Generative Engine Optimization (GEO) is still young enough that almost anyone can call themselves an expert. There’s no certification, no shared methodology,and many of the agencies pitching it today were doing traditional SEO eighteen months ago and have just relabeled the service. That means the burden is on you, the buyer, to ask sharp questions before you hire a firm.
We put this list of screening questions together based partly on what we wish more prospective clients asked us. Use some or all of the following as a vetting tool.
1. What metrics does the agency track for AI visibility?
This is the first filter, and it weeds out more agencies than you’d expect. If the answer is “mention count,” keep asking questions. Mention count tells you that your brand showed up somewhere in an AI response. It tells you almost nothing about whether that mention helped or hurt you, whether it appeared in a useful spot in the answer, or whether the same query is even worth tracking in the first place.
A team that’s built for this should be able to walk you through a stack of metrics, not just one. That means share of answer against named competitors, position within the answer (top of the response versus buried in a list of ten sources), frequency across repeated query variations, and some measure of how often your content gets pulled into answers. If they can only talk about one number, they’ve probably not done GEO work much. Modern metrics like narrative consistency or share of explanation should also be in the conversation
2. Do they have an actual framework, and can they walk you through the steps?
This is the question that exposes whether an agency has a methodology or just a service menu. Ask them directly: what’s your process, start to finish, and what does step one actually look like before you’ve signed anything? A team with a real framework will answer in order. Research and diagnosis first, then prioritization, then execution, then measurement, with a rough sense of how long each phase takes and what you should expect to see along the way.
What you don’t want to hear is a grab bag of tactics with no sequence behind them, schema markup here, some content production there, a Reddit push somewhere in the mix, with no explanation of why those tactics in that order or how they were chosen for your specific situation. If the agency can’t describe their own process without referencing your brand at all, that’s a sign they’re improvising a plan in real time rather than running one they’ve used before.
3. Do they start with customer research, or do they jump straight to tactics?
A lot of agencies skip a step that should come first: figuring out which AI tools your actual buyers use, how they query them, and where in the buying journey they reach for one. Without that, an agency is optimizing on assumptions, and assumptions about AI tool usage are often wrong. A buyer might lean on ChatGPT for everyday questions but switch to Perplexity specifically when researching a purchase in your category because they trust the sourcing more. Skip the research and you risk building a whole program around the wrong platform.
We borrow a lot from a customer research survey framework built to reveal which AI tools buyers use, how they search, and how AI shapes purchase decisions. Any agency worth hiring should be running something similar before a single piece of content gets written. Ask whether they survey your actual customers or prospects, what they ask about (tool usage, query phrasing, where AI shows up across the purchase funnel, how much your buyers trust an AI recommendation versus other sources), and whether they adjust the plan based on what comes back. If their answer is that they already know which platforms matter for your category without having asked anyone in it, push back.
4. Do they have a process for prompt universe mapping?
Before an agency can optimize anything, they need to know which specific questions your buyers are actually asking an AI model. Not vague categories like “questions about our product,” but the real list: the 100 to 200 actual prompts people type or speak when they’re researching a problem you solve, comparing you to competitors, or trying to figure out if you’re trustworthy. That list is what’s called a prompt universe, and mapping it is foundational work, not a nice-to-have.
Ask the agency how they build this. A real answer involves pulling from the customer research above, mining your own sales and support conversations for the language buyers actually use, and testing variations to see which phrasings surface your brand versus which ones surface a competitor instead. If they can’t show you a prompt universe for a client, or if their version of this is a dozen generic head-term keywords with “AI” pasted in front, they haven’t done the work. Everything downstream, content, citation gap fixes, measurement, depends on getting this list right first.
5. Do they use an AI-native monitoring tool, or is this SEO software with GEO bolted on top?
A lot of “GEO platforms” right now are existing rank-tracking or SEO tools that added a feature to check whether your brand shows up in an AI Overview, slapped a new label on the dashboard, and called it a GEO product. That’s not the same thing as a tool built from the ground up to understand how generative answer engines actually work, and the gap shows up fast once you start asking what the tool can actually do.
Ask the agency point blank what’s under the hood. Was this tool built specifically to query and parse responses from ChatGPT, Perplexity, Gemini, and AI Overviews, or is it a legacy SEO tool with a new module stapled on? We have seen bad data come from SEO-native tools.
You want to know if an agency’s preferred tool can track position and framing within a generated answer, or does it just confirm a citation exists somewhere in the sources list? Does it handle the fact that AI answers vary run to run for the same prompt, or does it treat one snapshot as gospel the way a traditional rank tracker treats a SERP position? An agency relying on repurposed SEO tooling isn’t lying to you, but they’re measuring a fundamentally different thing than what you’re paying them to optimize for.
6. Do they understand citation gap analysis?
This is the question that tends to separate the agencies who’ve actually built a methodology from the ones who are improvising. A citation gap analysis means looking at the sources an AI model is currently pulling from for a given topic or query cluster, figuring out what’s missing or thin in your own content compared to those sources, and building a plan to close that gap.
In practice this looks like reverse-engineering why a competitor’s comparison page or a third-party review site keeps getting cited instead of your product page. Sometimes it’s a structural issue (their page answers the question more directly). Sometimes it’s a trust issue (the model favors a source it’s seen cited elsewhere). Sometimes your content simply doesn’t exist in a format the model can easily extract and summarize. An agency that’s done this work will have a process for diagnosing which of these is happening and a plan that follows from the diagnosis, rather than a generic recommendation to “publish more content.”
7. What does a GEO-optimized content strategy look like?
There’s no single template here, and any agency that hands you a rigid formula is probably overselling it. But there are recognizable patterns in content that performs well across AI answer engines, and an agency with real experience should be able to describe them concretely rather than in vague terms like “high quality, authoritative content.”
Specifics worth listening for: clear, direct answers near the top of a page rather than answers buried under a long preamble. Structured formatting (lists, tables, defined sections) that’s easier for a model to extract and summarize cleanly. Original data, original research, or a genuinely distinct point of view, since models tend to favor sources that add something rather than ones that restate the consensus. And a publishing cadence and update process that keeps content fresh, since AI platforms weight recency heavily. If the agency’s answer to this question sounds like a paragraph from a generic SEO pitch deck, that’s a signal.
8. Is their own house in order?
When you’re evaluating an agency, ask them to pull up their own AI visibility numbers live on the call, not a client’s. Watch how comfortable they are doing it.
We hold ourselves to this standard. Before we built out our own GEO program, we evaluated several AI visibility platforms and landed on Scrunch to run it end to end. We identified nearly 700 prompts to prioritize across ChatGPT, Perplexity, Gemini, and Google AI Overviews and benchmarked ourselves against the competitive landscape, niche agencies and bigger established firms alike. The starting point wasn’t pretty. Our visibility in AI answers was close to zero. From there we rebuilt our own site for machine readability, invested heavily in third-party citations and placements, and tracked the results the same way we’d track a client’s.
By May 2026, we were the most visible agency in our category across those nearly 700 tracked prompts, with our own website tied for the most influential domain in the results and our citation consistency outranking sources like LinkedIn, Reddit, and Forbes. Leads grew fivefold after the program launched, and the deals coming in skewed larger too.
None of these questions have a single “correct” answer that proves an agency is good. What they prove is whether the agency has actually built a real methodology versus repackaged an SEO retainer with new terminology. A team that’s done the work will answer most of these without hesitation, point to specifics, and probably ask you a few sharp questions back about your own category and competitive set.
If you walk through this list on a sales call and get vague answers, a single dashboard screenshot, or a lot of confident language with no specifics underneath it, that’s useful information too.