I recently delivered two GEO audits back to back: a services brand and a high-end real-estate player. Same question on both sides, almost word for word: "When a prospect asks ChatGPT or Perplexity for advice in our field, do we show up?". In both cases the starting answer was fuzzy, and that's exactly the problem a GEO audit solves.
SEO tells you where you stand on Google. GEO, Generative Engine Optimization, tells you whether you exist inside the answers of generative AIs. In 2026 that's a business question, not a technical curiosity: a growing share of commercial-intent searches now starts inside an AI assistant rather than on a results page.
Ranking well on Google no longer guarantees being cited by ChatGPT. Two engines, two logics, two audits.
Why an SEO audit is no longer enough
When a user asks a question in Perplexity or triggers a Google AI Overview, they don't get ten blue links: they get a written answer with a handful of cited sources. If your brand is neither in the text nor in the sources, you are invisible, even if you rank first in classic SEO.
So a GEO audit answers three questions the traditional SEO audit doesn't cover:
- Do you appear in the generative answers of your industry, and on which questions?
- How do you appear: cited as a trusted source, mentioned in passing, or described with negative sentiment?
- Who shows up instead when you're absent, and why those competitors?
The 5-step method I use
1. Build the business prompt corpus
Everything starts here. We don't test keywords, we test real questions, the ones a prospect asks in natural language. I start from about thirty prompts grouped into families: discovery questions ("which solution for…"), comparison questions ("X or Y"), reassurance questions ("is [brand] reliable"), and pure brand questions ("reviews of [brand]"). This corpus is what makes the audit defensible: it follows the buying journey, not a theoretical search volume.
2. Query the four engines that matter
I submit each prompt to ChatGPT (with and without web browsing, because the two modes don't answer the same), Perplexity (which always cites its sources in real time), Gemini (tied to Google Search) and Google AI Overviews. For each answer we record: is the brand cited, at what position, as a source or in the body, with what sentiment, and which competitors appear.
3. Repeat to neutralize volatility
This is the #1 trap of amateur GEO: test once and conclude. LLM answers are extremely volatile. A brand cited on Thursday can vanish on Monday with no change at all. A serious audit replays each prompt over several days and computes an average citation rate, not a snapshot. That's the difference between an anecdote and data.
4. Compute the AI Share of Voice and map the gaps
We consolidate everything into a few decision-ready indicators:
| Metric | What it measures |
|---|---|
| AI Share of Voice | % of answers that cite your brand |
| Inbound citations | Answers listing your site as a source |
| Sentiment | Positive / neutral / negative when cited |
| Dominant competitors | Who occupies the space you don't |
| High-potential prompts | Strategic questions no one owns yet |
That last row is often the most valuable: the questions where no player is established yet are open ground to capture quickly, before a competitor anchors there.
5. Translate into a prioritized action plan
An audit that ends on a finding is useless. The output is a roadmap ranked by effort/impact: which pages to restructure (answer-first, FAQ, hard numbers), which E-E-A-T signals to strengthen (author bios, verifiable citations, external mentions), whether to aim for a Wikipedia page, and what content to create to capture the orphan prompts. The goal: move from invisible to cited on the questions that weigh on your business.
What a first audit almost always reveals
Three findings come up every time: (1) the brand ranks far better on Google than it is cited by the AIs (SEO doesn't transfer automatically); (2) the cited competitors aren't always the market leaders, but those with the most structured, most "citable" content; (3) there is always a handful of strategic questions no one dominates, so easy to preempt.
Mistakes that sink a GEO audit
- Testing from your own logged-in account: personalization and history skew the answers. We audit in a neutral session.
- Confusing "being mentioned" with "being cited as a source": a text mention without a link doesn't bring the same traffic or authority as a source citation.
- Ignoring sentiment: being cited negatively (a review, a dispute, a critical article) is an e-reputation matter, not a GEO win.
- Stopping at the audit: the snapshot is useful, but it's monitoring over time that lets you steer.
One-off audit or continuous monitoring?
Both have their place. The one-off audit is the right entry point: it gives a quantified baseline and a roadmap. For brands genuinely betting on their visibility inside the AIs, I then move to recurring monitoring that replays the critical prompts every week and alerts when the Share of Voice drops. I detailed that setup in my field report on GEO monitoring for ChatGPT, Perplexity and AI Overviews.
If you want to see exactly what the deliverable looks like and what it costs, I detailed the offer on the GEO Audit page. And for the broader strategic frame, the GEO expertise page explains the four pillars that decide whether an AI cites you.
Find out if the AIs cite you
I test your brand across ChatGPT, Perplexity, Gemini and Google AI Overviews, and hand you a prioritized roadmap.