The GEO Audit Checklist: 12 Categories for AI Search Readiness

Why Audit Before Optimizing

In short

Optimizing without auditing wastes effort on the wrong fixes. An audit reveals which of the 12 GEO categories are weakest, so the team works on highest-leverage problems first instead of guessing.

Most teams jump straight to writing FAQ blocks or adding schema. They skip the audit step.

The result is predictable. Effort goes into categories that were already fine, while the real bottleneck — often robots.txt, missing entity signals, or stale freshness data — stays unsolved.

A GEO audit forces a structured baseline. Every category gets scored. Every gap gets ranked by impact versus effort. The team starts with whatever moves the most citations per hour of work invested.

There is also a strategic reason. AI search is a citation game, not a click game. The SparkToro 2024 study found roughly 60% of Google searches end without a click. If your brand is not the source being cited, you are invisible — and an audit is the fastest way to find out where you stand.

The 12 Audit Categories

In short

Every GEO audit must cover 12 categories: schema, entity signals, llms.txt, robots.txt, citation-rich content, quick-answer boxes, FAQ structure, author schema, freshness, inbound citations, brand monitoring, and knowledge panel presence.

The 12 categories below come from triangulating three sources. The Aggarwal et al. (2024) GEO paper. Schema.org documentation. And recurring patterns we observe across 100+ Nordic enterprise implementations.

Each category is independently testable. Each maps to a specific signal that AI retrieval systems consume.

How Alice Labs Runs a GEO Readiness Audit

In short

The Alice Labs GEO Readiness Audit is a proprietary 30-point checklist that expands the 12 categories into specific, testable subchecks — delivered in roughly 1-2 days with a scorecard and prioritized backlog.

The Alice Labs audit framework breaks each of the 12 categories into 2-3 concrete subchecks. That gives a 30-point checklist that two consultants can complete in a working day for a typical enterprise site.

Phase one is data collection. We crawl the site, pull every JSON-LD block, fetch robots.txt and llms.txt, and run a 20-prompt audit across ChatGPT, Perplexity, and Claude.

Phase two is scoring. Each of the 30 subchecks gets a 0-3 score with evidence attached. We benchmark against the Alice Labs LLMO Citation Benchmark — our internal dataset covering 100 SaaS brands across the Nordic region.

Phase three is the backlog. We rank every gap by impact (citation potential) versus effort (engineering hours). The output is a sequenced 90-day roadmap with owners.

The framework is deliberately bias-corrected. We cap how much weight any single category can carry, because in our experience teams over-invest in schema and under-invest in entity signals and off-site citations.

Common Audit Findings on Enterprise Sites

In short

Most enterprise sites pass on basic Article schema but fail on llms.txt, author markup, entity sameAs links, and off-site citation flow. These are the four highest-leverage gaps to fix first.

We see a recurring pattern across enterprise audits. The findings cluster into four buckets.

Bucket one: missing llms.txt. Most enterprise sites have no llms.txt file. The Answer.AI standard launched September 2024 and adoption is still early. Publishing one takes under an hour.

Bucket two: weak entity signals. Schema.org Organization markup often exists but lacks the sameAs property linking to Wikipedia, Wikidata, Crunchbase, and LinkedIn. NAP data is inconsistent across the site, Google Business Profile, and directory listings.

Bucket three: incomplete author schema. Articles list a byline but do not emit Person schema with credentials or reviewer markup. This weakens E-E-A-T — Google's quality framework covering Experience, Expertise, Authoritativeness, and Trustworthiness.

Bucket four: no off-site citation flow. Brands rarely show up in Tier-1 publications, Wikipedia references, or authoritative directories. Off-site citations are the slowest-moving signal, but they compound — and they are heavily weighted by ChatGPT, Perplexity, and Claude.

Building the Prioritized Fix Backlog

In short

A prioritized backlog ranks every audit gap by impact versus effort, sequences quick wins first, and assigns owners with review cadences — so the team executes the right fixes in the right order.

The audit scorecard is the input. The fix backlog is the output. Without sequencing, even a perfect audit becomes a wishlist that nobody executes.

We rank fixes on a 2x2 matrix: impact (low-high citation potential) and effort (low-high engineering hours).

• Quick wins (high impact, low effort). Publish llms.txt. Fix robots.txt for AI crawlers. Add sameAs to Organization schema. Add datePublished/dateModified everywhere. Ship in week 1.
• Medium projects (high impact, medium effort). Roll out FAQPage schema across pillar pages. Add Person schema with reviewer markup. Add quick-answer boxes and entity definitions to top 20 pages. Ship in weeks 2-6.
• Long horizon (high impact, high effort). Build off-site citation flow. Earn Wikipedia references. Pursue Knowledge Panel claim and accuracy. Ship over months 3-12.
• Skip or defer (low impact). Most schema cosmetic tweaks, micro-format experiments, and exotic markup types. Revisit only after the high-impact backlog is clear.

Assign one owner per fix. Set a 30-day review cadence. Re-run the audit at 90 days to confirm the score actually improved. Without re-measurement, the audit becomes a one-off artifact instead of a feedback loop.