AI Search Optimization Checklist: 30 Steps to Get Your Website Cited by AI

A customer opens ChatGPT or Google and types: "recommend a reliable company to install commercial ventilation in New York City." A few seconds later there is a coherent answer on the screen with three to five company names, short explanations, rough price ranges and links to source websites. In Google it is an AI Overview on top of the results. In ChatGPT, Perplexity and Gemini it arrives in the chat window with citations. The customer reads the answer and chooses from the companies it names. Everyone else simply does not exist for them.

Optimizing your website for AI search means making sure AI can open it, understand it and use it as a source. Below are the 30 checks we use to assess whether a website is ready to be cited and recommended by AI. Work through them in order: problems in the first sections cancel out everything you do in the last ones.

Why AI passes over a good website

When AI answers a question about products or services, it usually does not rely on its "memory" alone. First it searches the web, then it picks a few relevant pages from what it found and retells them in its own words. Your website can drop out at any step. A crawler may be blocked, the right page may be missing from the index, or the text may be loaded by a script the crawler does not run. Or everything opens fine, but the price and location have to be guessed, so the AI picks a competitor's page where everything is stated plainly.

For a business, this turns into very real losses:

• Competitors in the answer instead of you. The customer chooses from the few names the AI mentioned and never reaches your website.
• Less traffic at the same rankings. People read the AI answer above the results and do not click, or they click the source the AI cited.
• Wrong facts about your company. The AI finds an old price list on a third-party site or confuses you with a company with a similar name.
• Wasted content budget. Good articles do not work if crawlers cannot see them.

The good news: optimizing a website for AI answers is not a lottery, and the way AI selects sources is predictable enough to act on. Generative engine optimization (GEO) is built on top of good old SEO. Most checks below will look familiar to anyone who has done search engine optimization; they just carry a different weight now.

How to use this AI search optimization checklist

Mark each item "yes," "no" or "partly." The sections are ordered by priority. The first two check whether AI can read your website at all. The next four check whether it will understand and trust it. The last two cover authority and measuring results.

Section 1. Crawler access: checks 1–5

1. robots.txt does not block search engine and AI crawlers

Open yourdomain.com/robots.txt and look for User-agent groups with a "Disallow: /" rule. Make sure Googlebot, Bingbot, OAI-SearchBot, Claude-SearchBot and PerplexityBot are not among them. These blocks are often left over from an old "anti-scraping" setup or copied from someone else's template. Training crawlers (GPTBot, ClaudeBot and the Google-Extended token) can be blocked if that is your company's decision, but that is a separate question, and blocking Google-Extended does not affect Google Search. We cover which crawlers to allow in detail in GPTBot and other AI crawlers: which ones to let in.

2. Bot protection does not silently block the crawlers you need

Sometimes robots.txt allows everything, but the server responds to the crawler with a 403 error or a "verify you are human" challenge. This happens because of a CDN, a hosting firewall, a DDoS protection module or geo-blocking of foreign IP addresses. Website owners usually have no idea. You can check your server logs: which response codes do requests from these crawlers receive?

3. Important pages are indexed in Google and Bing

Google AI Overviews and AI Mode are built on the Google index, and Copilot relies on Bing. If a page is not in the index, it will not be in the answer either. In Google Search Console and Bing Webmaster Tools, check that your service pages, pricing, contact page and articles are indexed and not excluded as duplicates or low-value pages. If you also work in Russia, add Yandex Webmaster, since Yandex Alice AI relies on Yandex Search.

4. Status codes, redirects and sitemap.xml are in order

Key pages return a 200 status code. There are no chains of multiple redirects, nonexistent URLs honestly return 404, and duplicates have a canonical URL. The sitemap lists only live canonical pages with accurate last-modified dates, with no junk, filter pages or system sections.

5. The website has an llms.txt file with a short description and key links

This file in the root of the website briefly describes the company and points to its key pages. To be upfront: the format was proposed in 2024, and the major AI platforms have not officially confirmed that they use it. But it takes about an hour to create, it does no harm, and writing it forces you to gather your key facts in one place. We explain how to write one in the complete guide to llms.txt.

