25-Point Website Audit Checklist for AI Search Visibility

When checking AI search exposure, the first thing many reps look at is robots.txt or JSON-LD. Both items are important, but allowing robot access or applying structured data alone does not make the website a source of answers or brand information.

In order for search engines and AI systems to understand a site, they must be able to read what company it operates, what questions each page answers, what the representative URL is, how it was created and the date of modification, what evidence was used, and how related documents are connected.

GEO is an operating method that checks and reinforces this technical structure, content, and trust information according to the context of the question. The basic concept and the difference from SEO can be found in the ‘What is GEO?’ guide, and this article focuses on the diagnostic items on the website itself.

The goal of the diagnosis is not to pass all 25 items or to produce a good score. The key is to find issues that prevent collection, issues that make it difficult to understand the topic, and issues that lack a basis for trust on the official page linked to important questions and determine the order of correction.

Structured data, robots.txt, or specific meta tags alone do not guarantee AI citations. Collection accessibility, page topic, operating entity, content basis, and relationship between documents must be consistent.

Identify access blocking issues before content quality. There is a difference between a page existing and being indexable, and what appears on the browser screen is different from what is output in the initial HTML. Metadata and JSON-LD must also be checked not only for their existence but also for whether they match the actual text.

Public results can be categorized as ‘immediate improvement’, ‘priority reinforcement’, and ‘normal or observation’. Items that can directly affect access and indexing, such as robots blocking or noindex on public pages, are considered targets of immediate improvement, and items with weak explanations and evidence are reinforced first, starting from important pages.

Pages that are important to your business are viewed first before the diagnostic score. After modification, we re-examine the server response, initial HTML, metadata and JSON-LD of the same URL and note that it may take time to reflect the actual index and search results. We do not assume that a single discovered item is the single cause of non-exposure.

The following table summarizes the 25 diagnosable items that can be disclosed in terms of collection, understanding, and trust. A good signal is not a guarantee of passage, but rather a status that can be seen now, while a problem signal is an observation that requires further verification.

The Collection, Index, and Server groups verify content before evaluating it. Even if the configuration file looks normal, it may be blocking the actual request, so examine the file contents, response headers, and server behavior separately.

Check for: User-agent: * rules, Disallow misconfigurations, sitemap declarations, major crawler rules and domains/paths. Why it's important: robots.txt guides the crawler to a range of URLs to access, but does not guarantee indexing or AI citation. Sign of a problem: Public pages or rendering assets are blocked or sitemaps are declared for other domains. Directions for improvement: Allow only necessary public paths and clean up unnecessary rules. Verification method: robots.txt Check the actual request results of the original text and target URL together.

Check for: Server firewall, CDN/WAF, country-specific blocking, repeat request limit, CAPTCHA, 403/429 response and User-Agent-specific policies. Why it's important: Even if robots.txt is Allow, the body cannot be taken if the security layer rejects the request. Sign of a problem: The regular browser opens, but certain requests are blocked or redirected to the login/challenge screen. Direction for improvement: Establish reasonable access policies for crawlers and public pages that can be confirmed through official documents while maintaining security. Verification method: Compare the server log and the status/response body of the same URL.

Check for: meta robots, X-Robots-Tag, per-page noindex, staging settings and canonical conflicts. Why it's important: noindex is a means of directing search index exclusion, unlike robots.txt which deals with crawl traffic. Problem signal: noindex left in public page HTML or response header. Direction for improvement: After confirming public intent, modify meta and header settings together. Verification method: Read the initial HTML head and HTTP response header directly and compare the results for each environment.

Check for: Includes public URLs, 200 response, normal XML, canonical URL, lastmod, excludes duplicates and 404·redirect·noindex URLs. Why it's important: A sitemap conveys URLs to be discovered and change information, but submission alone does not guarantee indexing or ranking. Signs of a problem: Mixing of private, duplicate, different host URLs, or repeated lastmods that are unrelated to the actual modification. Directions for improvement: Keep only public authentic URLs. Verification method: XML parsing results are compared with the actual page state.

