that’s been quietly circulating through executive inboxes — you felt it. That specific unease of someone telling you the future is already here, and you missed the memo.
He’s right. But he’s describing the human side of the equation: how professionals adapt, retrain, survive. What nobody is talking about is the other side — what happens to your website, your brand, your business, when AI agents start doing the visiting.
They already are.
Your Website Has a New Audience. It’s Not Human.
When someone asks an AI assistant to research your company, compare your pricing, or book a service through you, an agent visits your site. Not a human. Not a search crawler. An autonomous system with a task, a deadline, and a limited tolerance for ambiguity.
And here’s what that agent finds: a website built for humans.
Navigation menus it has to parse. Marketing copy it has to see through. FAQs it has to cross-reference. A contact form it can’t fill out. Pricing buried three clicks deep. Terms written to be read, not processed.
The agent does its best. It scrapes, infers, summarizes. And then it tells the user what it thinks your business offers — based on a best guess from an HTML page that wasn’t designed for this conversation.
You had no input. You had no control. You may not even know it happened.
This Is a Brand Problem, Not a Tech Problem
Think about what you’ve invested in your web presence. The copy that took months to get right. The positioning that differentiates you from competitors. The service details that matter to the customers who actually convert.
Now imagine an AI agent summarizing all of that in four sentences, getting two of them wrong, and presenting the result to a potential customer as if it were fact.
That customer may never visit your site directly. They asked their AI, and the AI answered. The interaction is over before you got a word in.
This is the version of AI disruption nobody’s preparing for. Not “will AI replace my employees?” but “will AI misrepresent my business to my customers, and will I have any way to know?”
The Handshake You’re Missing
There’s a reason we call it the Agent Handshake Protocol.
A handshake is mutual. Both parties show up. Both parties communicate. The interaction has structure, intent, and shared understanding.
Right now, when an AI agent visits your site, there’s no handshake. There’s just the agent, doing its best, with whatever it can scrape.
AHP gives your site a voice in that interaction.
You publish a simple manifest — a machine-readable file at a predictable address — that tells any AHP-compliant agent exactly what your business offers, how to ask for information correctly, and what actions are available. The agent reads it before doing anything else. Your content, your framing, your structure — not an inference from your HTML.
It’s the difference between a customer walking into your store and having a trained employee greet them, versus them wandering the aisles alone until they give up and go somewhere else.
What This Actually Looks Like
A travel company implements AHP MODE1: a structured description of their packages, destinations, and pricing. When an AI agent compares travel options for a user, it reads this directly — accurate, current, in your words — instead of guessing from your homepage.
A law firm implements AHP MODE2: a conversational endpoint backed by their knowledge base. An agent researching legal services can ask “do you handle commercial lease disputes in California?” and get a real, sourced answer. Not a hallucination. Not a scraped paragraph from a five-year-old blog post.
A SaaS company implements AHP MODE3: tool-enabled actions. An agent helping a user evaluate software can actually trigger a demo, check live pricing, or start a trial — all within the agent conversation, without the user having to switch contexts or navigate a signup flow.
Each level is opt-in. You start where you’re comfortable and expand as the value becomes clear.
The Window Is Narrow
Shumer’s essay makes the case that the people getting ahead of AI disruption are the ones moving now, not when it’s obvious. That’s true for professionals. It’s equally true for businesses.
The companies that implement AHP early will have accurate, structured representation across every AI platform that adopts the standard. Their competitors — the ones waiting to see how this plays out — will be represented by whatever an agent could scrape, whenever it visited, on whatever day the content happened to be indexed.
This is the same dynamic that played out with SEO. In 2004, most companies didn’t have an SEO strategy. By 2010, not having one was a competitive liability. The window for first-mover advantage was short.
AI agent traffic isn’t a future concern. The agents are already visiting. The question is only whether you’re speaking their language or hoping they figure it out on their own.
Learn more: agenthandshake.dev
Read the spec: agenthandshake.dev/spec
Try the reference implementation: ref.agenthandshake.dev
Nick Allain is building AHP — an open standard for agent-native web interactions. The spec, reference implementation, and test suite are all open source.