Local & Industries · 9 min read · July 15, 2026
The GEO Audit: How to Show a Client Their AI Visibility in 30 Minutes
The GEO audit is the fastest way to turn a hunch into a case you can sell: you ask the same AI systems your prospects already turn to for recommendations, then show whether the brand shows up, how it's described, and who gets named instead. Half an hour, one document, one moment where the client sees it for themselves.
Why GEO Is No Longer a Niche Topic for Agencies
Your clients' customers don't just google anymore, they ask. ChatGPT, Perplexity and Google AI Overviews. Their own customers are doing the same thing. When someone asks, "Which advertising agency in Munich is good for B2B software?", it's no longer just the top Google spot that decides the outcome, it's whether the AI includes the brand in its answer at all. That's exactly where your new service line comes in: Generative Engine Optimization, or GEO.
This matters twice over for you as an agency. First, you need your own AI visibility under control, or you'll be recommending something you haven't tested on yourself. Second, GEO is something you can actually sell, while classic SEO has turned into a commodity for a lot of clients. The GEO audit is your door-opener: a concrete, visible result you can produce in under an hour that gives you real talking points for the next pitch.
The good news: you don't need an enterprise tool or a data science team. You need a structured question catalog, three or four AI systems, and the discipline to document what you find. Everything else is craft, and you'll get it step by step in this article.
What the GEO Audit Actually Measures
A GEO audit answers three core questions. First: does your client's brand show up at all when you ask AI systems for providers in their industry? Second: how is it described, accurately, out of date, favorably, or with facts that are simply wrong? Third: who gets named instead, or alongside them, meaning which AI-visible competitors are taking the spotlight your client wants?
We call these three dimensions presence, portrayal, and competitive landscape. Presence is binary and blunt: named or not. Portrayal is qualitative, and often the emotional trigger, because clients are genuinely unsettled when an AI describes a service they discontinued years ago as a current offering. Competitive landscape is strategic, because it shows who you're actually up against in AI answers, and that's rarely the same set of names as the Google ranking.
The honest framing matters here: an audit is a snapshot. AI answers aren't deterministic, they shift. So you don't measure with a single prompt, you run a series, and you log the date, the model, and the exact wording. That's what turns a nice screenshot into a real consulting artifact you can compare against the next audit.
The Question Catalog: Your Most Important Tool
The whole audit lives or dies on the prompts. A common rookie mistake is asking, "Do you know the agency XY?" That tells you nothing, because you've already put the name in the AI's mouth. The right approach is need-based questions written from the end customer's point of view. For example, if your agency's client is a trades business: "I'm looking for a roofer near Augsburg for a renovation, who would you recommend?" That's how you measure real visibility.
Build eight to twelve prompts per client across three categories. Category one: pure need prompts with no brand names ("best providers for X in region Y"). Category two: comparison prompts ("provider A or provider B, which is better for Z?"). Category three: brand prompts that use the name directly ("What do you know about the brand XY?"), to check how it's portrayed. Together, this mix covers presence and reputation at once.
If you're running this audit across clients in different industries, a reusable template pays off fast. Build a prompt library where industry, service, and region are the only variables you swap out. That turns the 30-minute audit into a repeatable process a junior on your team can run reliably, without the quality dropping.
The 30 Minutes, Step by Step
Minutes 0 to 5: setup. Open ChatGPT, Perplexity, Google with AI Overviews, and optionally Gemini or Copilot in separate tabs. For ChatGPT and Perplexity, use a fresh chat or incognito mode where you can, so your own history doesn't skew the result. Have your prompt template ready, plus an empty table with columns for prompt, system, result, and notes.
Minutes 5 to 25: run it. Work through your question catalog systematically, every prompt in every system. Copy the relevant part of each answer into your table, not the entire response. Note immediately whether the brand was named (yes/no), which competitors show up, and whether the portrayal contains errors. Screenshots of the key answers are worth their weight in gold, because clients trust visual proof more than your summary.
Minutes 25 to 30: distill it. Calculate the hit rate (the share of need-based prompts where the brand was named), list the three competitors that come up most often, and note the two most serious portrayal errors. Those three numbers plus two screenshots are already a presentable result. Polish and formatting can come later, the substance is there after half an hour.