Check for: HTTPS certificate, 200 status, 301·302 chain, www·non-www, trailing slash, HTTP→HTTPS, redirect loop and soft 404. Why it matters: Multiple URL variations and long breadcrumbs can make it difficult to interpret and access representative documents reliably. Sign of a problem: The same page opens on multiple hosts or navigates to the final URL repeatedly. Directions for improvement: Use one representative host and short persistent redirects. Verification method: Record the status and Location header from the first request to the last 200.

The Rendering, Accessibility, and Technology group ensures that the results displayed in the browser are the same as the document at the time of collection. You need to look at the source, rendered DOM and mobile screen respectively.

Check for: H1/core body of page source, pre-JavaScript content, infinite scroll, post-click loading, API errors and hydration errors. Why it's important: The content you see on screen may not be in the initial HTML and may result in an empty document if execution fails. Signaling a problem: Only the shell is output and the important answer appears after the client request. Directions for improvement: Provide key descriptions as server-rendered or static HTML. Verification method: Compare the original HTML and the post-rendered DOM.

Check for: Initial server response, load failures, timeouts, excessive images, layout shifts, hydration errors and mobile network status. Why it's important: It's more important than a specific speed score that the core body is reliably served every time you request it. Signs of trouble: Intermittent 5xx, blank screen, long waits or large elements moving repeatedly. Directions for improvement: Reduce server errors and key asset bottlenecks. Verification method: Repeatedly check whether the status code and body are provided under several conditions.

Check for: Viewport settings, horizontal overflow, tables/code blocks, CTAs, font size, mobile menus, and image truncation. Why it's important: On mobile, truncated content can make it difficult for both users and collection systems to determine document relationships. Signs of a problem: The entire page is sliding left or right, or buttons and table columns are stuck off-screen. Directions for improvement: Tables provide their own horizontal scrolling and limit body width. How to verify: Check actual manipulation and horizontal overflow at key breakpoints starting at 320px.

Check for: html lang, body language, multilingual page, hreflang if present, view target URL and language code. Why it matters: Document language is the primary clue that readers and search systems use to interpret page context. Sign of a problem: The Korean body is given a different language code or the language-specific pages incorrectly share the same canonical. Direction for improvement: Unify language information to match actual content and URL policy. Verification method: lang·hreflang·canonical of the final HTML are compared together.

Check for: meaningful image alts, empty alts for decorative images, file names, captions, logo descriptions and text only on images. Why it's important: alt replaces the meaning conveyed by the image and is not a keyword repetition space. Signs of a problem: All images contain the same text or long screenshots and no text description. Directions for improvement: Concisely explain the purpose of the image and provide key information as HTML text. How to verify: Verify that you can understand the context without looking at the image.

Metadata and headings describe the main topic of the page. Rather than simply counting the existence of each element, it checks whether they point to the same question and URL.

What to look for: Unique titles per page, brand name, core topics, duplicate titles and duplicate templates. Why it's important: title is meta information that briefly identifies the main topic of the document. Sign of a problem: Multiple pages use the same title, such as ‘website’ or ‘Posts’, or the brand name is added twice. Where to improve: Create a natural distinction between the key questions and brand on the page. Verification method: Check HTML title and site-wide list of duplicates.

Check for: Summary of page content, brand/service, duplicate/empty values, hyperbole and text matches. Why it's important: The description is meta information that describes the page content, not a ranking formula that guarantees exposure. Sign of a problem: Every page uses the same advertising copy or promises services that aren't actually in the text. Directions for improvement: Specifically summarize the questions and scope the page answers. Verification method: Compare the final meta value with the first screen body.

Check for: self-referencing canonical, other URL·domain, http·https, www, see parameter and sitemap mismatch. Why it matters: canonical signals which representative URL is preferred among overlapping candidates. Sign of a problem: Detail posts point to a list or another post as canonical, and internal links use another URL. Direction for improvement: Unify canonical·sitemap·OG·internal links in line with the actual authentic policy. Verification method: Compare all URL signals with the final HTML link.

