How to choose a GEO agency: a scorecard you can use on any sales call
Six weighted criteria, four levels of evidence and 12 questions to ask, so you can compare GEO agencies on proof instead of promises.
On this page
The short answer: Judge a GEO agency mainly on proof you can check: measured, dated results for named questions, ideally including the agency's own visibility. Then look at whether its method is written down, who actually does the work, whether it fits your stage and market, how it reports, and whether its price and terms are clear. The scorecard below weights those six things and gives you the questions to ask.
Generative engine optimisation (GEO) is new, and the market is full of confident claims. Many "best GEO agency" lists are written by agencies that rank themselves first. That doesn't make them wrong, but it means you need your own way to compare. This is ours. Use it on us too.
The scorecard
Score each agency from 0 to 5 on every criterion, multiply by the weight, and add up. A perfect score is 100.
| Criterion | Weight | What good looks like | Ask them |
|---|---|---|---|
| Proof you can check | 30% | Measured results for specific questions, with dates, assistants and before/after numbers | "Show me a result for a named question, with the dates and how you measured it." |
| A method you can inspect | 20% | A written process: which questions, which assistants, how often, what changes they make | "What exactly will you do in the first 30 days, and why?" |
| Who does the work | 15% | You meet the people doing the work, and they stay on your account | "Who will work on our account each week? Can I talk to them now?" |
| Fit for your stage and market | 15% | Experience with companies like yours: size, category, region, language | "Which clients like us have you worked with, and what changed for them?" |
| Reporting | 10% | The same metric, measured the same way, every month, with sources | "Can I see a real (anonymised) monthly report?" |
| Price and terms | 10% | A written scope and price, a sensible minimum term, and a clear way out | "What's the price, the minimum term, and what happens if the numbers don't move?" |
How to read the total: 80 or more is a strong candidate. Between 60 and 80, check the weakest criterion before deciding. Under 60, keep looking, unless the gap is only about client history and everything else is excellent (see "new agencies" below).
Rate the evidence, not the confidence
Every agency has a story. What separates them is the kind of evidence behind it. Sort what you hear into four levels:
- Level 1: re-runnable proof. Specific questions, named assistants, dates and counts, which you could check yourself. For example: "named in 12 of 50 answers to these 10 questions in ChatGPT, Claude, Perplexity, Gemini and Google, 1–7 September, up from 2 of 50 in June." This is the strongest evidence.
- Level 2: named case studies with numbers. A real client, a metric and a time frame, but not something you can re-run yourself.
- Level 3: anonymous claims. "A fintech client tripled its AI visibility." It may be true, but you can't check it.
- Level 4: promises. "We'll get you to number one in ChatGPT." This is a red flag, not evidence.
The single most telling check is an agency's own visibility. GEO agencies sell exactly this, so ask: "When I ask AI assistants about agencies like yours, do you come up? Show me." An agency that measures and publishes its own numbers is practising what it sells.
12 questions to ask on the call
- Which questions would you track for us, and how did you choose them?
- Which assistants do you measure, and do you use web search mode?
- How do you deal with answers changing between runs?
- What's our baseline today, and when will you measure it?
- What changes will you make on our website in the first month?
- What will you do off our website (publications, review sites, communities)?
- Who writes the content, and who approves it?
- How do you decide which third-party sites to approach?
- What will the monthly report show? Can I see an example?
- Which tools do you use for tracking, and will we get access to the data?
- What results would make you say this isn't working?
- What's the minimum term, and what happens after it?
Good answers are specific and slightly boring: named assistants, fixed question lists, weekly checks, written plans. Vague answers ("our proprietary AI optimisation framework") deserve a follow-up question.
Red flags
- Guarantees. No one controls what an AI assistant says. Google's own guidance treats optimising for its AI features as ordinary SEO with no special tricks, so be wary of anyone selling secret techniques.
- No baseline. If they don't measure where you are before they start, they can't show you what changed.
- A dashboard and nothing else. Tracking is useful, but on its own it changes nothing. Ask what work they will do.
- Long lock-ins without milestones. A 12-month contract with no agreed checkpoints shifts all the risk to you.
- Tactics that break platform rules. Fake reviews, paid placements dressed up as editorial content, or spam in communities can backfire, and some break review sites' terms.
- Mass-produced content. Research on GEO found that adding statistics, quotations and sources helps content get used in AI answers, and keyword stuffing doesn't. Volume for its own sake isn't a strategy.
New agencies: how to judge them fairly
GEO is only a few years old, so many good agencies are new. A new agency will score lower on client history. That's expected, and it doesn't have to rule them out. Look harder at the other criteria instead:
- Do they publish their own visibility numbers and update them over time?
- Can they show you sample deliverables: a snapshot, an audit, a report?
- Is the method written down in enough detail to check?
- Will a senior person do the work, and will you be one of their first clients (often a good thing)?
- Do they offer founding-client terms that share the risk, such as a lower price, a shorter minimum term or a free first step?
Where Quoth stands on this scorecard
In the interest of fairness, here's how we'd score ourselves:
- Proof you can check: We're new, so we don't yet have client case studies. We've started measuring our own visibility from zero and will publish the weekly numbers on this blog. Until those numbers move, this is our weakest criterion.
- Method: Written down. We track a fixed list of buyer questions across ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews, in fresh sessions, every week. You can open a sample snapshot, audit and monthly report.
- Who does the work: Our founder, Amreet, sets the strategy for every account and is there to talk it through. Our team does the tracking, content, outreach and reporting.
- Fit: B2B software and service companies competing with bigger brands, in English-language markets.
- Reporting: The same questions, asked the same way, every month.
- Price and terms: Published: a free snapshot, a $900 audit, and $2,500 a month with a three-month minimum.
Use the scorecard on us, and on everyone else you're considering.
Frequently asked questions
How much does a GEO agency cost?
It varies widely, and many agencies don't publish prices. Ask for a written scope with a price before you sign anything long-term, and compare what's actually included: tracking only, content, outreach, or all three.
How long before a GEO agency shows results?
For assistants that search the web, changes can show up within weeks once the pages they read change. Google AI Overviews tends to follow normal search rankings, which take longer for newer sites. Ask the agency what they expect to see at 30, 60 and 90 days, and how they'll measure it.
Can my SEO agency do GEO?
Often, yes. The two overlap heavily. Ask them the same questions: which AI questions they'll track, how they'll measure answers, and what they'll do off your website. If they only talk about keywords and rankings, they may not be set up for it yet.
Do I need a tracking tool as well as an agency?
Not necessarily. A good agency will measure for you and share the data. A tool becomes worthwhile when you need to track hundreds of questions or many markets. See our comparison of agencies and tools.