← Back to Lab Notes
Agentic Search Optimization

From Cited to Chosen: How Agentic Search Optimization (ASO) Makes AI Agents Pick Your Business

GEO got your business into the conversation. Now there's a bigger prize: being the one the AI actually picks.

For the last few years, the goal has been getting cited — making sure ChatGPT, Claude, Gemini, and Perplexity mention your business when someone asks a question. That's Generative Engine Optimization (GEO), and it matters. But there's a second, newer job your website has to do, and most businesses haven't started it yet.

That job is being chosen. When a customer stops asking questions and starts asking the AI to do something — "find me a clinic near me that's open Thursday and start a booking" — the AI has to pick a single business to act on. The business it picks is the one whose information it can read as data, not the one whose website merely reads well to a human. That's Agentic Search Optimization, or ASO. (One heads-up on the acronym: in the app world, "ASO" usually means App Store Optimization. That's a completely different thing — we mean Agentic Search Optimization.)

GEO vs. ASO: cited is not the same as chosen

Here's the distinction in one line: GEO makes you the answer. ASO makes you the choice.

GEO is about showing up when someone asks a question. ASO is about being the business an AI agent can actually act on — book, schedule, route, or recommend with confidence. A business can be perfectly cited and still lose the moment the customer tries to take the next step, because the AI can't read its prices, hours, or booking link as data. If an agent can't confirm you're open, can't see your prices, and can't open your booking page, it will quietly hand the job to a competitor whose website it can read.

The agent-readiness ladder

Being agent-ready isn't one switch. It's a ladder, and most businesses are sitting on the bottom rung without realizing it.

Level 0 — Invisible. No structured data. An AI has to guess your name, address, and phone — and it usually gets them wrong.

Level 1 — Identified. A machine can tell who you are and where you are. Your name, address, phone, and hours are marked up so an AI can read them as data.

Level 2 — Screenable. The AI can answer the routine questions — open hours, prices, quality, service area. This is where you become reliably citable, and it's the foundation everything else is built on.

Level 3 — Reachable. There's a stable, machine-usable front door to starting a booking or enquiry. This is the most-skipped rung, and the best value after Level 2 — because it's where a customer's intent turns into action.

Level 4 — Verifiable. A machine can tell whether you actually have openings. Not "contact us for availability" — a real signal the agent can read. This is where selection is won.

Level 5 — Transactable. The agent completes the booking end to end. This is only worth building on top of a real scheduling system — and we'd never sell you Level 5 before you've cleared Level 2.

For almost every business, the highest-return work sits at Levels 3 and 4. That's the gap between being mentioned and being chosen.

What an agent-ready business looks like

The difference between a citable business and a chosen one comes down to a handful of concrete, machine-readable signals:

Exact prices, not ranges. "$1,275" beats "affordable." An agent can quote a specific price. It can't do anything with a vague promise.

A machine-readable service menu. Every service is its own item, with its own price and description — not one block of prose an agent has to parse and guess at.

Hours in a format machines read. Not a sentence ("open most days"), but structured hours an agent can check against "open Thursday."

One unambiguous location. If your address appears four different ways across the web, an agent can't tell which is real — so it moves on.

A bookable link a machine can open. Deep-linkable, works without JavaScript, and doesn't hide behind a bot-block that also blocks the agents you want.

An availability signal. Can a machine tell you have openings right now? If it can't, you're asking the agent to gamble on you — and it won't.

Why this matters now

The shift from asking to acting is already happening. Customers don't just ask ChatGPT who to hire anymore — they ask it to do the next thing. Book the appointment. Check availability. Start the enquiry. Every one of those moments is a selection event, and the business the AI picks is the one whose website hands it clean, readable, unambiguous data.

The businesses that treat their website as something an AI has to read — not just something a human has to admire — are the ones quietly winning those selections today. Everyone else is still fighting to be the first answer, while the real transaction walks out the door.

Getting cited was step one. Getting chosen is the whole point. And the ladder between the two has a clear first rung: make sure an AI can tell who you are, what you charge, when you're open, and how to book you.