The AI Visibility Audit: A 12-Point Checklist for Brand Citation Readiness

12 checkpoints separate brands that AI answer engines cite from brands they ignore. That is the core of an AI visibility audit: a structured pass through every signal that determines whether ChatGPT, Perplexity, Google AI Overviews, and Copilot can discover your content, understand who you are, and reference you by name in their answers. Most brands have never run one. The ones that have tend to find at least three critical gaps they did not know existed.

I built this checklist because the question I hear most often from new clients is some version of “are we showing up in AI?” The honest answer is usually no, and the reasons fall into predictable categories. Blocked crawlers. Missing schema. Thin entity signals. Content that reads fine to humans but gives AI models nothing clean to extract. An AI visibility audit finds these gaps before your competitors do.

This article walks through all 12 checkpoints, explains what each one measures, and shows you how to run the audit yourself. If you want context on how this audit fits into a broader maturity framework, the AEO Maturity Model covers the scoring system that sits on top of these checkpoints.

Why you need an AI visibility audit now

AI answer engines are not a future trend. They are a current discovery channel. When a potential customer asks ChatGPT “who is the best [your category] in [your market]?”, your brand either gets named or it does not. There is no second page of results. There is no scroll. You are in the answer or you are invisible.

The problem is that most brands optimized for a search model that assumed 10 blue links. That model is splitting. Google still matters, but ChatGPT, Perplexity, and Copilot are absorbing query volume that used to flow exclusively through traditional search. Your SEO work gives you a head start, but it does not automatically translate to AI citation readiness. The gaps between “ranks well on Google” and “gets cited by AI” are exactly what this audit is designed to find.

Unlike an SEO audit, which has mature tools and established workflows, AI visibility auditing is still new. There is no equivalent of Google Search Console for AI citations. No centralized dashboard that tells you which AI engines mention your brand and how often. That absence makes a structured checklist more important, not less. Without one, you are guessing.

The four audit categories

The 12 checkpoints are organized into four categories. Each category maps to a different failure mode. Some brands fail on access: AI crawlers literally cannot reach their content. Others fail on structure: the content exists but AI models cannot extract clean answers from it. A third group fails on identity: AI models have no way to recognize them as a distinct entity. And a fourth group fails on measurement: they might be getting cited already but have no way to know.

The categories, in order:

1. Technical access (checkpoints 1 through 4). Can AI crawlers reach and render your content?
2. Content structure (checkpoints 5 through 7). Is your content formatted for machine extraction?
3. Entity authority (checkpoints 8 through 10). Does your brand exist as a recognizable entity across the web?
4. Citation tracking (checkpoints 11 and 12). Are you monitoring where and how often AI engines cite you?

A passing score on all 12 does not guarantee you will be cited. But a failing score on any one of them can block citation entirely. That asymmetry is why the audit matters. One misconfigured robots.txt line can erase the value of everything else you have built.

Category 1: Technical access

Technical access is the foundation. If AI crawlers cannot reach your content, nothing else in this audit matters. These four checkpoints confirm that the door is open.

Checkpoint 1: AI crawler access in robots.txt

Open your robots.txt file at yourdomain.com/robots.txt. Search for these user agents: GPTBot, ClaudeBot, PerplexityBot, and Google-Extended. Each one should be explicitly allowed or, at minimum, not blocked.

What to look for:

• An explicit User-agent: GPTBot block with Disallow: entries means OpenAI’s crawler is blocked from those paths. If the disallow is /, your entire site is invisible to ChatGPT’s retrieval system.
• A blanket User-agent: * with Disallow: / blocks everything, including AI crawlers, unless you add specific allow rules below it.
• Missing entries for AI crawlers is fine as long as no blanket block exists. The default is “allowed.”

This is the single most common failure point I see in AI visibility audits. It is also the fastest to fix. A single line change can be the difference between complete invisibility and full access.

Checkpoint 2: Server-side rendering

If your site uses a JavaScript framework like React, Vue, or Angular, your content may only exist after JavaScript executes in a browser. AI crawlers do not always execute JavaScript. Some do. Many do not. The safe assumption is that they see the raw HTML your server sends.

