GEO Audit
GEO Audit Checklist 2026: 25 Checks to See If Your Page Gets Cited by AI
27 March 2026
8 min read
GEO Analysis Tool
Most pages that rank on Google will never be cited by ChatGPT, Perplexity, or Google AI Overviews. The signals that matter for AI citation are fundamentally different from traditional SEO. This checklist covers the 25 checks that actually determine whether AI models use your content as a source.
Traditional SEO audits check things like keyword density, meta descriptions, and backlink profiles. A GEO audit checks something different: can an AI model understand this page, trust it, and extract a specific, attributable claim from it?
The difference matters because AI engines don't rank pages. They synthesise answers from multiple sources, selecting content based on how easily it can be parsed, how credibly it signals expertise, and how specifically it answers the question being asked. A page that ranks number one on Google can score zero in AI citations if it's full of vague qualitative claims with no citable data.
This checklist is built on the Princeton and IIT Delhi GEO research (Aggarwal et al., 2024), which tested specific content interventions across thousands of queries and measured their impact on AI citation rates. The top-performing signals are the foundation of every check below.
The five categories that determine AI citation
Before running through the checks, understand the framework. AI citation likelihood is determined by five factors, each contributing roughly equally to whether your page gets used as a source.
| Category | What it measures | Impact on citation |
|---|
| Citation Hooks | Statistics, quotable claims, original data | Very high |
| Content Structure | Headings, FAQs, passage clarity | High |
| Content Clarity | Authoritative tone, specific claims | High |
| Technical Access | Schema markup, crawler permissions | Medium |
| Entity and Authority | Author credentials, E-E-A-T signals | Medium |
The Princeton research found that adding statistics to a page improved AI citation rates by up to 40% - the single most impactful intervention tested. The checks below are ordered by impact within each category.
The checklist
1
Citation Hooks
Statistics present: Does the page contain at least one specific number, percentage, or measurable data point that an AI could extract as a citable fact? Vague claims like "significantly improves" are uncitable. "Improves by 23%" is citable.
Original data or research: Does the page cite or present original data, a study, or a survey result? Pages that are primary sources get cited far more than pages that summarise secondary sources.
Quotations from named sources: Are there direct quotes from named individuals, organisations, or studies? Attributed quotes are one of the strongest citation signals the Princeton research identified.
Outbound citations to authoritative sources: Does the page link to peer-reviewed research, government data, or established industry sources? Paradoxically, citing others increases how much AI models cite you - it signals that your content is research-backed rather than opinion.
Specific, verifiable claims: Are the page's main claims falsifiable and specific enough that a fact-checker could verify them? "This method works well" fails this check. "This method reduced load time by 1.4 seconds in our testing" passes it.
2
Content Structure
FAQ section with self-contained answers: Does the page include a FAQ section where each answer makes sense without requiring the reader to have read the rest of the article? AI models extract FAQ answers and serve them independently. If an answer requires context from elsewhere on the page, it won't get cited.
Heading hierarchy is logical: Does the page use H1, H2, H3 correctly with one clear topic per section? AI models use heading structure to understand what each passage is about before deciding whether to cite it.
Direct answers lead each section: Does each section open with the most important information first rather than building to a conclusion? AI models pull the first substantive sentence of a section as a candidate citation. Burying the key point at the end of a paragraph means it gets missed.
Comparison tables present where relevant: If the page compares options, products, or approaches, is there a structured table? Tabular data is highly retrievable by AI engines and rarely appears in pages that rely purely on prose.
FAQ schema implemented: Beyond having FAQ content, is the page's FAQ section marked up with JSON-LD FAQ schema? Schema makes the Q&A structure explicit to AI crawlers rather than requiring them to infer it from the HTML.
Passage-level clarity: Can individual paragraphs be extracted and understood without surrounding context? AI models break pages into passages and evaluate each one. A passage that requires three other paragraphs to make sense will never be cited.
3
Content Clarity
Authoritative tone throughout: Does the writing make claims confidently rather than hedging everything? "This might help" signals uncertainty. "This reduces X by Y" signals expertise. AI models prefer the latter.
Technical terms used correctly: When technical terms appear, are they used precisely and defined on first use? The Princeton research found that appropriate use of technical terminology improved AI citation rates - it signals domain expertise.
Qualitative claims are supported: Every subjective claim ("this is the best approach") should be followed by a specific reason. Unsupported assertions are the most common reason pages get paraphrased without attribution rather than cited directly.
Fluent, readable prose: Is the writing clean and easy to follow? This matters more for AI citation than most people expect. Awkward sentence construction reduces how much an AI model can confidently extract and use.
4
Technical Access
AI crawlers not blocked in robots.txt: Check your robots.txt file to confirm GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot, and Google-Extended are not blocked. Many sites accidentally block AI crawlers through overly broad disallow rules.
