My Approach to AI Visibility Audits with citelayer®

As of June 2026. An AI visibility audit isn’t a ChatGPT test prompt. If you ask, „Recommend providers for X,“ and your brand doesn’t show up, that’s a clue—but not a definitive finding. If your brand does show up, that’s also just a clue—not a victory.

This is exactly where the real work begins. AI Visibility It doesn't just measure whether a brand is mentioned anywhere. It asks: Is it understood correctly? What sources are shaping this perception? Which competitors are recommended instead? Which technical signals are helpful, and which are just smoke and mirrors? And which gaps can you truly close without flooding the web with flimsy AI filler?

The Summary

• An AI Visibility Audit is an analysis, not a prophecy. It examines how AI-powered response systems currently categorize a brand, product, or service.
• A single prompt isn't enough. The platform, model, timing, wording of the question, and available sources can all influence the results.
• Measurements are taken at several levels: Actual mentions, sources, competitors, brand and entity clarity, content, technology, and priorities.
• Technical signals are important, but they are not the only deciding factor. Schema, llms.txt, Markdown, robots.txt, and crawler access are only helpful if the content itself is clear, up-to-date, and verifiable.
• A score is a map, not the landscape itself. It helps with prioritization. It is not a substitute for expert assessment.
• This is especially handy for WordPress: Many of these issues can be specifically addressed: key pages, internal links, structured data, authors, categories, outdated content, and machine-readable formats.

My take: A good audit doesn't make you dependent on a tool. It reveals which assumptions are currently valid, which aren't, and what the next logical steps should be.

What is an AI Visibility Audit?

An AI Visibility Audit examines whether and how a brand appears in AI response systems. In this context, „visible“ means more than just “the name appears somewhere.” Visible can mean that the brand is mentioned, correctly described, cited as a source, compared to competitors, recommended, or simply overlooked.

It’s the combination that matters. An audit looks not only at the surface-level results but also at the underlying causes: website structure, brand and product clarity, external references, structured data, access policies for crawlers, search data, content gaps, and competitor presence. Only then does a clear picture emerge that you can work with.

The difference from a traditional SEO audit isn’t that SEO suddenly doesn’t matter. On the contrary. Google itself states that SEO fundamentals remain relevant for generative AI features in Google Search. But AI Visibility raises additional questions: What sources does the system use? What answer is generated without a click? Which competitors are mentioned even though you’re ranking? And how clear is the entity behind your name?

What an Audit Is Not

An audit is not a screenshot from ChatGPT with the headline „We Are Invisible.“ An audit is not a tool score without sources, a time frame, or platform details. And an audit is not a list of 50 recommendations where everything sounds equally urgent.

AI responses are dynamic. They depend on the query, language, location, model, time of retrieval, available sources, and sometimes even the interface. That’s why an audit must clearly document what was examined. Otherwise, a snapshot becomes a judgment that it cannot possibly support.

A good audit also distinguishes between findings, interpretations, and actions. „The brand was not mentioned in 8 out of 20 prompts“ is a finding. „The brand is poorly defined as an entity“ is a possible interpretation. „Consistently update the About Us page, product pages, external profiles, and Schema Graph“ is an action item. This distinction may sound dry, but it prevents every metric from immediately triggering knee-jerk reactions.

The Five Questions I Ask First

In my work with citelayer® audits and products, five opening questions have proven effective. They are intentionally simple. Precisely for that reason, they quickly reveal whether we are dealing with a genuine visibility problem, a measurement problem, or a positioning problem.

1. What should the brand be known for? Are you looking for a category, a specific product, a local service, a problem, a recommendation, or a technical question?
2. Who or what is the brand? A person, company, product, plugin, method, store, studio, or multiple entities that need to be properly linked?
3. What sources currently describe this brand? Company website, Search Console, structured data, third-party profiles, reviews, public code repositories, marketplaces, press coverage, videos, industry directories.
4. Which competitors are emerging instead? Not only on Google, but also on ChatGPT, Perplexity, Claude, Gemini, and Google AI.
5. Which gap is the most relevant from a business perspective? Is there a missing definition, reference, comparison page, technical approval, central offer page, or clear answer to questions that influence purchasing decisions?

This is where AI Visibility becomes pleasantly unromantic. Not every missing mention is a big deal. Not every mention is valuable. And not every technical optimization has the same impact.

Why One Platform Isn't Enough

ChatGPT, Claude, Perplexity, Gemini, and Google AI Search do not function identically. They have different user interfaces, retrieval logic, crawlers, source preferences, security rules, and response styles. A brand may appear prominently in Perplexity, be absent from ChatGPT, and appear in Google AI Overviews only as a source, without being recommended.

That’s exactly why I break down platform responses. A combined metric can be useful if you want to identify trends. For the actual diagnosis, however, it’s important to know where the problem occurs. Is the brand missing everywhere? If so, that suggests a fundamental problem. Is it missing only on one platform? If so, we need to take a closer look at that platform’s traffic, sources, citation behavior, and response logic.

