AgentVitals Field Notes

The agentic web, decoded

Practical writing on agent-readiness, task completion, and selling to the machine majority — from the team building the Agent Experience Optimization platform.

The agency playbook: selling agent-readiness as a recurring service line

Shopify and Webflow agencies are sitting on the most natural new retainer line in years: every client site now needs to work for machine buyers, and clients can't evaluate it themselves. How to package, price, and deliver it.

Putting agent-readiness in CI: how a merge gate stops silent regressions

Audits decay. The only durable version of agent-readiness is a deterministic check that runs on every pull request and blocks the ones that break your score — here's how to wire it without hating it.

Your next enterprise evaluation will be run by an agent. Will your signup survive it?

PLG assumed a curious human clicking around your product. Increasingly the first visitor is an agent building a shortlist for a buyer — and most SaaS signup flows fail it in the first thirty seconds.

SEO, AEO, AXO: a working taxonomy for machine-era optimization

Three acronyms, three different questions: can you be found, can you be cited, can you be operated? Where each discipline starts and stops — and why the third is where transactions happen.

The Product JSON-LD that makes your catalog machine-buyable

Copy-paste structured data for products and offers, the properties agents actually read, the validation mistakes that get your markup silently discarded, and how to keep it correct at scale.

How AI shopping agents actually buy: an anatomy of autonomous checkout

Step by step through a real agent purchase attempt — discovery, extraction, cart, checkout — and the specific point in the flow where most e-commerce sites lose the sale.

What is an Agent Readiness Score? Inside the seven categories

The 0–100 number is the headline, but the category breakdown is where the money is. What each of the seven categories measures, why it's weighted the way it is, and which failures cost the most.

GPTBot, ClaudeBot, PerplexityBot: the robots.txt lines that decide your AI traffic

A practical guide to the crawler user-agents behind AI assistants — what each one feeds, what blocking actually costs, and a sane default policy for a commercial site.

ChatGPT is in your referral report. Here's what that traffic does next.

Growth teams everywhere are seeing the same new referrer line. The uncomfortable part isn't the volume — it's what happens after the click, and why your analytics can't see the failures.

llms.txt, explained: what it is, what goes in it, and a template you can ship today

The proposed standard for telling AI agents what your site can do — what it actually is, what belongs in it, the mistakes we see in the wild, and a copy-paste starting point.

Should you block AI agents, or build for them?

The fight over who's allowed to crawl and act on the web is real — and mostly a distraction from the question that pays: when a legitimate buying agent shows up, can it actually transact with you?

Why AI agents abandon your checkout — and the fixes that bring them back

The failure modes are surprisingly consistent across sites: unlabeled forms, invisible buttons, missing schema, ambiguous CTAs. Here are the six that cost the most, and the code-level fix for each.

Task completion is the new conversion rate

You can't optimize what you don't measure. Here's how to put a number on whether agents can actually finish the jobs your site exists to do — and why that number predicts revenue better than traffic.

57% of web traffic is automated. Most sites are built for the other 43%.

Agents are now the majority of visitors to the web. The infrastructure under your site still assumes a human with eyes, a mouse, and patience. That mismatch is where revenue leaks.

From around the web

Articles and discussions we found worth your time.