AEO and GEO

How do you make a business usable by AI agents that book and buy?

Making a business usable by AI agents starts with open doors: letting AI crawlers in (18 publishers back it), serving plain HTML rather than pages that only render with JavaScript, and keeping CAPTCHAs and logins out of the path to your facts and your booking. For agents that act, 16 publishers back an MCP server, while OpenAI's agent checkout draws more doubt (9 against, 1 for). On 14 November 2026, a fresh AI agent given only our web address booked a check on our site through our JSON API, with no human help.

Updated 19 November 2026

An AI agent can only recommend or book your business if it can read your pages and finish the action without hitting a wall built for humans. Across 320 publishers in our corpus, the strongest agreement is on the basics: 18 back letting the crawlers in through robots.txt, and many warn against content that only appears after JavaScript runs or behind a CAPTCHA. For direct transactions, 16 back an MCP server, while OpenAI's Agentic Commerce checkout has 9 publishers against and 1 for. For many small businesses, being on a booking platform the assistants already reach is simpler than building your own.

What our source count shows

Each row below is one tactic, and each cell counts how many of the 320 publishers in our 435 saved sources took that stance. A publisher can take more than one stance, so the columns do not add up to a total.

StepRecommend itMeasured it workingCall it useless or warnMeasured no effectMeasured harmOur label
Let the AI crawlers in (robots.txt)180001Recommended, little measured
Content in the HTML, not behind JavaScript172001Recommended, little measured
No CAPTCHA, login wall or forced account in the agent's path; guest checkout70201Contested
Forms and buttons a browser agent can operate90100Recommended, little measured
An MCP server for the business143000Recommended and measured
A public, documented API for booking or ordering40100Recommended, little measured
Be on a platform that is already wired in203210Recommended and measured
Live price, stock and availability in machine-readable form161100Recommended, little measured
An app or connector in the ChatGPT directory60410Contested
Agentic Commerce Protocol and Instant Checkout (OpenAI, Stripe)10522Mostly called useless or risky
Agent payment rails80230Contested

What happened when an AI agent visited our site

We built our own site so agents can use it, with 5 doors for them next to the pages people read. On 14 November 2026, we gave a fresh AI agent nothing but our address and asked it to book a check; it found the booking path and booked through our JSON API, with no human help.

What this means for your business

We saved 435 guides, studies, forum threads and platform docs from 320 publishers in full, recorded 1592 word-for-word method statements across 99 tactics, and logged 216 reported results with numbers. Every count is by publisher, not by article, so one prolific blog only gets one vote per stance, and every quote is checked word for word against our saved copy. A high count is not proof that a tactic works: many GEO writers sell GEO services or software, and a measured claim usually means one site's before-and-after numbers rather than a controlled test.

Common questions

Should you block AI crawlers in robots.txt?

If you want AI assistants to cite your pages or check your prices and hours, their crawlers need to be allowed in robots.txt; 18 publishers in our corpus back this. Crawler names change, so check the list each assistant publishes before you edit the file.