Why a GEO audit became essential in 2026
In early 2025, fewer than 15% of Google queries in France triggered an AI answer at the top of the results (source: Google SERP analysis, January 2025). By May 2026, that figure exceeds 40% on high-volume informational and commercial queries. The shift is structural, not cyclical.
I've guided a number of small and mid-sized businesses through this diagnostic in 2025 and early 2026. The pattern is consistent: sites that are technically sound in SEO terms, ranking well for their main keywords, yet completely invisible in ChatGPT on queries like "recommend a provider for X near me." The reason is always the same: content that is too generic, with no proof of expertise, no citable sources, and no structure that lets an LLM build a reliable answer from their pages.
That is exactly what a GEO audit detects — and what a classic SEO audit never sees.
SEO vs GEO: what the audit doesn't cover the same way
A classic SEO audit examines your site from the angle of Google's crawler: technical structure, speed, indexing, internal linking, domain authority, SERP rankings. It's a machine-to-machine diagnostic.
A GEO audit takes a different angle: it focuses on what language models can learn and synthesize from your content. The question isn't "does Google index this page correctly?" but "can an LLM build a reliable, sourced answer about your expertise from what you publish?"
Key point: ChatGPT and Google AI Overviews don't read your site in real time. They rely on a combination of historical training data and, for models with real-time browsing (ChatGPT with Bing, Perplexity), a third-party index. A GEO strategy targets both channels: feeding future training data AND being citable in real-time searches.
The two audits share common dimensions — baseline technical quality, content structure, structured data — but diverge on several critical points:
| Dimension | Classic SEO audit | GEO audit |
|---|---|---|
| Objective | Ranking in Google SERPs | Citability in AI answers |
| Primary metric | Average position, GSC impressions | Citation rate on target queries |
| Content | Keyword density, length, freshness | Citable structure, direct answers, named sources |
| Authority | Domain Rating, backlink profile | Demonstrated E-E-A-T, media mentions, proof of experience |
| Structured data | Schema.org for rich snippets | Schema.org for entity signal + FAQPage + HowTo |
| Result testing | Rank tracker (Ahrefs, Semrush) | Manual ChatGPT test + AI Overviews scraping |
For an in-depth look at how the two approaches fit together, see our complete guide to Generative Engine Optimization.
The 7 dimensions of a complete GEO audit
Here is the structure we use at Cicero Studio for our client GEO audits. Each dimension is scored from 0 to 10. The overall score is the weighted average using the weights shown.
This is the most decisive dimension. An LLM cites content when it can extract a direct, precise and sourced answer to a question from it. Positive signals are: direct answers at the start of a section, the use of figures backed by a named source, an explicit Q&A structure (FAQ, definition boxes), and the absence of undefined jargon.
- Check: can each H2 stand alone as the answer to a question?
- Check: do factual claims have a named source (not "according to a study")?
- Check: is the content reusable as-is in an AI answer without distortion?
- Check: the density of FAQ and HowTo markup on high-potential pages?
Experience, Expertise, Authoritativeness, Trustworthiness. LLMs assign an implicit weight to a source's reliability. A site with no identified author, no proof of hands-on experience and no visible trust signals will consistently be cited less than a competitor carrying the same information but with stronger credibility anchoring.
- Author pages with a full bio, photo, and LinkedIn links
- Case studies with verifiable, quantified results
- Certifications, accreditations, training mentioned with a source
- Mentions in third-party media (press, podcasts, conferences)
- Schema.org Person and Organization correctly filled in
AI engines actively use Schema.org markup to identify entities, understand relationships and structure answers. A GEO audit checks not only the presence of the markup but its quality and its consistency with the visible content.
- Article / NewsArticle with datePublished, author, publisher
- FAQPage on every page with Q&A content
- HowTo on tutorial / methodology pages
- Organization with sameAs (LinkedIn, Wikidata where relevant)
- BreadcrumbList consistent with the URL structure
- Product / Service on commercial pages with a detailed description and offer
LLMs associate a source with a topic when it covers that subject deeply and consistently. A site with 3 scattered blog posts on different subjects will be perceived as a generalist. A site with 20 articles organized into semantic clusters around 2-3 themes will be perceived as a reference on those themes.
- Map existing articles onto a topical coverage matrix
- Identify gaps: sub-topics with no dedicated article
- Check the internal linking between pillars and spokes
- Assess the semantic consistency of the vocabulary per cluster
GEO citability isn't only on-site. LLMs draw on the whole indexed web corpus. Mentions in recognized media, editorial backlinks, well-maintained profiles on specialized directories — all of these signals feed how AI engines perceive a brand's authority.
