How to Choose a Generative Engine Optimisation Agency

How do I choose a generative engine optimisation agency?

Choose a GEO agency on evidence, not deck promises. Ask it to run your category live on the first call and show where AI names you today. A capable agency reports your named-in-answer rate, builds the content and citations models trust, then re-scores monthly. If it cannot show a current reading, keep looking.

That is the short version. The longer version matters, because most buyers approaching this market have never bought generative engine optimisation before, and the suppliers know it. The category is young enough that there is no established way to tell a serious agency from one that has rebranded an SEO package and added the letters AI to the deck.

This guide gives you the criteria to judge on. None of it requires you to take a vendor's word for anything.

Where the usual ways of picking an agency fall short

The "best GEO agency" listicles you find through search are mostly paid placement. A directory ranks the agencies that pay to appear, or that wrote the directory themselves. The ranking tells you who spent money on listings, not who moves a client from absent to recommended in AI answers. Treat those pages as adverts.

Referrals are thin too, because few buyers have run a GEO programme long enough to judge one. The discipline is new. An agency that has been doing it for eighteen months is an old hand. So the safest filter is whether the agency can demonstrate the work in front of you, on your category, before you sign anything.

That sets the bar for everything below. A capable agency proves its method on your data in the first conversation. A weaker one stays in generalities and asks you to trust the process.

How to hire a GEO agency: the seven questions to ask

Put these to any agency on your shortlist. You are not trying to catch anyone out. You are separating a measurable programme from a content retainer wearing new language. Each question has an answer you want and a red flag that should end the conversation.

Question to ask	The answer you want	The red flag
Can you run my category live on this call?	Yes. They open ChatGPT, Gemini or Perplexity and ask your buyers' questions there and then.	They promise an audit "in the proposal" and never show you a live reading.
What metric do you report against?	Named-in-answer rate: how often AI names you across the questions your buyers ask.	Traffic, rankings, or impressions with no measure of whether AI recommends you.
How often do you re-score?	Monthly, question by question, against a baseline they set at the start.	One audit at the start, then content delivery with no re-measurement.
What do you actually build?	Content the models draw on, plus the citations and structured signals that earn trust.	Content alone, or link building alone. Half the job each way.
Do you understand my regulatory context?	They can speak to your sector's compliance and evidence expectations without coaching.	Generic content that ignores the claims a regulated buyer is allowed to make.
What results can you show, even anonymised?	Specific movement on real programmes, with the metric named, even if the client is not.	Vague growth claims, vanity multipliers with no source, or borrowed SEO case studies.
What happens when a competitor surfaces in an answer I used to own?	A monitoring and rapid-response model that catches the slip while it is small.	No answer. They deliver content and move on.

If an agency answers the first question well, the rest usually follow. The live audit is the single most revealing test, because it shows whether the agency works from measurement or from assertion.

Why the live audit on the call matters most

A live category reading is the one thing a weak provider cannot fake. Ask the agency to open an answer engine on the call and put your buyers' real questions to it. You will watch, in real time, which of your questions name you and which name a competitor instead. That reading is your starting position, and a serious agency treats it as the baseline for everything that follows.

Here is how it tends to play out from your side of the table. You ask, "which platform is best for [your category] in a regulated setting," and the model returns three names. Yours is not among them. It is an uncomfortable reading, and a useful one, because you now know your real starting position. An agency willing to show you that is confident in its method.

The named-in-answer rate is the metric that follows from it. Count the buyer questions where AI names you, divide by the total questions in your set, and you have a number that maps directly to whether buyers find you when they ask. That number is what should move. Rankings and traffic are proxies carried over from the old model. In a market where a buyer reads the AI answer and stops, the citation is the outcome that decides the deal.

What good GEO work looks like underneath

The strongest agencies run generative engine optimisation as a programme rather than a one-off project, in four connected steps. It helps to know the shape, so you can check an agency is doing all of it rather than selling you one slice.

First, the audit: a live reading of where AI names you across your buyers' questions. Second, content built to answer the real questions your buyers put to AI, written in the format models pull from when they form a recommendation. Third, citation building, which earns the references and structured signals that teach the models to trust your name, so you appear as a cited source rather than an afterthought. Fourth, monthly re-scoring, where the agency re-runs your question set and records the movement question by question.