Test this yourself: open your site in Chrome, then open DevTools and disable JavaScript (Settings, then Debugger, then check “Disable JavaScript”). Reload the page. If your content disappears, AI crawlers see an empty page.

The fix is server-side rendering (SSR) or static site generation (SSG). Both ensure that the full content exists in the initial HTML response. If a full migration is not feasible, pre-rendering the pages you most want AI engines to index is a pragmatic middle ground.

Checkpoint 3: Schema markup coverage

Schema markup is how you tell AI models what your content is, who created it, and what entity it belongs to. Run your pages through Google’s Rich Results Test and check for these types:

• Organization schema on your homepage. Name, URL, logo, description, sameAs links to your social profiles and directory listings.
• Article or BlogPosting schema on every content page. Headline, author, publisher, datePublished, dateModified.
• FAQPage schema on pages with FAQ sections. Each question-answer pair marked up individually.
• Person schema for your authors. Name, jobTitle, worksFor, knowsAbout, sameAs.

Schema does not guarantee citations. But it makes your content machine-readable in a way that unstructured HTML does not. AI models parsing your page can distinguish the author from the publisher, the publish date from the modified date, the FAQ from the body copy. That structural clarity gives you an edge over pages where the same information is buried in unmarked paragraphs. For a deeper look at how schema connects to AEO, the schema markup for AEO guide covers the full implementation.

Checkpoint 4: llms.txt file

The llms.txt standard is relatively new. It is a plain text file at your site root (yourdomain.com/llms.txt) that provides AI models with a structured overview of your organization and your most important content.

Think of it as a cover letter for AI crawlers. Instead of forcing them to discover your site structure by following links, llms.txt hands them a map. It typically includes a one-paragraph description of your organization, links to your most important pages, and a summary of what your brand does and what topics you cover.

If you do not have one, add it. The effort is minimal (15 minutes for a basic version), and it removes one more friction point between your content and AI indexing.

Category 2: Content structure

Technical access gets AI crawlers to your content. Content structure determines whether they can extract anything useful once they arrive. These three checkpoints evaluate how well your content is formatted for machine consumption.

Checkpoint 5: Answer-first content structure

Pull up your five most important content pages. Read the first paragraph of each. Does it directly answer the question that someone searching for this topic would ask? Or does it start with context, history, or a broad introduction before getting to the point?

AI models extract the first comprehensive answer they find on a page. If your answer lives in paragraph six, a competitor whose answer lives in paragraph one will get the citation instead. The first 150 words of every page carry disproportionate weight.

Check these specifics:

• Does the first paragraph contain a complete, standalone answer to the primary query?
• Do your H2 headings match the kinds of questions users type into AI engines? “How to run an AI visibility audit” is better than “Our process.”
• Are FAQ sections present with questions that mirror real user queries?
• Are comparison points in tables rather than buried in prose?

Content reformatting is often the highest-return activity in the entire audit. You do not need to write new content. You need to restructure what you already have so AI models can parse it.

Checkpoint 6: Content freshness and depth

Check the publish date and last-modified date on your top 10 pages. Anything older than 12 months in a fast-moving category is at risk. AI models penalize stale content, especially when a fresher source covers the same topic.

Depth matters too. Open a competitor’s page for the same topic and compare coverage. If they address 12 sub-topics and you address 5, the gap is obvious to any AI model evaluating source authority. Depth is not about word count. It is about completeness. A 2,000-word article that covers every angle of a topic outranks a 5,000-word article that pads three angles with filler.

Also check for original insights. Does your content include data, case studies, or frameworks that exist nowhere else? Primary sources get cited. Summaries of other people’s work get skipped. Proprietary frameworks are especially powerful here. When you name a concept and define it (like the AEO Maturity Model), AI models associate that concept with your brand.

Checkpoint 7: Author and publisher signals

AI models evaluate source credibility. Part of that evaluation is who wrote the content and who published it. Check for:

• Named authors on every content page. “By Brendan Hunt” carries more weight than “By Admin” or no attribution at all.
• Author bios with verifiable credentials. A two-sentence bio linking to a LinkedIn profile and listing relevant experience gives AI models a trust signal that an anonymous byline does not.
• Person schema connecting the author to their professional profiles and the publishing organization.
• Consistent authorship across your content. If the same expert writes about the same topic area repeatedly, that reinforces topical authority for both the person and the organization.