Section 2. Speed and HTML that reads without JavaScript: checks 6–10

6. The main text is in the initial HTML

Many AI crawlers do not execute JavaScript, or execute it only partially. They need text the server delivers right away. The check is simple: open "View page source" (the actual source, not the developer inspector) and search it for the service name, the price and the first FAQ question. You can also load the page with JavaScript turned off. If the text is missing, the crawler sees an empty shell. This is common with single-page applications built on modern frameworks without server-side rendering. The fix is server-side rendering or prerendering.

7. The server responds quickly

Crawlers have limited time to wait. A fetcher that opens a page on a user's request, such as ChatGPT-User, does so while the person is waiting for an answer in the chat. A slow server easily loses to a faster source. What helps: caching, lightweight images, fewer third-party scripts and decent hosting.

8. The mobile version is complete

Search engines look at the mobile version first. If descriptions, pricing tables or FAQ sections are cut from it, that information barely exists for the index.

9. Important information is not hidden behind clicks or in images

Prices on an image banner, a price list only in a PDF, terms in a slider, answers in tabs that a script loads on click. A person will get to all of it, but a crawler probably will not. Duplicate everything important as plain text on the page.

10. The markup is semantic

Each page has one h1 heading. The h2 and h3 subheadings follow the logic of the text. Lists use list tags and tables use table tags, not a set of blocks that only look like a table. This markup lets a machine tell where the question is, where the answer is and where the price is.

Section 3. Service page structure: checks 11–15

11. One service, one page

An "All our services" page describing ten offerings in a couple of lines each is nearly useless to AI. It needs a dedicated page for a specific question: "restaurant ventilation installation," not "building systems."

12. The first paragraph answers "what, for whom, where and how much"

The first two or three lines of the page should answer the customer's question on their own: what you do, for whom, in which city or region and in what price range. AI often takes exactly this opening into its answer, so "welcome to our website" and company history do not belong there.

13. Prices, timelines and terms are specific

"Price on request" loses to "from $2,500 per site up to 1,000 sq ft, completed in 5 to 7 business days" (a hypothetical example). If you cannot give an exact price, give a range and explain what it depends on. AI readily repeats specifics and skips vague wording.

14. The process, guarantees and limitations are described

The stages of work, what is and is not included in the price, what guarantees you offer, which jobs you do not take on. This kind of text answers the follow-up questions ("what if...") people ask AI assistants and makes the page more trustworthy as a source.

15. Pages are connected with internal links

Each service page links to related services, case studies, articles and FAQ, and those link back. This makes it easier for a crawler to navigate the website and understand which pages matter most.

Section 4. Direct answers, tables and FAQ: checks 16–20

16. Prompt research is done

Prompt research is the list of questions customers ask AI assistants about your industry, the AI-search equivalent of keyword research. These prompts are longer and more conversational: "for a small cafe, is a supply air unit or a split system better?" Without prompt research, content gets written by guesswork.

17. The answer comes in the first sentence

For every tracked prompt, there is a place on the website where the answer is given right away, with details afterward. A hypothetical example. Question: "Does a cafe need a ventilation design plan?" Answer: "In most cases, yes: a commercial kitchen and dining room usually need a design plan, because..." A paragraph like this is easy to quote verbatim.

18. Comparisons and prices are formatted as tables

"Plan A vs. Plan B," "material X vs. material Y," "pricing by square footage." AI parses tables reliably and often answers in table form itself. An HTML table is worth more than the same information spread across three paragraphs.

19. Service pages have an FAQ section

Five to ten real questions, taken from prompt research and from your sales team, with short, to-the-point answers. The answers are visible on the page right away rather than loaded by a script.

20. Expert articles have an author and an update date

An article signed by a specialist who has their own page on the website earns more trust than an anonymous text. An update date shows the information is current. Old articles with outdated prices should be updated or removed.

Section 5. Company facts: checks 21–24

21. The About page lists verifiable facts

Full and short company name, year founded, legal entity, addresses, service area, licenses and certifications, and number of employees or clients if you are willing to share it. You need facts that can be verified, not slogans like "we are the market leader."

22. Your name and contact details are identical everywhere

The name, address and phone number on your website, map listings, directories, social media and third-party websites match exactly. If one place says "Company N LLC" and another says "CompanyN Group," AI may treat them as two different companies.

23. There is a single source of truth for company facts

This is an internal document with the current approved wording: company name, one-line description, list of services, prices, addresses and key numbers. Everything published on the website and beyond is checked against it. When a price changes, it changes in this document first, then everywhere else.

