How to Run a 5-Layer AEO/GEO Audit: Score Your Site's AI Visibility in Under an Hour

Most businesses know AI engines are citing brands, but almost none know where they stand. They’ve heard terms like “generative engine optimization” and “AI citations,” but they don’t have a framework to measure their current position — let alone improve it.

That’s what the 5-layer AEO/GEO audit solves. It takes a website from “I think we’re probably fine” to a concrete score (0-100) with prioritized fixes, all in under an hour.

This article walks through the exact methodology StayCitable uses for client engagements. It’s the same framework that produces the scored reports we deliver to vacation rental property managers, ecommerce brands, and SaaS companies — adapted here so you can run it yourself. For a worked example of this framework applied to a real domain, see our 5-Layer AEO/GEO Self-Audit of StayCitable.

The 5 Layers (and Why Each One Matters)

AI engines don’t cite websites the way search engines rank them. Google’s organic algorithm evaluates hundreds of ranking signals. AI engines evaluate a much smaller set: can I access the content, is it structured so I can extract answers, is the entity clear, and are other engines already citing it?

The 5-layer framework maps directly to how AI crawlers and language models consume your site:

Layer	What It Measures	Weight
1. Technical Foundation	Can AI crawlers access and parse your content?	20%
2. Content Structure	Is your content organized so AI can extract answers?	25%
3. Entity Optimization	Does your site clearly declare who and what you are?	20%
4. AI Visibility	Are AI engines actually citing you right now?	25%
5. Competitive Position	How do you compare to competitors on AI citations?	10%

The weights aren’t arbitrary. Content structure and AI visibility carry the most weight because they’re the two factors that most directly determine whether an AI engine cites you. Technical foundation is the gatekeeper — if crawlers can’t access your content, the other layers don’t matter. Entity optimization bridges the gap between having good content and getting attributed for it. Competitive position contextualizes your score.

Layer 1: Technical Foundation (20 points)

This is binary in practice: either your site is crawlable and parseable by AI systems, or it’s invisible. Check these six items:

Schema Markup (5 points): Open your homepage. View source. Search for application/ld+json. You should find at least one JSON-LD block, and it should parse as valid JSON (copy it into a JSON validator). The block should declare your organization type — Organization, LocalBusiness, or an industry-specific type like LodgingBusiness. Zero JSON-LD blocks = 0 points. One valid block with entity type = 3 points. Multiple blocks covering organization + website + breadcrumbs = 5 points.

llms.txt (4 points): Visit https://yoursite.com/llms.txt. This file tells LLM crawlers which pages they should read. If it returns 404: 0 points. If it returns HTML disguised as text (common on WordPress sites): 1 point — you have the file but it’s useless. If it returns valid plain text with URLs: 3-4 points depending on whether entries include descriptions.

robots.txt (3 points): Visit /robots.txt. It should not disallow any LLM-specific crawlers (Google-Extended, GPTBot, Claude-Web, PerplexityBot). A clean Disallow: line targeting AI crawlers costs you points. Score: fully permissive = 3 points, partial blocks = 1-2 points, aggressive blocking = 0 points.

Sitemap (3 points): Visit /sitemap.xml or /sitemap-index.xml. It should return valid XML with recent lastmod dates. Stale sitemaps (lastmod > 30 days ago) cost 1 point. Missing sitemap = 0 points.

HTTPS (3 points): Your site must serve over HTTPS. This is table stakes — AI crawlers typically won’t index non-HTTPS content. HTTP-only = 0 points. HTTPS with mixed content warnings = 2 points.

Mobile Responsiveness (2 points): Open your site on a phone-width viewport. If content is cut off, buttons overlap, or text requires horizontal scrolling, deduct points. AI engines increasingly use mobile-first crawling.

Layer 2: Content Structure (25 points)

This layer answers: can an LLM extract clear, attributed answers from your pages?

