What a GEO audit checks — and what it does not
A GEO audit is a structured self-review of whether your website can be reached by AI search systems, understood in content terms, and cited as a source. That means systems like ChatGPT Search, Perplexity, Google AI Overviews and Microsoft Copilot — not the classic blue link list on Google.
Important upfront: an audit is neither a substitute for classic SEO nor legal advice. It does not guarantee AI citations — measurement is still fragmented in 2026. What it does: you find blockers (blocked crawlers, missing schema, unclear passages) and prioritize levers before investing in content or budget.
As an AI SEO agency, we use this checklist internally before every major content release. It builds on our foundation articles on GEO, AI search and AI Overviews — here the focus is the operational checklist.
GEO audit vs. classic SEO audit
| Aspect | SEO audit | GEO audit |
|---|---|---|
| Primary goal | Rankings & clicks in Google/Bing | Citability in AI answers |
| Success metric | Position, impressions, CTR (GSC) | Prompt tests, referral traffic, brand mentions |
| Technical focus | Crawlability, Core Web Vitals, index | + AI crawlers, schema, citable passages |
| Content focus | Keywords, E-E-A-T | + FAQ structure, 40–80 word answer blocks, tables |
| Off-page | Backlinks | + consistent entity, directories, mentions |
GEO builds on SEO — without indexable, fast pages there is no AI visibility. GEO alone with neglected technical SEO optimizes into a void.
Block A: Technical (checks 1–5)
Check 1 — robots.txt: do not blanket-block AI crawlers
AI systems use their own user agents. Relevant ones include GPTBot and OAI-SearchBot (OpenAI), ClaudeBot and Claude-SearchBot (Anthropic), PerplexityBot (Perplexity) and Google-Extended (training for Google AI features, not identical to Googlebot for classic search).
Review your robots.txt: does Disallow: / for User-agent: * block everything — including AI bots? Are GPTBot or PerplexityBot explicitly excluded? Training crawlers and search/retrieval crawlers can be controlled separately: to be cited in ChatGPT Search you must at least allow retrieval bots.
Pass: no blanket block of the named agents on money pages and blog.
Fail: GPTBot/PerplexityBot/OAI-SearchBot with Disallow: / — you are invisible on those channels.
Check 2 — Indexability of money pages
Every transactional page (SEO, web agency, industry hubs) must return HTTP 200, carry no noindex, and appear in the sitemap. Soft 404 (200 with “page not found” content) is a GEO killer — AI systems may index the junk but will not cite it.
Check 3 — Schema: Organization, WebSite, Service, FAQPage
Check with Google Rich Results Test or schema validators: Organization/LocalBusiness sitewide, FAQPage on pages with visible FAQ, Service on service pages. Duplicated, contradictory schema confuses parsers — one consistent set per page.
Check 4 — Server-side HTML (SSR/SSG)
AI crawlers are not guaranteed to be JavaScript-capable like Googlebot. Next.js Server Components and prerendered HTML are a structural advantage — pure client-only SPAs risk empty first fetches.
Check 5 — Core Web Vitals & crawl budget
Slow pages are less often fully captured. LCP under 2.5s, clean sitemap, no millions of parameter URLs. GEO benefits from the same technical hygiene as SEO.
Block B: Content (checks 6–10)
Check 6 — H2/H3 as answerable questions
Every heading should carry a concrete user question: “How do I connect Propstack to my website?” instead of “Integration”. AI systems match prompts to headings and the first sentences beneath them.
Check 7 — Answer blocks in 40–80 words
Directly under H2/H3: a standalone core statement citable without context. Each section should start that way.
Check 8 — FAQ sections with real value
At least five FAQs on money pages; questions from customer enquiries, not keyword stuffing. FAQPage JSON-LD only when the FAQ is visible in HTML — otherwise rich results risk.
Check 9 — Lists, tables, numbered checklists
Wall-of-text paragraphs are extracted less often than tables and numbered lists. Comparisons (SEO vs. GEO, CRM options) as tables.
Check 10 — As-of date & cannibalization
“As of June 2026” in intro or excerpt. One owner URL per search intent — two pages for “SEO agency Hamburg” dilute signals (see hub-spoke logic on your SEO service page).
