KNOWLEDGE BASE · GUIDE
The 40-Point GEO Audit Checklist
A GEO audit checks 40 things across seven points: whether AI answer engines can fetch, read, verify and trust your site well enough to cite it. This page lists the seven point groups, how scoring works, and what a ranked fix list looks like.
The seven point groups
- Crawler access (checks 1-6): GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User, Google-Extended, Applebot-Extended — allowed on public pages, private pages disallowed. If a bot is blocked, you do not exist in that engine's answers (/pillar/become-the-answer, live 11 Oct 2026).
- Plain HTML without JavaScript (checks 7-12): facts in raw HTML — headings, text, tables; script-only pages read as empty shells.
- Answer-first structure (checks 13-19): each page answers its question in the first two sentences; key claims not buried in PDFs.
- Schema and machine-readable facts (checks 20-26): Product + Offer with real prices, Organization, Person; engines rarely quote what they must reconstruct.
- Discovery files (checks 27-31): llms.txt present, dated, current; sitemap canonical and clean.
- Entity clarity (checks 32-36): one consistent description of the company everywhere — site, directories, profiles, press.
- Off-site corroboration (checks 37-40): third-party pages repeating your exact description; one credible mention beats ten self-published claims.
How the score works
| Band | Meaning | Typical state |
|---|---|---|
| 0-39 | Effectively invisible | Blocked crawlers, no schema, no llms.txt |
| 40-69 | Partially readable | Some facts extractable, gaps in corroboration |
| 70+ | Citable | Consistent facts, access open, corroboration building |
Our worked sample (fictional firm, method illustration): score 42 of 100 — GPTBot and PerplexityBot disallowed in robots.txt, no llms.txt, key claims buried in PDFs, no entity paragraph. Day-30 target after the sprint: 70+ (skillpilotadvisory.ai/geo, fetched 10 Oct 2026).
Sources and context
- GEO spend is estimated at about US$1.2 billion in 2026, up from about US$848 million in 2025 (modelled, industry estimates, skillpilotadvisory.ai/geo, fetched 10 Oct 2026).
- Around 55% of marketers report active GEO budgets (directional, vendor survey 2026, same page).
- Search behaviour is shifting: buyers increasingly ask ChatGPT, Gemini and Perplexity who to hire — the engines compose answers from sources they can read, verify and trust (/pillar/become-the-answer, live 11 Oct 2026).
- Malaysia's National Guidelines on AI Governance and Ethics (MOSTI, Nov 2024) push transparency and explainability — the same properties GEO structuring delivers.
What the audit cannot do
Force a citation. It removes the reasons engines skip you; engines choose what to cite. Any firm that guarantees citations is lying (skillpilotadvisory.ai/geo FAQ, fetched 10 Oct 2026). Scale rechecks fixed items at day 30 so the score moves on the record; founder-reviewed within 48 hours of intake.
Run the checklist free (self-serve version)
Ask your site: can a bot fetch my main pages without JavaScript? Do product pages state price and currency in schema? Does robots.txt allow the AI crawlers? Two "no" answers = start the fix list today.
FAQ
Is a GEO audit the same as an SEO audit? No. SEO wins a ranked link; GEO wins a mention inside a composed answer. The mechanics overlap; GEO adds crawler access, JS-free rendering and llms.txt gates.
How long do fixes take? Days for the structural fixes; engine re-crawls and third-party corroboration set the pace after that.
What do I get? A scored audit, a ranked fix list, a copy-paste fix pack (llms.txt, schema, restructures) and a day-30 recheck.
Do you need site access? No — the audit reads public pages only.