Google has also shifted its focus on this topic: Since June 2026, Google has been rolling out its own Search Console reports for generative AI features. This is important because it means that AI visibility is no longer discussed solely in terms of manual prompt tests and third-party tools. But the same rule applies here: Search Console shows data from Google—not automatically from ChatGPT, Claude, or Perplexity.

Brand and Entity: Is it clear who is being referred to?

At AI Visibility, „brand“ isn’t primarily a design term. It’s a classification term. A system must be able to recognize that a name belongs to a specific person, company, website, service, or product. If this classification falters, everything that follows falters as well.

This applies especially to small businesses, studios, medical practices, local service providers, SaaS products, and WordPress plugins. People often understand context through conversations, recommendations, or experience. AI systems need public, accessible, and consistent signals. Is the product name always spelled the same way? Is there a central website? Do the organization, founder, brand, domain, social media profiles, and structured data all align? Are old names still in use? Is there external evidence?

An audit isn't looking for beauty here, but for clarity. When multiple entities overlap, a system may have to guess. And machines sometimes guess with remarkable self-assurance.

Sources, Citations, and Competitors

An AI response is rarely just a reflection of your website. Especially when it comes to recommendation questions, third-party sources matter: comparison lists, reviews, forums, marketplaces, GitHub, YouTube, press coverage, industry directories, documentation, and other websites that rank you or your competitors.

That's why it's not enough to simply count how many times your URL has been cited. A source can be cited and still strengthen the competition. In the article on Comparison Lists and Listicles in AI Search I described exactly this risk: Self-promoting lists can, under certain circumstances, provide an AI with a neatly sorted list of competitors without the brand itself coming out on top.

In the audit, therefore, three questions are of interest at the same time: Which sources are used? Which brands are mentioned? And what role does your brand play in the response: source, option, recommendation, side note, or not mentioned at all?

Technology: Schema, llms.txt, Crawler, Canonicals

Technology is important. But it's not the magic solution it's often made out to be. Schema Data can clarify entities and relationships. llms.txt This can be useful for systems and agents that use such files. Markdown output can make content easier to read. robots.txt can control crawling. Canonical tags can reduce conflicting signals. For WordPress, I’ve broken down this technical aspect separately: Making WordPress More Readable for AI Systems.

However, Google explicitly states for AI Overviews and AI Mode that no special additional machine-readable files are required to appear in these Google features. At the same time, this doesn’t ignore the practices of other systems, agents, and technical retrieval methods. It’s precisely this tension that needs to be clearly explained; otherwise, you’ll end up either concluding that „it’s all nonsense“ or that „install file X and you’ll become visible.“ Both approaches are too simplistic.

In my methodology, technical signals therefore serve two purposes: Some are factored into the evaluation if they demonstrably improve clarity, structure, or accessibility. Others serve as warning or diagnostic indicators. For example, the presence of an llms.txt file is not proof of visibility. However, the absence of an access policy for AI crawlers can still be a practical problem if search bots cannot access important public content.

Contents: Fan-Out, Decision Questions, and Gaps

AI Visibility rarely fails simply because of a missing file. Very often, the answers needed to generate a recommendation in the first place are missing. This is exactly where Query Fan-Out As a conceptual model: A question is broken down into subquestions. Systems search not only for the main keyword, but also for definitions, risks, alternatives, prices, evidence, target audiences, limitations, and next steps.

An audit therefore checks whether your website answers the relevant decision-making questions. Not all of them. Just the relevant ones. The difference is important. If you blindly work through every possible sub-question, you’ll quickly end up with a jumble of content. If you clearly answer the real decision-making questions, you’ll build substance.

For WordPress, this means specifically: pillar articles, clusters, FAQs, product pages, offer pages, categories, and internal links must all work together. A good page explains not only „what it is,“ but also „who it’s useful for,“ „what its limitations are,“ „how to recognize quality,“ and „why I should trust this source.“.

A Short WordPress Mini-Audit Checklist

If you don't want to do a full analysis, you can start with a small sample. This isn't a substitute for a thorough audit, but it will quickly show you where the problems lie.