- Presence and quality of profiles on Google Business, LinkedIn, Trustpilot
- Mentions in professional press and sector media
- Participation in podcasts, webinars, industry conferences
- Citations in other sites' articles with a link or a named mention
- NAP consistency (Name, Address, Phone) across every directory
LLMs struggle to cite a brand whose positioning they don't clearly understand. If your site says "innovative digital solutions" without ever spelling out "we do X for Y in sector Z," the model has nothing to cite. Clear positioning is a prerequisite for citability.
- An explicit definition of the offer right on the homepage
- A consistent message across every page (no contradictions between "agency" and "software")
- An "About" page with story, team, mission — written to be citable
- A glossary or definitions of the company's own domain terms
This dimension measures your ability to track your GEO visibility over time. An audit with no baseline can't measure progress. It includes: defining a panel of 10-15 target queries, the initial test in ChatGPT and Google AI Overviews, and setting up a monthly monitoring system.
- A defined panel of target queries (recommendation, comparison, best-of queries)
- A documented initial test with screenshots or logs
- A baseline citation score calculated (% of queries that mention the brand)
- A formalized monthly re-test protocol
Request a free GEO audit of your site. We analyze the 7 dimensions and deliver a full report with a score and prioritized recommendations.
Get my free articleFree articleThe GEO scorecard: how to use it
The scorecard below is the simplified version of our internal tool. It's designed to be usable by an SEO practitioner in 2-3 hours on a mid-sized site.
For a deeper look at how scoring fits a broader visibility plan, see our 2026 GEO strategy framework.
| Dimension | Score /10 | Weight | Weighted score | Action priority |
|---|---|---|---|---|
| 1. Content citability | _ / 10 | 25% | _ × 0.25 | High |
| 2. E-E-A-T signals | _ / 10 | 20% | _ × 0.20 | High |
| 3. Structured data | _ / 10 | 15% | _ × 0.15 | Medium |
| 4. Semantic coverage | _ / 10 | 15% | _ × 0.15 | Medium |
| 5. External mentions | _ / 10 | 12% | _ × 0.12 | Medium |
| 6. Brand consistency | _ / 10 | 8% | _ × 0.08 | Secondary |
| 7. Testability / baseline | _ / 10 | 5% | _ × 0.05 | Secondary |
| Overall GEO Score | _ / 10 | |||
Interpreting the scores
- 8-10: Strong position. Maintenance and marginal optimization are enough. Focus on extending topical coverage.
- 6-7.9: Untapped potential. Targeted actions on the weak dimensions can produce citability gains in 2-3 months.
- 4-5.9: Significant vulnerability. A rework is needed on dimensions 1 and 2. Competitors hitting 7+ are already capturing the citations you're missing.
- Below 4: Near-total invisibility in GEO. A structured 6-month rebuild plan is needed.
How to test your visibility in ChatGPT and Google AI Overviews
The direct test is the benchmark measure. No third-party tool replaces the manual test for now — existing GEO tracking tools measure proxies, not actual citation.
ChatGPT test protocol
Use an account with no conversation history (or a private session) to avoid personalization bias. Test with the GPT-4o model and with GPT-4o browsing enabled separately.
Query types to test (adapt to your sector):
- Direct recommendation: "What are the best [your sector] agencies?"
- Comparison: "[Your competitor] vs [your brand]: what are the differences?"
- Geographic best-of: "Recommend a [your activity] provider near me"
- Problem-solution: "I need [your solution], where do I start?"
- "Who does X" query: "Who offers [your offer] for SMEs?"
For each query, note: (1) is your brand cited? (2) in what context? (3) are competitors cited in your place?
Google AI Overviews test protocol
Google AI Overviews (formerly SGE) appears inconsistently — not every user sees it on every query. Test from a "clean" Google account (no search history) to get a neutral measure.
Queries to test: the same as ChatGPT plus the informational queries in your sector where you have blog articles. Document each visible AIO with a screenshot.
Pitfall to avoid: don't test your own branded keywords ("Cicero Studio SEO agency"). Test the generic queries in your sector — that's where GEO citability makes the difference. Branded queries answer a navigational intent, not a recommendation intent.
Prioritizing actions after the audit
A GEO audit rarely surfaces a single weak point. Most sites show gaps across 3-4 dimensions at once. Prioritization is therefore critical to avoid scattering your efforts.
Our rule: impact × effort × time-to-result. Here's the simplified matrix:
| Action | GEO impact | Effort | Time to result | Priority |
|---|---|---|---|---|
| Add direct answers to existing articles | High | Low | 2-4 weeks | P0 |
| Complete FAQPage markup on key pages | High | Low | 2-6 weeks | P0 |
| Create or enrich the author / about page | High | Medium | 4-8 weeks | P1 |
| Produce articles for the cluster's missing angles | High | High | 6-12 weeks | P1 |
| Earn mentions in sector press | Medium | High | 8-16 weeks | P2 |
| Set up monthly monitoring | Low | Low | Immediate | P0 |
Our six-pillar GEO strategy framework details the end-to-end process, from data collection to report delivery.