Block C: Entity & off-page (checks 11–13)
Check 11 — NAP & Google Business Profile
Name, address, phone identical on website, GBP and legal notice. Exact, consistent company name — for AI systems entity clarity is a trust signal.
Check 12 — Brand mentions (no invented awards)
Directory listings, real press, guest posts. No invented seals or rankings. Brand mentions correlate more strongly with AI citation probability than isolated long-tail backlinks — but quality over quantity.
Check 13 — llms.txt honestly assessed
llms.txt is optional. It describes your site for LLMs but does not control crawler access (robots.txt does). As of mid 2026 there is no solid evidence that Google or ChatGPT use llms.txt for answer generation. Treat it as preparation — not a ranking lever.
Block D: Measurement (checks 14–15)
Check 14 — Prompt test matrix (weekly)
Define 10–20 typical questions from your audience — neutrally worded, e.g. “What does SEO cost for mid-sized companies?” or “How do you connect a broker website to Propstack?” Ask the same prompts weekly in ChatGPT, Perplexity and Google (AI Overview if triggered). Document: was your domain cited? Which passage? Which competitors?
Check 15 — Referral traffic in GA4
Check referrals from chatgpt.com, perplexity.ai and copilot.microsoft.com. The share is often small but growing — and sessions are frequently high-intent.
Prompt test template (copy-paste)
Columns for a sheet: Date | Prompt | ChatGPT cited? | URL | Perplexity cited? | Google AIO? | Notes. Example prompts for agencies:
- “How do I optimize a website for ChatGPT and Perplexity?”
- “What is the difference between SEO and GEO?”
- “What robots.txt settings do I need for AI crawlers?”
- “How do I build FAQ schema for AI Overviews?”
Avoid prompts like “Best SEO agency Hamburg” — they do not yield reproducible audits and raise promotional expectations.
Crawler strategy: training vs. search
OpenAI, Anthropic and Perplexity increasingly distinguish crawlers for model training and live search. Blocking training only (e.g. GPTBot) can still allow appearance in ChatGPT Search — if OAI-SearchBot is allowed. Decide consciously; clarify privacy and IP questions with legal counsel.
Example snippet (illustration only — verify before going live):
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /30-day plan after the audit
- Week 1: robots.txt + indexation check of all money pages
- Week 2: sharpen FAQ + FAQPage on two key service pages
- Week 3: one new spoke article with table + FAQ (internal links)
- Week 4: prompt retest, document referral baseline
What a GEO audit cannot do
- No guarantee of citations or rankings
- No substitute for GSC data on classic visibility
- No legal advice on crawler blocking or GDPR
- Tool metrics (AI visibility scores) are aids — not truth
Frequently asked questions
What is the difference between an SEO audit and a GEO audit?
An SEO audit checks rankings, indexation and on-page for Google. A GEO audit adds AI crawlers, citable passages, entity signals and prompt tests for ChatGPT, Perplexity and AI Overviews. Both together make sense in 2026.
Is llms.txt enough for AI visibility?
No. llms.txt does not replace robots.txt or good content. It can describe your site but does not prove measurable citation lift. Priority: technical indexability, schema, FAQ, clear answer blocks.
Do I have to allow GPTBot in robots.txt?
Not strictly for search visibility. For citations in ChatGPT Search, OAI-SearchBot is more relevant. Training (GPTBot) and retrieval are separable — align strategy with privacy.
How do I measure GEO success without expensive tools?
Weekly prompt matrix, GA4 referrals, GSC for AI overview impressions where available, plus qualitative notes. Specialist tools can help but are optional.
Can an agency run the GEO audit?
Yes — typically as a starting point before a content push or relaunch. We deliver prioritized findings, no success guarantee.
Conclusion
GEO in 2026 is not magic but disciplined work: crawlers reachable, content citable, entity consistent, measurement honest. These 15 checks give you clarity in an afternoon — before you spend budget on content that never arrives technically.
Free GEO assessment. We review whether your website meets the technical and content prerequisites for AI visibility — concretely, without performance promises. More about our work as an AI SEO agency in Hamburg. Request a no-obligation assessment.
Note: This article is a professional checklist, not legal advice. Clarify crawler policies and data protection with your legal counsel. As of June 2026.