1. Define three key questions: What would you like to be known for? For example, „Best WordPress Agency for WooCommerce,“ „SEO Plugin for Small Online Stores,“ or „AI Visibility Audit for B2B.“.
2. Try out several platforms: Ask similar questions in ChatGPT, Perplexity, Claude, Gemini, and Google AI Search. Record the date, language, exact question, and result.
3. Don't just write down "Yes" or "No": Is your brand mentioned? Where in the text? In which sentence? Which competitors are mentioned? Which sources are cited?
4. Check your main entity page: Is there a page that clearly explains the brand, offerings, target audience, location, people, products, and supporting documentation?
5. Check structured data: Do "Organization," "Person," "Article," "Product," "FAQ," or "LocalBusiness" match the visible content?
6. Check indexing and snippets: Are important pages indexable, not accidentally marked as "noindex," and generally available in Google Search with a snippet?
7. Check out robots.txt and AI crawlers: Are you blocking search bots even though you want to be visible? Do you separate search, training, and user-triggered requests? The basics on this topic are covered in the article on AI crawlers, robots.txt, and content signals.
8. Search for content gaps: Which questions does the competition answer more effectively? Where are comparisons, evidence, price ranges, target audiences, limitations, or examples missing?
9. Rate third-party sources: Are there profiles, reviews, mentions, documentation, repositories, or industry sources that independently validate your brand?
10. Prioritize strictly: What is the next step that will provide the greatest benefit: improving the main page, cleaning up the schema, adding clusters, updating old content, or building external references?

The great thing about this list is that it brings AI Visibility out of the fog. Suddenly, it’s no longer about „GEO-Vibes,“ but about very specific questions, pages, and signals.

How the citelayer® Plugin and Audit Work Together

citelayer® for WordPress bridges this very gap between traditional SEO plugins and AI Visibility. The plugin makes WordPress content more machine-readable: Schema.org, llms.txt, Markdown output, UCP Discovery, WebMCP, and compatibility with SEO plugins such as Yoast, Rank Math, or All in One SEO (AIOSEO). That’s the technical foundation.

The citelayer® AI Visibility Audit is the diagnostic level above that. It asks not only whether the technology is in place, but also whether the brand is understood, mentioned, cited, or overlooked in relevant AI response scenarios. This includes platform testing, competitor analysis, source analysis, structured data, content architecture, crawler policy, and a prioritized roadmap.

This distinction is important to me because it’s honest: A plugin can improve technical signals. An audit can identify and prioritize gaps. Neither of these replaces the actual work on branding, content, and documentation. Unfortunately. It would be more convenient. I’d take it too. But that’s not the reality.

Limitations: What No Audit Can Honestly Promise

A reputable audit does not promise a guaranteed mention in ChatGPT, Claude, Perplexity, Gemini, or Google AI. No one can definitively control which source a model will select tomorrow. Anyone who claims otherwise is selling reassurance rather than analysis.

What an audit can do: It can document the current state of affairs, identify patterns, uncover technical and editorial obstacles, highlight gaps in the competition, and prioritize actions. It can also prevent you from spending a lot of time on superficial distractions while the actual brand remains unclear.

That’s why I view scores as a decision-making tool—not as a judgment on a brand’s value, not as a trophy, and certainly not as a substitute for common sense. A score is useful when it explains why it turned out the way it did and what action should be taken as a result.

FAQ

Do I need an AI Visibility Audit if my SEO metrics are good?

Maybe. Good SEO metrics are an advantage, but they don't automatically tell you whether AI systems are correctly naming, recommending, or citing your brand as a source. An audit can be particularly worthwhile when it comes to recommendations, comparisons, and decision-making.

Is it enough to test my brand in ChatGPT?

No. That's a good starting point, but it's not an audit. You need multiple questions, multiple platforms, documented results, an analysis of sources, and an assessment of the causes.

Is llms.txt a ranking factor?

For Google Search, Google explicitly states that llms.txt is not required and is ignored by Google Search. For other systems, agents, and machine-readable workflows, a well-maintained llms.txt file can still be useful. That’s why I don’t view it as a magic solution, but rather as a technical signal within its context.

Which is more important: structure or content?

Both serve different purposes. Content provides the substance. Schema helps to structure that substance more clearly. If the content is thin, schema won't suddenly make it useful. If the content is good, a well-designed schema can make relationships and entities clearer.

Can a small business even build AI visibility?

Yes. Not necessarily for broad, generic terms, but certainly for specific niches, local inquiries, special offers, product categories, and decision-making questions. Small providers, in particular, often have real-world experience but too few publicly available, well-structured examples.

Sources and Verification

This classification is based on my citelayer® audit and product work, as well as on publicly available primary sources. I use my own analyses to provide a technical classification; publicly available factual claims can be verified through the following sources.

• Google Search Central: AI Features and Your Website.
• Google Search Central: Optimizing Your Website for Generative AI Features on Google Search.
• Google Search Central Blog: Introducing Search Generative AI performance reports in Search Console.
• Google Search Central: Introduction to robots.txt.
• OpenAI Developers: Overview of OpenAI Crawlers.
• Anthropic Privacy Center: Does Anthropic crawl data from the web?.
• Perplexity Docs: Perplexity Crawlers.
• citelayer®: AI Visibility Plugin for WordPress.
• citelayer®: AI Visibility Audit.
• Our own citelayer® audit and product work—including platform tests, source analysis, entity verification, technical WordPress signals, query fan-out gaps, and prioritization—is incorporated into this classification as a practical methodology.