Publisher signals matter too. Your Organization schema should connect your content to your brand entity. Every Article schema should reference the publisher. These connections create a graph that AI models can follow from a single blog post all the way up to your organization’s identity.

Category 3: Entity authority

Entity authority determines whether AI models recognize your brand as a distinct, identifiable thing. You can have perfect technical access and perfectly structured content, but if AI models do not know who you are, they have no reason to cite you by name. These three checkpoints evaluate the strength of your entity signals.

Checkpoint 8: The ChatGPT brand test

This is the simplest and most revealing test in the entire audit. Open ChatGPT and type: “What is [your brand name]?” Then: “What does [your brand name] do?”

Three outcomes are possible:

• Accurate response. ChatGPT knows your brand, describes it correctly, and mentions your key products or services. This means you have meaningful entity presence. Your entity signals are reaching AI training data or retrieval sources.
• Vague or partially accurate response. ChatGPT has heard of you but confuses details, merges you with a similarly named brand, or gives a generic description. Your entity exists but is weak or fragmented.
• No response or fabricated details. ChatGPT does not recognize your brand at all, or it makes up information. This means your entity footprint is too thin for AI models to work with.

Run the same test in Perplexity and Copilot. Perplexity is especially useful because it shows source citations inline. If Perplexity cites your website when answering a question about your category, that is a strong signal. If it cites your competitors instead, you know exactly who you need to outperform.

Checkpoint 9: Knowledge graph and directory presence

Search your brand name on Google and look for a Knowledge Panel on the right side of the results page. If one exists, your brand has been recognized as an entity in Google’s Knowledge Graph, which directly feeds Google AI Overviews.

Then check these sources:

• Wikidata. Search for your brand at wikidata.org. A Wikidata entry is one of the strongest entity signals available. Even brands that are not notable enough for a Wikipedia article can often establish a Wikidata item.
• Crunchbase. Relevant for technology and B2B companies. A claimed, verified Crunchbase profile feeds multiple AI training pipelines.
• Industry directories. Whatever directories are relevant to your category (Clutch for agencies, Avvo for lawyers, Healthgrades for physicians), check that your listing exists, is claimed, and has consistent information.
• LinkedIn company page. A complete, active LinkedIn company page is a strong entity signal for B2B brands.

The common thread is consistency. Your brand name, description, URL, and category should match across every directory and profile. Inconsistencies fracture your entity signal. If Google sees “AEO Hunt” on your website, “AEO Hunt LLC” on Crunchbase, and “AEOHunt” on LinkedIn, it takes longer for AI models to recognize these as the same entity.

Checkpoint 10: Third-party mention footprint

Search your brand name on Google News. Search it on Reddit. Search it on YouTube. How often is your brand mentioned on sites you do not own?

Third-party mentions are one of the strongest signals AI models use to validate entity authority. Mentions on authoritative sites (industry publications, news outlets, educational institutions) carry more weight than mentions on low-authority sites. But volume matters too. A brand mentioned across 50 independent sources has a stronger entity signal than a brand mentioned on 5, even if the 5 are individually more authoritative.

What counts as a meaningful mention:

• A named reference in a news article, blog post, or industry report on a site you do not control.
• A podcast appearance where the host names your brand and your URL appears in the show notes.
• A guest post on an authoritative site with an author bio linking back to your brand.
• A citation in a research paper, case study, or industry benchmark published by a third party.

If your third-party mention footprint is thin, that becomes a priority action item. Building entity authority through mentions is slower than fixing a robots.txt file, but it is often the gap that separates brands that occasionally get cited from brands that consistently get cited.

Category 4: Citation tracking

The first three categories are about making your brand citable. This final category is about measuring whether AI engines actually cite you, and building the infrastructure to track changes over time. Without measurement, you cannot tell whether your AEO work is producing results.