Check for: og:title, og:description, og:url, og:type, og:image, og:site_name, absolute path and image. Why it matters: Open Graph is primarily about shared previews and metainformation that helps identify documents. Signs of trouble: Different page title/URL left or image 404. Directions for improvement: Use absolute URLs and representative images that match the current page. Verification method: Verifies the HTML meta and HTTP response of each asset, which in itself is interpreted as not guaranteeing AI citation.

Look for: clear H1s, page topic, semantic match to title, hidden/multiple H1s and titles within images. Why it's important: H1 is the main title that users see on screen. Signs of a problem: H1s are missing or visually hidden, and card titles are haphazardly repeated with H1s. Directions for improvement: Have a screen title that explains your key questions. Verification method: Check the number and contents of H1 visible in the rendered DOM.

Check for: H1→H2→H3 order, design misuse, question-type subheadings, table of contents links and empty headings. Why it's important: Headings explain information relationships and sub-questions in a longer body of text. Signs of a problem: Headings are skipped for font size or titles are repeated without content. Directions for improvement: Place topics and detailed criteria at the same level under representative questions. Verification method: Verify that the document flow is established just by reading the heading.

The Content/Document Relationships group looks at question resolution and connections between pages rather than character count. Quick responses, sufficient evidence, and internal links should support the same user intent.

Check for: key answers in the introduction, repetition of questions, definitions, comparisons, conditions, summary boxes and conclusion. Why it's important: A quick conclusion isn't just one short sentence, it's structured to answer the user's key questions first. Sign of trouble: The background and blurbs are long and the answer is only given at the end. Direction for improvement: Conclusions, conditions, and limitations are presented first and the rationale is explained later. Verification method: Check whether you can understand the answer and scope of the article by just reading the first screen.

Check for: Title and text consistency, key/sub-questions, specific explanations, examples/conditions/limitations, and advertising sentences. Why it's important: If a long character count doesn't address the question, it weakens topic understanding and user value. Sign of a problem: Most posts repeat keywords or promote services that are different from the title. Direction for improvement: Add conditions and grounds necessary for users to make decisions. Verification method: Match the list of questions expected in the title with the actual answer paragraph.

Check for: Comparison tables, step lists, checklists, on-screen FAQs, interpretation of table bottoms, and FAQPage matching with mobile tables. Why it's important: Structural elements should clarify the comparison, order, and relationships of information. Sign of a problem: Tables for keyword repetition are added or the on-screen answer is different from the schema answer. Direction for improvement: Use only the structures necessary for the actual text and add interpretation sentences. Verification method: Check the consistency of mobile operation and screen·JSON-LD questions and answers.

Check for: Core/hub pages, orphan pages, meaningful anchors, broken links, draft/hidden links and context relationships. Why it's important: Internal links describe thematic relationships and representative content between pages. Signs of a problem: Repeating ‘Learn More’ or linking to private/irrelevant pages. Directions for improvement: Specify in the anchor the purpose for which the reader will check next. How to verify: Check link response and visibility, and review page relationships.

The Trust, Operator, and Structured Data groups check whether the facts visible on the screen match the description read by the machine. We don't create individual authors or external channels that don't exist.

Check: author, publisher, datePublished, dateModified, match JSON-LD to screen, see author introduction. Why it's important: You can determine the context of information only when you can see who was responsible for writing and editing it and when. Sign of a problem: Writing random expert names or just updating the date without any actual modification. Direction for improvement: If there is no actual individual, a verifiable organization is indicated as the author/publisher. How to verify: Verify the screen date against the Article JSON-LD values.

Check: Company introduction, official company name, Korean/English brand, representative/contact information, URL, service scope, footer, and external notation. Why it's important: When operator information varies from page to page, it can weaken the link between your brand and official documentation. Sign of a problem: Multiple company names are mixed on the same site or contact information is different. Direction for improvement: Set the original notation as SUMMITFEED for the first mention, and SUMMITFEED for subsequent mentions. Verification method: Compare About, Footer, Home, and External official channels.