Article or webpage schema implemented: Is the page marked up with Article, BlogPosting, or WebPage schema that includes author, datePublished, and dateModified? This helps AI engines understand the page's type, recency, and authorship.
Page loads under 3 seconds: Verify with Google PageSpeed Insights. Slow pages are partially crawled or skipped by AI bots with strict timeout limits.
Content is server-side rendered: Is the main content rendered in the initial HTML response rather than loaded by JavaScript? AI crawlers often don't execute JavaScript, meaning dynamically loaded content is invisible to them.
5
Entity and Authority
Named author with credentials: Does the page have a clearly identified author with a brief statement of relevant expertise? Anonymous content from a generic "team" signals lower authority than content from a named person with verifiable credentials.
First-hand experience signals: Does the author mention personal testing, direct experience, or specific involvement with the topic? The E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) framework rewards first-hand experience signals highly.
Publication and update dates visible: Are dates clearly shown? AI models weight recent content more heavily for topics that evolve over time. A page last updated in 2022 competing with a 2026 page will lose on recency.
Entity clarity - what is this page about?: Can an AI model definitively understand the topic, scope, and purpose of the page from its structure alone without reading every word? Entity ambiguity is one of the most common reasons pages get ignored despite strong content.
How to score your page
Go through each of the 25 checks and mark them pass or fail. A rough scoring guide:
| Score | Interpretation | Priority action |
|---|
| 20 or more | Strong AI citation readiness | Monitor and maintain |
| 13 to 19 | Developing - competitive topics need improvement | Focus on Citation Hooks first |
| 8 to 12 | Weak - unlikely to be cited for competitive queries | Structural and citation overhaul |
| Below 8 | Not citation-ready | Comprehensive rewrite required |
The highest-leverage improvements, based on the Princeton research, are in Citation Hooks. Adding three specific statistics and two outbound citations to a page that currently has none can move it from the "weak" range to "developing" in a single revision.
What makes a GEO audit different from an SEO audit
The distinction is worth being precise about because the two overlap significantly but are not the same thing.
A traditional SEO audit checks whether search engines can find, crawl, and rank your content. It focuses on technical health, keyword optimisation, and backlink signals. A GEO audit checks whether AI engines can extract specific, attributable information from your content and trust it enough to cite. The difference in focus produces different recommendations.
SEO gets you ranked. GEO gets you cited. Both matter, but they require different optimisations.
A page can rank in position one on Google and score near zero on this GEO checklist. That's common for pages that are well-optimised for keywords but built on qualitative assertions, generic advice, and no original data. Conversely, a page with detailed original research, specific data points, and clear attribution can score highly on this checklist even before it has significant backlinks or traditional SEO authority.
The practical implication is that GEO optimisation is accessible to smaller sites and newer domains in a way that traditional SEO often isn't. You don't need domain authority to be cited by AI. You need citable content.
Frequently asked questions
How long does it take for GEO improvements to show results?
AI models update their knowledge and citation patterns continuously. Pages with strong citation signals can start appearing in AI-generated answers within days of being crawled. Unlike traditional SEO where ranking changes can take months, AI citation patterns respond faster to content changes - though this also means they can change without warning as AI models update.
Do I need to optimise for each AI engine separately?
The core signals - statistics, clear structure, authoritative tone, schema markup, and named expertise - work across ChatGPT, Perplexity, Claude, and Google AI Overviews. There are platform-specific differences in how each engine weights recency and source diversity, but the foundational signals are consistent enough that a single optimisation effort improves performance across all platforms.
Is GEO the same as AEO (Answer Engine Optimisation)?
The terms overlap significantly and are often used interchangeably. GEO (Generative Engine Optimisation) was coined by researchers at Princeton and IIT Delhi in 2024 to describe optimisation for AI-powered generative search engines specifically. AEO is a broader term that predates generative AI and includes optimisation for featured snippets and voice search. In practice both refer to making content extractable and citable by AI-powered answer systems.
What is the single most impactful GEO improvement?
According to the Princeton GEO research, adding specific statistics to a page produced the largest measurable improvement in AI citation rates - up to 40% in controlled testing. If your page has only one GEO improvement to make, replacing qualitative claims with quantified, sourced data points is the place to start.
How do I know if my page is currently being cited by AI engines?
The most reliable method is manual testing - run 10 to 15 queries related to your page's topic in ChatGPT, Perplexity, and Google, and check whether your page or site is mentioned or linked. Dedicated GEO monitoring tools like Otterly.ai and Profound automate this at scale, but manual testing gives you an immediate baseline at no cost.
Run a free GEO audit on any page
Paste any URL and get a scored audit across all five categories - Citation Hooks, Content Structure, Content Clarity, Technical Access, and Entity and Authority. Free score in under 60 seconds.
Get your free GEO score →