Heading Hierarchy (5 points): Every page should have exactly one H1. H2s should organize major sections. H3s should nest under parent H2s. Run a heading audit: document.querySelectorAll('h1,h2,h3,h4') in browser console. Missing H1 = 0 points. H1 present but H3s without H2 parents = deduct. Clean hierarchy = 5 points.

Answer Positioning (8 points): For your top 5 target queries, find the answer on your site. Is it above the fold? Is it in the first 200 words? Is it in a dedicated section with a descriptive heading? AI engines extract from the first substantive content they encounter. Answers buried below ads, hero images, or long introductions earn lower scores.

FAQ Sections (5 points): Does your site have FAQ sections with structured question-answer pairs? The format matters: H2: "Frequently Asked Questions" followed by bolded questions and paragraph answers. This pattern is directly extractable by AI crawlers. Zero FAQ sections = 0 points. One FAQ section with proper formatting = 3 points. FAQ sections on multiple pages = 5 points.

Readability (4 points): Grade-level analysis (Flesch-Kincaid, 8th-10th grade target). Long paragraphs (>4 sentences), jargon-heavy language, and wall-of-text layouts all hurt extractability. Quick check: scan a page. Can you identify the answer to “what does this page do” in under 10 seconds? If not, an AI crawler can’t either.

E-E-A-T Signals (3 points): Experience, Expertise, Authoritativeness, Trustworthiness. Author bylines with bios, publication dates, citations to primary sources, and privacy/terms pages all signal credibility. AI engines favor content from verifiable sources.

Layer 3: Entity Optimization (20 points)

Entity optimization determines whether AI engines attribute content to your brand or to “an unnamed source.”

JSON-LD Entity Types (8 points): Beyond presence (Layer 1), evaluate quality. Does your Organization schema include sameAs links to Wikipedia, Wikidata, social profiles? Does it include address, logo, url? A bare Organization block with no properties = 2 points. Fully enriched with knowledge graph connections = 8 points.

Knowledge Graph Presence (6 points): Search your brand name in Google. Is there a Knowledge Panel on the right side? Do you have a Wikidata entry (wikidata.org/wiki/Q...)? Google’s Knowledge Graph and Wikidata feed entity understanding to every major AI system. No Knowledge Panel and no Wikidata = 0 points. Knowledge Panel present = 3 points. Both Knowledge Panel and Wikidata entry = 6 points.

Entity Linking (6 points): Do your pages link to authoritative external sources? Do they use consistent brand naming (no “Acme Inc” on one page and “Acme Incorporated” on another)? Internal linking to related content creates entity relationships. Strong internal link graph with consistent naming = 6 points.

Layer 4: AI Visibility (25 points)

This is the layer that surprises most businesses. You can have perfect technical foundations, great content, and clean entities — and still get zero AI citations if you haven’t appeared in the training data or real-time search indices.

Manual AI Engine Testing (15 points): Open ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude, and Meta AI. For each engine, ask a question your content should answer. For example: “what’s the best [your product category] for [your audience]?” Record whether your brand appears, whether it appears with a link, and whether the description is accurate.

Scoring: cited in 5-6 engines = 15 points, 3-4 engines = 10 points, 1-2 engines = 5 points, zero engines = 0 points.

Citation Quality (5 points): When your brand appears, is it a named citation (“According to Brand X…”) or a vague reference? Does it include a link? Are the facts correct? Named citations with links = 5 points. Unnamed references = 2 points. Incorrect facts = 0 points (this is worse than not being cited — it means the AI has learned something wrong about you).

Citation Consistency (5 points): Test the same query across multiple sessions and days. Does your brand appear consistently, or is it hit-or-miss? Consistent appearance across sessions = 5 points. Appears sometimes = 2 points. Never appears = 0 points.

Layer 5: Competitive Position (10 points)

Your score only matters relative to competitors. A 70/100 is excellent if competitors score 40, but poor if competitors score 85.

Competitor Citation Audit (4 points): Identify your top 3 competitors (the ones ranking for your target queries in Google). Run the Layer 4 AI engine tests for them. How many engines cite them? If competitors appear in more engines than you do, deduct points.