Checking for: name·alternateName·url·logo·@id·sameAs in Organization and publisher·inLanguage in WebSite See duplicate output. Why it's important: Global entities complement the website's relationship with the site operator. Signs of a problem: Company information not on screen, unresolved sameAs, outdated logos or a mix of @ids. Directions for improvement: Only use values ​​from actual screens and official URLs. Verification method: Collect JSON-LD for each page and check for overlap/connection of Organization and WebSite.

Check for: headline, author, publisher, date, mainEntityOfPage, image, Breadcrumb, FAQ matches, duplicate schemas and page types. Why it's important: Structured data helps machines understand document and site hierarchies, but it shouldn't be any different than the screen. Sign of a problem: Add a FAQ that isn't on screen or show AboutPage as Article. Directions for improvement: Connect the minimum correct type for the page type. Verification method: Compare parsing and verification tools with screen contents.

Check for: official documents, own experience/original data, source URLs, long direct citations, copying of the same article, duplicates from multiple domains, canonical, demo/real data and advertising notices. Why it matters: E-E-A-T is not a single schema, but rather a state of experience, expertise, external verifiability, and transparency connected across the entire site. Sign of a problem: Negative or identical articles are copied on multiple domains without distinction of source. Directions for improvement: Specify official primary sources and authentic URLs. Verification method: Check the basis for each claim and representative signals of duplicate URLs.

llms.txt, RSS or feed, IndexNow, web app manifest and separate AI data feed are auxiliary elements that can be optionally used after confirming the operational purpose and scope of official support. Applying this file or submission method alone does not guarantee exposure or citations.

If your site isn't a PWA, there's no need to force a manifest, and llms.txt isn't even listed as a required ranking factor. If the required function is not present, first strengthen the consistency of access, representative URL, body, and operating entity among the 25 core features.

In general, you can review in the following order: access/index blocking, representative URLs and redirects, HTML body and rendering, title·canonical·H1, Organization·Article structured data, body that answers key questions, author/source/company information, internal links, duplicate content, and mobile/performance/auxiliary elements. However, the order varies depending on the business-critical pages and the actual extent of the error.

When viewed together with target questions, brand mentions by platform, source citations, and competitor analysis, the website diagnosis results make it easy to prioritize your actual work.

You can check which items to improve first among collection blocking, metadata, JSON-LD, content, and internal links based on the current website status.

1. I think it will be quoted as long as robots.txt Allow is enabled. Since robots deals with scope of access, you need to double check the actual server response, indexability, body and trust information.

2. Consider noindex and robots.txt as the same settings. Since robots deals with crawling and noindex deals with index exclusion, you need to check the file, HTML, and response headers separately.

3. If it appears on the screen, it is assumed to be present in the initial HTML. You need to compare the page source and the DOM after rendering to see when key answers are output.

4. Use the same title and description on all pages. You must create unique meta information based on the questions and body of each page.

5. Link canonical to another page or the wrong host. You should check whether the sitemap, Open Graph, and internal links point to the same representative URL.

6. Enter information not on the screen in JSON-LD. Structured data is not a hidden promotional space, so it should only describe the company, document, and FAQ information that is actually displayed.

7. The screen FAQ and FAQPage answer are different. You must ensure that questions and answers are generated from the same data and that there are no duplicate schemas.

8. Only long articles are published without author or modification date. The actual author/issuer and change history must match the screen and Article data.

9. Copy the same original to an external channel. You must determine the authentic URL and role for each channel, and review canonical·content differences in duplicate documents.

10. Isolate posts without internal links. You need to make sure that relevant hubs and featured articles link to context.

11. To increase your score, start by modifying the items with the least impact. Access blocks and key gaps on important conversion pages should be addressed first.

12. Describe optional elements such as llms.txt as if they were required elements. The scope of official support and operational purpose must be confirmed and distinguished from core technology, content, and trust items.