Turning Raw Data Into a Story You Can Sell
Raw prompt output convinces no one in a meeting. Your job as the agency is to translate it into a story with three acts. Act one: the status quo, usually a sobering one (for example, the brand didn't come up at all across most of the need-based prompts). Act two: the competitive picture ("Instead, the AI keeps recommending these three competitors"). Act three: the opportunity ("With the right work, this gap can close within a few months").
The strongest moment in a client conversation is almost always the wrong or outdated portrayal. When ChatGPT claims a client doesn't offer a service that's actually their core business, or references an old company name, that's the emotional trigger that gets people ready to act. Document findings like this especially carefully, because they sell the follow-on project almost by themselves.
Package the result as a tight document: one page of executive summary with the three headline numbers, then the proof screenshots, then a short outlook on next steps. Resist the urge to build a forty-page report. The audit works because it's clear and fast, not because it's thick. Depth belongs in the paid follow-up engagement.
Common Mistakes That Undercut the Audit
The first classic mistake is a personalized history. If you test from your own ChatGPT account, one where you've mentioned the client's brand dozens of times already, you're measuring your own history, not reality. Use fresh sessions and document that you did. Otherwise you present an inflated version of visibility that falls apart the moment the client checks it themselves, and that costs you credibility.
The second mistake is a single measurement. One prompt on one day is chance, not a finding. AI systems answer inconsistently. Run every core prompt at least two or three times, across several systems, and talk about tendencies rather than one-off results. Being upfront about this protects you when the client sees a slightly different answer on their own next attempt.
The third mistake is confusing visibility with ranking. GEO isn't SEO under a new name. A top Google position doesn't guarantee an AI mention, and conversely, AI answers often surface providers that rank further down on Google but have well-structured, citable content and strong third-party mentions. Mix these two up and you end up selling the wrong fix.
From Free Audit to Paid Retainer
The audit is meant to be a door-opener on purpose, but it can't turn into an endless loop of free analysis. A model that works: the first audit is part of your pitch and free or low-cost. Everything after that, defining the plan of action, doing the implementation, and above all the ongoing monitoring, is paid work. That's how a single aha moment turns into recurring revenue.
Monitoring is the core of the business model here. GEO isn't a project with an end date, it's an ongoing process, because the models, training data and the competitive field keep shifting. A quarterly or monthly report tracking the hit rate over time is a strong retainer argument. You make progress visible and give the client a reason to keep working with you every month.
For your agency, this is strategically valuable, because GEO is still a young field with little competition and clients willing to pay for it. Whoever builds a clean, repeatable audit process now and packages it as a standard offer positions themselves as the go-to expert right as clients start asking these questions on their own.
Your Next Step: Run the Test Yourself
Don't start with a client, start with yourself. Run the 30-minute audit on your own agency. Ask the AI systems which advertising agency they'd recommend for your specialty and your region. The result is your first real data point, your practice round, and an honest look in the mirror that's often more uncomfortable than expected.
After that, pick a client you have a good relationship with for the second run. Present the result to them not as a sales pitch, but as a gift: "We took a look at how you show up in AI systems, you should see this." That approach opens doors, because you deliver value first and only talk about working together afterward. Most clients ask about next steps on their own.
Keep your process in a living document: prompt template, table structure, report template. After three or four audits, you'll have a routine the whole team can run, and a product with a clear price tag. That's how thirty minutes of curiosity turns into a new line of business that matures right as the market starts to need it.
Common questions
Do you need paid AI subscriptions or specialized tools to run a GEO audit?
For getting started, the free versions of ChatGPT, Perplexity, and Google AI Overviews are enough. Paid tiers sometimes pull more current web data and help with consistency, but they're not required. Dedicated GEO monitoring tools only start to pay off once you're running the audit as an ongoing retainer across several clients and need automated tracking over time.
How do I explain to a client the difference between GEO and the SEO we already do for them?
SEO optimizes a website to rank high in a results list and get clicked. GEO makes sure an AI actually names the brand in its generated answer and describes it accurately, often with no click involved at all. The two are related but not the same: a strong Google ranking doesn't guarantee an AI mention. The audit is what makes that difference concrete for the client.
What if the client is completely invisible across the AI systems?
That's not a problem, it's your strongest selling point. Total invisibility shows the client in black and white that action is needed, and positions you as the one who spotted the gap. What matters is presenting it factually, as a starting point, and pairing it with a clear path forward: structured, citable content, strong third-party mentions, and consistent business information are the levers that build presence.
Read on