Competitor Technical Audit (3 points): Quick-check competitors’ schema markup, llms.txt, and sitemaps. If competitors have strong technical foundations and you don’t, the gap is widening.

Market-Specific AI Presence (3 points): For your industry, are AI engines answering category-level queries at all? Test: “best [your industry] companies” or “top [your category] providers.” If AI engines return generic listicles with no named companies, the entire category is under-optimized — this is an opportunity. If they already name specific competitors, the window is closing.

Scoring Your Site

Add up the points from each layer. The maximum is 100.

Score Range	What It Means
80-100	AI-ready. You’re being cited and your infrastructure supports it. Maintain and monitor.
60-79	Good foundation, gaps in 1-2 layers. Target the lowest-scoring layer first.
40-59	Significant gaps across multiple layers. Start with technical foundation — it gates everything else.
20-39	Mostly invisible to AI engines. Technical and content issues need immediate attention.
0-19	Effectively invisible. AI crawlers cannot access, parse, or attribute your content.

What Happens After the Audit

The audit is diagnosis, not treatment. Once you have your score and layer breakdown:

1. Fix Layer 1 first. If crawlers can’t access your content, nothing else matters. Add JSON-LD, create llms.txt, verify robots.txt. For step-by-step implementation guidance, see our Structured Data for AI Citation Guide and the AI Citation Readiness Checklist.

2. Restructure your top 3 pages for Layer 2. Add FAQ sections, ensure answer-first content structure, fix heading hierarchies.

3. Enrich your entity layer. Get a Wikidata entry. Add sameAs links to your schema. Claim knowledge panels.

4. Test AI visibility monthly. Set calendar reminders. AI citation landscapes shift fast — a new engine release or training update can change your visibility overnight.

5. Re-run the full audit quarterly. Scores should trend upward. If they plateau, the remaining gaps are likely competitive (everyone in your space has caught up on basics).

How StayCitable Uses This Framework

This isn’t a theoretical model. It’s the exact methodology we use for every client engagement. The 5-layer audit produces a scored report that becomes the strategic roadmap for the next 90 days.

For vacation rental property managers in the Pacific Northwest, we run this audit and typically find scores in the 20-40 range — good businesses with strong local reputations that are effectively invisible to the AI engines their guests are using to plan trips. The most common pattern: solid content (Layer 2) with zero technical foundation (Layer 1) and zero entity optimization (Layer 3). The content exists, but AI engines can’t find it and don’t know who wrote it.

We also offer a free 5-business-day citation audit that applies this framework to a single domain. It’s the same methodology, just scoped to produce a lightweight deliverable. If you want to see your own score without running the audit yourself, that’s the fastest path.

FAQ

How long does a 5-layer audit take?

For a single domain with standard architecture: 45-60 minutes. Multi-domain or enterprise sites with hundreds of pages: 2-3 hours, mostly spent on Layer 2 (content structure) sampling across the site.

Do I need technical expertise to run this?

Layers 1, 4, and 5 require only a browser and the ability to view page source. Layer 2 requires reading your own content critically. Layer 3 is the most technical — JSON-LD validation and Wikidata editing have a learning curve, but both are well-documented.

How often should I re-audit?

Quarterly at minimum. Monthly if you’re in a competitive industry (ecommerce, SaaS, travel). The AI citation landscape shifts fast — quarterly is the minimum safe cadence.

What’s the most common mistake?

Auditing once and filing the report. An audit is a snapshot. The value comes from re-auditing after fixes and measuring improvement. The second audit (showing a score increase from 35 to 58) is more valuable than the first — it proves the methodology works and identifies the next layer to tackle.

My score is low but my Google rankings are fine. Does this matter?

Yes, and this is the most common objection we hear. Google rankings measure traditional search visibility. AI citations measure a different channel — the one where travelers ask ChatGPT “plan my trip” instead of searching “best hotels Portland.” Both channels matter, but the AI channel is growing while traditional search is shrinking. A site ranking #1 on Google but invisible to AI engines is winning yesterday’s game.