GEO audit frequency and cadence
A GEO audit is not a one-off action. AI engines evolve fast — models are updated, source-selection algorithms change, and competitors' practices improve. A brand that reached good GEO visibility in January 2026 may have regressed by July if it hasn't kept producing quality signals.
Recommended cadence
- Complete audit (7 dimensions): every 6 months. Ideally in January and July, to capture the major AI model updates.
- Light monthly monitoring: test 5-10 target queries in ChatGPT, note brand and competitor citations. 30-45 minutes.
- Continuous watch: monitor major model updates (GPT-5, Gemini Ultra, Claude 4, etc.) to anticipate changes in citation behavior.
For B2B teams with long sales cycles, monthly monitoring matters especially: it's often during the early research phase that buyers query ChatGPT to build their shortlist. Being cited at that stage can influence who gets invited into the RFP. Our GEO guide details these specific stakes.
What this audit does not cover
Limits of this method
- Perplexity and other AI engines: this audit focuses on ChatGPT and Google AI Overviews, the two channels with the widest coverage. Perplexity, Claude.ai Search and Gemini Advanced are not systematically covered.
- Past training data: we cannot know with certainty what is in a model's training data. Direct tests are proxies, not certainties.
- Geographic variations: ChatGPT and AIO results vary by user location. This protocol tests from one location — results may differ for targets in other countries.
- Stability of AI answers: LLMs produce non-deterministic answers. Test each query 3 times and note the citation frequency, not just a single answer.
- Technical SEO audit: the technical SEO dimension (Core Web Vitals, indexing, robots.txt) is not covered here. For a combined SEO + GEO audit, see our complete guide.
Frequently asked questions about GEO audits
What is a GEO audit for a business?
A GEO audit (Generative Engine Optimization) evaluates a company's ability to be cited, referenced and recommended by AI engines such as ChatGPT and Google AI Overviews. It analyzes 7 dimensions: content citability, E-E-A-T signals, structured data, semantic coverage, external mentions, brand consistency, and direct testability. A GEO audit is distinct from a classic SEO audit: it focuses on the quality of the signals perceived by LLMs, not just by Google's crawlers.
How long does a complete GEO audit take?
A complete manual GEO audit takes between 4 and 8 hours for a mid-sized site (50-200 pages). The content analysis and ChatGPT visibility testing phase accounts for roughly 40% of the total time. The technical phases (schema.org, semantic structure) make up the remaining 60%. With a structured methodology like the one presented in this article, an experienced practitioner can cut that time to 3-4 hours.
How do I know if my business is cited by ChatGPT?
The most reliable method is the direct test: ask ChatGPT questions about your sector, your city and your services, and note whether your brand appears in the answers. Use query variations: "best [your service] in [your city]", "who do you recommend for [your activity]", "[your product] reviews 2026". Dimension 7 of our scorecard covers exactly this diagnostic with a grid of 10 sample queries.
Does a good SEO score guarantee good GEO visibility?
No. A high SEO score increases the probability of being indexed in LLM training data, but it does not guarantee citability. Sites with decent technical SEO scores but content that is too generic, with no demonstrable expertise and no strong E-E-A-T signals, stay invisible in ChatGPT and Google AI Overviews. GEO requires an additional layer: content structured to answer directly, citable sources, and demonstrated topical authority.
How often should a GEO audit be run?
A complete GEO audit is recommended at least every 6 months. AI engines evolve fast: what was citable in January 2026 may no longer be in July. Between two full audits, monthly monitoring of the key signals (ChatGPT visibility testing on 5-10 target queries) is enough to catch regressions early. The scorecard presented in this article is designed for these quick monthly checks (30-45 minutes).
What is the difference between a B2B and a B2C GEO audit?
In B2B, GEO queries are more specific and LLMs give more weight to mentions in professional media. In B2C, brand awareness and customer reviews carry more weight. Our dedicated GEO guide details these specifics with concrete examples.
Sources and references
- Google SERP analysis — AI Overviews display rate, January and May 2026 (internal Cicero Studio analysis across 500 commercial and informational queries). Method reference: Google Search Central — AI features and your website.
- Google Search Central — official Schema.org documentation: schema.org/FAQPage and schema.org/HowTo (May 2026).
- Google Search Central — E-E-A-T framework and the Quality Rater Guidelines (latest published version).
- OpenAI — ChatGPT documentation, behavior of GPT-4o models with web browsing: OpenAI web search tool guide (May 2026).
- Google — AI Overviews and generative AI in Search, display conditions and cited sources (May 2026).
