How to Choose a GEO Agency: 7 Criteria That Matter

Choose a GEO agency the way you would audit one: make them prove they can do for themselves what they promise to do for you. Seven criteria cover it: their own AI-answer visibility, named client citations, an inspectable method, answer-first content skill, structured-data competence, measurement built on AI answers rather than rankings, and honesty about what nobody can guarantee yet. An agency that clears all seven is worth a conversation. One that dodges the first two is not.

The reason buyers are asking this question now is that the ground has moved. Generative engine optimisation (getting your business named and cited inside ChatGPT, Perplexity and Google’s AI answers) has split off from classic SEO, and business owners are shopping for help in a category almost nobody has a long track record in. That makes the selection harder, not easier: every agency can claim GEO expertise, because the field is too young for most claims to be checkable. The seven criteria below are the checks that still work.

What does a GEO agency actually do?

A GEO agency’s job is to make your business the answer AI assistants give, cited by name when someone asks ChatGPT, Perplexity or Google’s AI for what you sell. The work behind that is concrete: restructuring pages so an assistant can lift a clean answer, adding the schema and entity signals machines read, building the evidence and citations assistants prefer, and measuring whether the assistants actually name you.

The reason this is now a separate purchase from SEO is measurable. Analysis of real browsing behaviour by Pew Research Center found that “Users who encountered an AI summary clicked on a traditional search result link in 8% of all visits”, while “Those who did not encounter an AI summary clicked on a search result nearly twice as often (15% of visits)”. When the answer appears above the links, being the answer beats being the link.

Ahrefs measured the same shift from the publisher’s side: “the presence of an AI Overview now correlates with a 58% lower average clickthrough rate for the top-ranking page” (Ahrefs). Ranking first and being invisible in the answer is now a real, common outcome. It is precisely the gap a GEO agency exists to close.

How do you shortlist agencies when every “best GEO agency” list disagrees?

Ignore the lists and run the test yourself. Open ChatGPT, Perplexity and Google’s AI mode and ask them the questions your customers would ask: “best GEO agency in Australia”, “who can improve my visibility in AI search”. The agencies the assistants actually name, and the sources they cite, are the real shortlist. It costs ten minutes and it measures the exact skill you are buying.

Treat any ranked list you did not compile yourself as marketing until proven otherwise. The format is easy to assemble, impossible for you to verify, and nothing obliges its author to be neutral. Run your own comparison once, checking the list’s names against the names the assistants actually give you, and you no longer need to take anyone’s word for it. An agency that cannot be found by the machines it claims to influence is telling you something its case studies will not.

The same test works on their clients. Ask the assistants about the client industries the agency claims wins in. If the agency’s case-study clients never appear in actual AI answers, the case studies are describing effort, not outcomes.

What proof should a GEO agency show before you pay?

Three things: their own AI-search visibility, named clients being cited today, and per-client evidence that the citations followed the work. Every one of those is checkable from your side of the table, which is the point. In a field this young, verifiable beats impressive.

Concretely, ask for screenshots or live demonstrations of client names appearing in ChatGPT, Perplexity or AI Overview answers, with dates. Ask what the client’s visibility looked like before the engagement. Ask which specific pages earned the citations and what was changed on them. An agency doing the work well will have this evidence lying around, because collecting it is the reporting. An agency that responds with rankings charts is selling you the previous decade.

The honest version of this conversation includes a qualifier you should want to hear: nobody can guarantee placement inside a specific AI answer, because the assistants’ selection changes between model versions and even between phrasings of the same question. An agency that promises guaranteed citations is promising something the field cannot deliver yet. We would rather tell you that plainly. If a guarantee is the deciding factor, no honest provider will win the comparison.

What should their method actually include?

An inspectable GEO method has four layers: answer-first content structure, machine-readable markup, evidence density, and entity clarity. Ask the agency to walk you through each on a real page they have shipped. If the walkthrough is vague on any layer, or leans on a proprietary secret you may not see, the method is thinner than the pitch.

The reference point for the content layer comes from Google itself: “Google’s systems prioritize helpful, people-first content, not content designed solely for search engine rankings.”

“Google’s systems prioritize helpful, people-first content, not content designed solely for search engine rankings.”

Google Search Central

That principle transfers directly to assistants: they cite pages that state a complete, self-contained answer near the top, back it with sourced evidence, and mark up their structure so a machine can parse it. A competent agency will show you answer-first blocks, FAQPage and organisation schema, cited statistics, and consistent entity details across your site, then explain why each earns citations. We have written up how GEO and traditional SEO pulled apart if you want the mechanics behind that shift before you sit in the meeting.

How should a GEO engagement be measured?

By answers, not rankings. The reporting you want counts how often your business is named or cited in AI answers to your money questions, tracked across the major assistants on a fixed question set, plus referral traffic from assistant surfaces and enquiries that mention finding you through an AI tool (the number that pays for everything).

Rankings still matter as a supporting signal, but a report that stops at rankings is measuring the input, not the outcome you are paying for. Insist on the fixed question set: the same customer questions, asked the same way, month over month. Citation presence on a stable question set is the closest thing this field has to a rank tracker, and agencies doing real work have already built one.

This is also where connected work beats siloed work. Citations compound the way a business does: the content earns the citation, the citation earns assistant traffic, the traffic earns links and mentions that make the next citation more likely. Like a molecular bond, the strength of the outcome depends on how the elements connect. A marketing engagement built as one system captures that loop; a bolt-on GEO add-on usually does not.

What are the red flags?

Guaranteed placements, secret methods, rankings-only reporting, and pricing that cannot survive the question “what exactly do I get each month?” Each one signals the same underlying problem: the agency is selling certainty in a field that currently rewards evidence, iteration and honesty about limits.

Add one more: an agency that never says “no”. GEO is not the right first spend for every business. If your site lacks the service-page foundations or your customers do not research before buying, the honest advice is to fix that first. If we are not the right fit for what you need, we will tell you, and the agency you hire should pass the same test. Start with a straight conversation about where you stand, or see how we scope and price this work. Either way, take the seven criteria into every meeting you book.

Frequently asked questions

What is generative engine optimisation (GEO)?

Generative engine optimisation is the practice of making a business visible inside AI-generated answers: getting named and cited when ChatGPT, Perplexity or Google’s AI answers a customer’s question. It overlaps with SEO but optimises for being quoted in the answer rather than ranked in the links, which changes how pages are structured, what evidence they carry, and how success is measured.

How do you test whether a GEO agency is any good?

Ask the AI assistants yourself. Put your customers’ questions, plus “best GEO agency” queries, to ChatGPT, Perplexity and Google’s AI mode, and see whether the agency and its case-study clients actually appear. Then ask the agency for dated evidence of client citations and the specific page changes that earned them. Ten minutes of direct testing beats any ranked list you did not compile yourself.

Can our existing SEO agency handle GEO?

Sometimes. The skills overlap but are not identical. Ask them the same questions you would ask a specialist: show your own AI-search visibility, show a client being cited, walk through the answer-first and structured-data method on a live page. An SEO agency that has genuinely added GEO capability can answer all three quickly. One that rebadged its existing retainer will reach for rankings charts.

How long does GEO take to show results?

Expect movement in months, not days, and treat anyone promising instant citations with caution. Assistants have to recrawl and re-weigh your pages before answers change, timelines vary by how much structural work your site needs, and results arrive question by question rather than all at once. A trustworthy agency will show citation progress on a fixed question set from the first months while the compounding effects build.