Checkpoint 11: Cross-engine citation test

Pick your 10 most important target queries. These should be the questions that potential customers ask when they are considering your product or service category. Then query each one in ChatGPT, Perplexity, Google AI Overviews, and Copilot.

For each query on each engine, record:

• Were you cited? Yes or no.
• If yes, which page was referenced?
• Were any competitors cited instead?
• Was the citation accurate? Did the AI engine describe your brand correctly?

This produces a 10-by-4 grid that gives you a snapshot of your current AI visibility. Suppose a brand runs this test and finds citations in Perplexity for 3 of 10 queries, zero citations in ChatGPT, and one mention in Google AI Overviews. That tells a clear story: there is some visibility in Perplexity (which uses real-time web retrieval), limited visibility in Google’s AI layer, and no presence in ChatGPT’s training data or retrieval system. Each gap points to a different action item.

If you want to formalize this measurement as an ongoing metric, the Share of AI Voice framework gives you a formula for converting raw citation counts into a trackable percentage. It also shows you how to benchmark your citation rate against competitors on a query-by-query basis.

Checkpoint 12: Ongoing citation tracking setup

A one-time audit gives you a snapshot. Ongoing tracking gives you a trend line. Set up a recurring process (monthly at minimum, weekly if resources allow) to re-run your target queries across all four AI engines and record the results.

What to track over time:

• Citation frequency. For each target query, are you being cited more often, less often, or the same as last period?
• Citation accuracy. Are the AI-generated descriptions of your brand getting more accurate or drifting?
• New query appearances. Are you being cited for queries you did not previously appear in?
• Competitor movement. Are competitors gaining or losing citations on your target queries?
• Source page shifts. When AI engines cite you, which pages are they referencing? Is it always the same page, or are different pages being pulled for different queries?

The AI citation tracking guide covers the full methodology for building a repeatable tracking workflow, including how to structure your query set, how to normalize results across engines with different citation formats, and how to build a dashboard that surfaces changes without requiring you to manually check every query every week.

Running the audit: step by step

Here is the sequence I recommend. The order matters because later checkpoints depend on earlier ones passing.

Week one: Technical access (checkpoints 1 through 4). Start here because everything else is pointless if AI crawlers cannot reach your content. Review robots.txt. Test server-side rendering. Validate schema markup. Check for llms.txt. Fix any blockers immediately. These are typically quick changes that do not require content creation or strategic decisions.

Week two: Content structure (checkpoints 5 through 7). With access confirmed, evaluate your content. Audit answer-first structure on your top 10 pages. Check freshness dates. Verify author and publisher signals. Build a list of pages that need restructuring, and prioritize by traffic and strategic importance.

Week three: Entity authority (checkpoints 8 through 10). Run the ChatGPT brand test. Audit your Knowledge Panel, Wikidata, and directory presence. Map your third-party mention footprint. Entity gaps take longer to close than technical gaps, so the earlier you identify them, the sooner you can start building.

Week four: Citation tracking (checkpoints 11 and 12). Run the cross-engine citation test on your target queries. Set up your ongoing tracking process. This gives you the baseline that all future measurement will be compared against.

Four weeks, twelve checkpoints, one clear picture of your AI visibility.

What the results look like

After running the full audit, brands typically fall into one of four profiles.

The invisible brand

Fails on multiple checkpoints across all four categories. AI crawlers are blocked. Content is not structured for extraction. No entity presence in knowledge graphs. No citation tracking in place. This is where most brands start. The gap is wide, but that also means the upside from fixing foundational issues is large.

The technically ready brand

Passes technical access (categories 1 and 2) but fails on entity authority and citation tracking. The content is there. The schema is there. But AI models do not recognize the brand as a distinct entity, so they have no reason to cite it by name. The fix is entity building: directories, third-party mentions, Knowledge Panel, Wikidata.

The authority brand with broken plumbing

Strong entity presence, recognizable brand, third-party mentions across the web. But a restrictive robots.txt blocks AI crawlers, or JavaScript rendering hides the content, or schema markup is missing. The brand has earned the right to be cited, but a technical barrier prevents it. These are the most frustrating gaps because they are the easiest to fix, and the longest to go unnoticed.