24. Your experts have their own pages

The CEO, the lead engineer, the medical director: each has a page with experience, specialization, education and publications. AI considers not only the authority of the company but also the authority of the people behind it.

Section 6. Schema markup: checks 25–26

25. Core schema types are implemented

Organization with the name, logo, contacts and links to profiles. Service or Product for services and products. FAQPage for question-and-answer blocks. Article and Person for articles and authors. BreadcrumbList for the site structure. Markup does not replace content, but it removes ambiguity: a machine will not confuse a price with a product code, or your address with that of a similarly named company. We explain which fields matter in schema markup, FAQ and your company as an entity.

26. Markup matches the visible text and passes validation

The markup contains no prices, reviews or addresses that are not visible on the page. Pages pass the Schema.org validator and Google's Rich Results Test without errors. If markup contradicts the text, that is an error, not a clever trick.

Section 7. Brand mentions: checks 27–28

27. The company has complete listings on maps, directories and review platforms

Your Google Business Profile, Bing Places listing, industry directories and review platforms are fully filled out, and the data matches your website. Reviews get responses. When AI checks whether a company exists and what people say about it, these are often the places it looks.

28. Third-party websites write about the company and its experts

Articles in industry media, expert columns, inclusion in rankings and roundups, mentions by partners. AI trusts not what a company says about itself, but what others confirm. These publications need to be created systematically and aligned with your tracked prompts.

Section 8. Tracking: checks 29–30

29. Analytics tracks traffic from AI assistants

In Google Analytics 4, visits from AI assistant domains are grouped into a separate segment or channel, with conversions attached: leads, calls, purchases. We walk through the setup step by step in tracking AI traffic in Google Analytics 4.

30. Tracked prompts are checked regularly

Once a month, your tracked prompts are run in the AI assistants your customers use. You record whether the company is mentioned, whether your website is cited, who is named instead of you and which facts are wrong. Traffic shows only part of the picture, because many people read the answer and never click. The second metric is your share of mentions across tracked prompts.

Why you cannot run this checklist just once

All 30 checks can be completed in a few weeks, but six months later some of them will be failing again. A developer ships an update and accidentally blocks a section in robots.txt. The hosting provider turns on new bot protection. Prices change on the website but not in directories. AI platforms launch new crawlers and new answer formats. That is not a reason to give up; it is simply the normal life of a website.

There is a second reason. AI evaluates a website as a whole, not one successful article. It will not recommend a website with slow pages, errors and conflicting facts, no matter how many FAQ sections it has. That is why optimizing a website for AI search only works alongside ongoing website care, with one team responsible for the technical side, content and third-party presence.

What you get once the checklist is done

Nobody can guarantee a spot in a ChatGPT answer or a Google AI Overview: answers change from one wording to another and from week to week. What can be guaranteed is that there are no reasons for AI to skip you. The website is open to the right crawlers, serves its text quickly, answers customer questions directly, company facts match everywhere and third-party websites confirm them. A website like this becomes a strong candidate to be cited. From there, you see AI referral traffic in your analytics, and your tracking report shows which prompts already mention your company and where it is still missing.

How we optimize websites for AI search

AI Search Optimization has been doing SEO since 2002. We have 100+ clients around the world, Google and Yandex certifications, and one team that handles your website and your AI visibility end to end. We optimize for Google AI Overviews and AI Mode, ChatGPT, Gemini, Claude, Perplexity and Microsoft Copilot, in any market. This checklist is part of our everyday work:

• AI visibility audit: we check your website against all 30 items and see what AI assistants already say about you and your competitors;
• prompt research: the questions your customers ask AI;
• technical fixes on the website: AI crawler access, llms.txt, schema markup, speed, HTML, error fixes;
• new content built for AI answers: service pages, FAQ, tables, expert articles;
• brand mentions on third-party sites;
• articles in external sources that build the authority of your company and its experts;
• ongoing AI visibility tracking across all your tracked prompts.

We take care of the entire website from day one, from technical fixes to GEO, because AI looks at the website as a whole. We guarantee delivery of GEO tasks, error fixes and fast adaptation of your website to changing AI search requirements.

The fee is $290 per month, including the first 10 hours of work. Every additional hour is $45. GEO work is billed time and materials: we invoice only the work completed and delivered that month. No discounts or free months; the price is honest from the start.

Book a call. We will run the checklist on your website, show you who AI recommends instead of you to customers from New York City, and start with the issues that hold you back the most.