An AI search website checklist you can give your developer

The short answer

Start with a page that customers and relevant search systems can access, understand and use. Ask your developer to show the checks below on one important service page. A score from an AEO tool is a diagnostic opinion, not an engine’s promise to recommend you.

1. Establish access before changing infrastructure

Check the live URL, redirects, page status, robots rules, indexing directives and any firewall challenge. Keep private account pages private. Google’s AI search features use its search eligibility rules; no DNS switch is inherently required.

• Google: AI features and your website

2. Choose crawler controls deliberately

OpenAI separates OAI-SearchBot, used for search, from GPTBot, used for potential training collection. ChatGPT-User handles certain user-triggered visits. A training preference and a search preference are different decisions. Check current provider documentation before editing rules or firewall allowlists.

• OpenAI: crawler roles and controls

3. Make one page useful to a real prospect

Our suggested editing exercise: ask the owner to supply the exact service, actual coverage area, availability, contact route and any qualifications that can be checked. Then compare those facts with the page.

Illustrative example: replace “quality service you can trust” with a specific description of what an appointment includes, the areas served and how the estimate works. Obtain approval before publishing prices, response times or credentials. This is a content exercise, not a measured performance result.

4. Check the rendered page

Open the page on a phone and without JavaScript. Confirm the main answer, contact route and internal links remain usable. Inspect the live canonical and structured data as well as the source code. Our site serves guide text in the initial HTML to make this easier to inspect.

Give people and crawlers the same substantive information. Different technical rendering is not automatically deceptive, but showing bots extra claims or keywords to manipulate search creates a different risk.

• Google: spam policies

5. Avoid unsupported requirements

Google says special AI files, including llms.txt, do not improve its search visibility. It also specifies no special schema for generative search. These statements concern Google; they are not a universal claim about every agent or future system.

FAQ answers can still help visitors. Google retired FAQ rich results in May 2026, so do not buy FAQ markup on the promise of that display feature.

• Google: optimizing for generative AI search
• Google: search documentation updates (FAQ retirement)

What a schema study can—and cannot—tell you

Ahrefs tracked 1,885 pages that added JSON-LD and compared them with 4,000 control pages. Its matched analysis found no clear positive citation uplift across the studied search surfaces. The pages were already heavily cited, so this does not settle what happens to a new or undiscovered website.

The practical decision: use accurate markup where it describes visible content and supports a relevant feature. If citation improvement is the objective, measure that separately. Do not sell a schema installation as a proven visibility gain.

• Ahrefs: schema additions and AI citations (May 2026)

A handoff you can send

“Please review this service page’s access, rendered text, indexing directives and business details. Show the findings, propose one change, provide a preview and explain how to roll it back. Please do not change DNS or add unverified claims as part of this review.”

• Save the old page and the approved change.
• Record publication time and the person responsible.
• Check the live page after publication.
• Keep visibility measurements separate from technical acceptance.