FOR BRANDS HIRING
How to choose a GEO agency.
If your brand is losing AI recommendations to competitors and you are about to hire someone to fix it, these are the five questions that separate an agency doing the work from one selling you a dashboard.
A GEO agency — generative engine optimization — works to get your brand recommended inside AI assistants like ChatGPT, Gemini, Perplexity and Claude, rather than ranked in a list of links. The job splits in two: measuring how often you are named and recommended across a fixed set of buyer questions, and then earning presence in the third-party sources those assistants already trust.
You will also see the same practice sold as an AEO agency, an AI visibility agency, an AI search agency or an LLM optimization agency. The labels are close to interchangeable. The methods are not, and that is what this page is about.
We are not an agency. We build the measurement and commerce layer that several of them run on, which means we see a lot of engagements from the inside — the ones that work and the ones that quietly renew on a chart nobody acts on.
Five questions to ask before you sign.
Ask all five on the first call. A good agency will answer them without flinching, because they have already had the argument internally.
01
“What is the prompt set, and can I see it?”
Everything downstream is built on this list. If it was auto-generated from keywords and you cannot inspect it, you do not know what is being measured — and you cannot tell whether it reflects how your buyers actually ask.
A good answer sounds like: They hand you the list, it contains questions you recognise from sales calls, and it is fixed before measurement starts.
02
“How many runs per prompt, and what is the spread?”
Assistant answers are non-deterministic. The same prompt, same model, three runs, three different brand sets. A tracker that fires each prompt once and draws a line is plotting noise.
A good answer sounds like: Multiple runs per prompt, with variance reported. A four-point move that sits inside normal run-to-run spread is not a result.
03
“Are mentions and recommendations counted separately?”
Being named in a list of five is a mention. Being put forward as the answer is a recommendation. Both increment a blended score; only one sells anything. A brand mentioned constantly and recommended rarely looks healthy on the dashboard while losing every head-to-head.
A good answer sounds like: Two numbers, always reported apart. If the agency has never noticed a gap between them, they are probably not measuring one.
04
“Is it broken out per engine?”
ChatGPT, Perplexity, Gemini and Claude do not draw on the same sources and do not behave alike. A single cross-engine percentage averages away the only actionable detail — which surface you are losing, and to whom.
A good answer sounds like: Per-engine reporting, and a different plan for each. “Improve AI visibility” is not one project.
05
“Which sources did each answer cite?”
This is the half that tells you why, and it usually points off your own domain — to a listicle, a review platform, a trade publication or a forum thread you are absent from. Without it you get a diagnosis and no treatment.
A good answer sounds like: A cited-source log per prompt, and a plan that includes work you cannot do on your own website.
We hold ourselves to the same five. Our answers are on the methodology page, and the reasoning behind them is in why a single AI visibility score is a vanity metric.
Five things that should end the conversation.
None of these mean the agency is dishonest. Most mean they are selling a category norm that the people doing this work seriously have already abandoned.
“We'll get you ranked #1 in ChatGPT”
There is no ranked index inside ChatGPT to hold a position in. Nothing is at number three. An agency promising a rank is describing a structure that does not exist — which tells you what the rest of the reporting will be worth.
A single visibility score, with no methodology
Two agencies can both report 62 while measuring entirely different things. Ask how the number was built. If the answer is vague, the number is decoration.
The whole plan is on-page
In audits run across this category, the large majority of citations come from third-party sources rather than the brand's own site. An agency whose plan stops at your PDPs and schema is working on the minority of the problem.
“Add an llms.txt and you'll get cited”
It will not hurt, but it is nowhere near the lever it is sold as, and it does nothing about the actual reason you are invisible. Treat it as housekeeping, not strategy.
Reporting that cannot connect to revenue
Most AI-influenced visits arrive with the referrer stripped and land in analytics as Direct. If the agency has no answer for that beyond a visibility chart, you will not be able to defend the budget at renewal.
Agency, platform, or both?
They solve different halves, and conflating them is how brands end up paying twice for neither. A platform gives you the measurement and the technical layer — a catalog an agent can read, a checkout an agent can complete, and numbers that tie back to revenue rather than traffic. An agency gives you the people: publishing, digital PR, seeding into the third-party sources that actually get cited, and running the loop every month.
If you have fewer than three people on this internally, you will probably want both. If you are choosing one first, choose based on which half you are least able to staff.
Either way, start from a baseline you did not have to take on trust. Our free Citation Rank scan gives you your own numbers before anyone pitches you — useful as a brief, and useful for checking what you are told later. If you want the technical half, Visibility Score is where the measurement lives, and House of Zelena is what the loop looked like end to end.
If you are the agency
Run this on our platform.
If you deliver GEO, AEO or AI visibility for clients, you can operate Tru Commerce on their behalf: Share of Voice and Citation Rank tracking, feed and PDP work, MCP and agentic checkout builds, and reporting that shows agent-driven revenue rather than a visibility percentage. You keep the client relationship. We stay underneath.
There are three tracks — referral, agency, and technology partner — and terms are agreed before any client work starts.