An agency that does content but not citation building is doing half the job. One that audits once and never re-scores cannot tell you whether the work is paying off. If you want the full operating model laid out in detail, read the B2B GEO playbook. For the underlying discipline, start with what generative engine optimisation is.

Regulated-sector fluency is not optional

For a regulated B2B company, generic content is a liability. A healthcare compliance platform or a fintech operates inside strict rules about what it can claim and how it must evidence those claims. An agency that does not understand your sector will write content that either oversteps the line or stays so cautious it never earns a citation.

This is where many general-purpose agencies fall down. They can produce volume without producing the kind of precise, evidence-backed content that a compliance-minded buyer trusts and an answer engine cites. Ask directly whether the agency has worked in regulated B2B. The answer shapes whether the content will hold up.

What to ignore when choosing

Skip the "top ten agencies" directories. Most are paid placement and tell you nothing about results.

Discount any vendor quoting performance multipliers without a named source and year. Figures like "4.2 times more citations" circulate around this market. They trace back to vendor marketing rather than any measured result. A claim with no source is a claim you cannot use.

Ignore agencies that lead with tools. A monitoring dashboard is not a programme. Plenty of providers will sell you a subscription to a tracker and call it GEO. The tracker tells you where you stand. It does not do the work of moving you, which is building the content and citations, then responding when the answer shifts.

And be wary of any agency that talks about AI search but reports on the same metrics it always has. If the monthly report is still rankings and sessions, the agency has changed its language and kept its method.

How Forge meets these criteria

Forge Together runs generative engine optimisation as a programme for regulated B2B companies, on exactly the criteria above. We run your category live on the discovery call and set your named-in-answer rate as the baseline. From there we build the content and citations the models trust. Every month we re-score, so you watch your position climb question by question.

The proof is in the movement. A UK digital identity platform grew its monthly AI reach from around 39,000 to 2.9 million in six months and won a seven-figure contract directly through AI search. A healthcare compliance platform lifted its pipeline from £240,000 to £645,000, with its largest contract arriving through ChatGPT. Other programmes are still early, where a category sits open and we are building a client into the answer for the first time. Client names are withheld by agreement. The metric and the movement are real.

The market context is moving fast. According to G2's 2026 buyer study, a majority of B2B buyers now begin supplier research with AI chatbots more often than Google, and a large share chose a different vendor on the strength of AI guidance. (G2, 2026, exact figures pending verification against the published release.) Responsive and Digital Commerce 360 found that around one in four B2B buyers use generative AI more than conventional search for supplier research. (Responsive / Digital Commerce 360, 2025.) The agencies worth hiring are the ones already measuring against that shift.

Frequently asked questions

How do I choose a generative engine optimisation agency?

Start from a live reading of your category on the first call. Have the agency open an answer engine and put your buyers' questions to it, so you see where you stand before signing. Then check it reports named-in-answer rate, builds content and citations, re-scores monthly, and knows your sector.

How do I hire a GEO agency, and what should I ask before signing?

Ask whether it will run your category live on the call, what metric it reports against, how often it re-scores, what it actually builds, whether it knows your regulatory context, what results it can show even anonymised, and how it responds when a competitor surfaces in an answer you used to own. The answers separate a measurable programme from a content retainer.

What is the difference between a GEO agency and an SEO agency?

An SEO agency works to rank your pages in classic search results and reports on rankings and traffic. A GEO agency works to get your company named inside AI answers and reports on how often that happens. The foundations overlap, though the job and the metric are different. For the full distinction, see whether GEO is the same as SEO.

How much should a GEO programme cost?

Cost depends on how many buyer questions you are competing for, how open your category is, and how much content and citation work the position needs. The better starting question is the cost of being absent while the answer hardens around a competitor. For how to weigh value against cost, read whether GEO is worth it and what it costs.

Are "best GEO agency" lists reliable?

Mostly not. Most "best agency" listicles are paid placement or self-published directories, so the order reflects who paid or who wrote the page, not who delivers results. Use them as a starting list of names if you like, then judge each agency on a live audit and the seven questions above.

---

Sources: G2 buyer study, 2026 (figures pending verification against the published release); Responsive / Digital Commerce 360, 2025.

---

Book your free AI-visibility audit

Ask us the seven questions on a discovery call. We will answer every one of them live.

30 minutes. No pitch deck. We'll run your category live and show you exactly what AI says about you today.

Book a discovery call: forgetogether.agency/GEO