The citation-ready brand

Passes all 12 checkpoints. AI crawlers can reach the content. Content is structured for extraction. Entity signals are strong. Citation tracking is in place. This brand is positioned to be cited and has the measurement infrastructure to know when it is working. The next step is not more auditing. It is ongoing optimization: expanding topic coverage, building deeper authority, and monitoring citation trends to defend and grow their position. The AEO Maturity Model provides the scoring framework for tracking that ongoing progress.

Common audit findings

After running this audit across dozens of brands, patterns repeat themselves. Here are the most common findings and what they mean.

Blocked AI crawlers hiding in plain sight

The single most common issue. A developer added a broad disallow rule years ago, or a CMS plugin set a restrictive robots.txt by default, and nobody checked. It takes two minutes to open robots.txt and search for GPTBot. It is the highest-impact two minutes in any AI visibility project.

Schema markup exists but covers only the homepage

Many brands implemented Organization schema when they first launched their site and never expanded it. Their homepage has clean structured data. Their blog posts have none. Their team page has no Person schema. Their FAQ page has questions and answers in the HTML but no FAQPage schema wrapping them. Schema coverage needs to be site-wide, not page-one only.

Content depth is competitive but structure is weak

The brand has genuinely good content. Long, well-researched guides. Real expertise. But the content is formatted as continuous prose with vague headings. No FAQ sections. No comparison tables. No definition callouts. The information is there, but it is locked inside a format that AI models struggle to parse efficiently. Reformatting this content (without rewriting it) often produces the biggest short-term improvement in AI visibility.

Zero third-party mentions in the past 12 months

Entity authority does not build itself. If nobody outside your organization has mentioned your brand in the past year, AI models have very little independent signal to work with. Building a third-party mention strategy (guest posts, podcast appearances, press outreach, industry directory listings) is a medium-term investment, but it is the one that separates brands with occasional citations from brands with consistent ones.

No baseline citation data

The brand has never queried ChatGPT, Perplexity, or Google AI Overviews for their target terms. They do not know whether they are being cited, which pages are being referenced, or how their citation rate compares to competitors. Without a baseline, you cannot measure progress. The cross-engine citation test (checkpoint 11) produces that baseline in a single afternoon.

After the audit: prioritizing your action items

Every audit produces a list of gaps. Trying to fix all of them at once is a recipe for slow progress on everything and completion on nothing. Prioritize based on two factors: impact and effort.

High impact, low effort (do first):

• Fix robots.txt to allow AI crawlers.
• Add llms.txt to your site root.
• Add FAQPage schema to pages that already have FAQ content.
• Rewrite the first paragraph of your top 5 pages to lead with a direct answer.

High impact, medium effort (do next):

• Implement Article and Person schema across all content pages.
• Restructure your top 10 pages with comparison tables, numbered lists, and definition boxes.
• Claim and optimize your Crunchbase, LinkedIn, and industry directory profiles.
• Set up monthly citation tracking across all four AI engines.

High impact, high effort (plan for):

• Build third-party mention presence through guest posts, press, and speaking.
• Create a Wikidata entry for your organization.
• Develop proprietary frameworks and original research that make you a primary source.
• Migrate JavaScript-rendered content to server-side rendering.

The first group can be done in a single week. The second takes a focused sprint of two to four weeks. The third is ongoing work measured in months. But the compound effect of all three groups working together is what moves a brand from invisible to consistently cited.

The audit as a recurring practice

An AI visibility audit is not a one-time exercise. AI models update their training data on different schedules. Retrieval-augmented generation systems (like Perplexity) pull fresh content in real time. Google AI Overviews draw from an index that changes daily. Your competitive landscape shifts as other brands start doing AEO work of their own.

Run the full 12-point audit quarterly. Between full audits, maintain your monthly citation tracking cadence. Treat the audit results as a living document, updating your scores as you close gaps and as the AI search landscape evolves.

The brands that treat AI visibility as an ongoing practice, measured and optimized the same way they measure paid media or organic search, are the ones that will own the answers in their category. The brands that run one audit and forget about it will watch their competitors fill the gap.

Your AI visibility audit starts with checkpoint 1. Open your robots.txt. Everything else follows from there.