How to Evaluate a GEO Agency

Comparing agencies in this category is difficult because almost every claim sounds the same. Everyone says they do generative engine optimization. Everyone reports citations. Almost nobody publishes how they count them. This page is the framework we would want a buyer to use on us, which is why it is written to be applied to any agency rather than to favour one.

Six dimensions worth comparing

Evidence and case study depth

What can be verified independently, rather than described in a proposal.

Strong signalPublished research with a registered DOI, named clients with verifiable outcomes, or a methodology anyone can check. Rejected data points disclosed alongside accepted ones.

Weak signalAnonymised case studies with impressive percentages, offered only after an NDA. Confidentiality is legitimate; it is also the easiest way to make unverifiable numbers sound exclusive.

Measurement framework

Whether there is a stated method, or only a number.

Strong signalA named framework with defined components, a stated counting rule, and a frozen prompt panel handed to the client. The client can reproduce the measurement without the agency.

Weak signalCitation counts with no method attached, or a screenshot of one favourable AI answer. If the measurement cannot be repeated by the client, it cannot be audited.

Platform coverage

How many AI systems are measured, and whether separately.

Strong signalSix platforms measured and reported individually: ChatGPT, Gemini, Perplexity, Claude, Copilot and Grok. Per-platform reporting shows where you are strong and where absent.

Weak signalOne platform, usually ChatGPT, or an aggregate figure that averages away the platforms where you do not appear at all.

Schema and entity architecture

Whether the technical foundation is present or merely mentioned.

Strong signalOrganization, Service and LocalBusiness schema implemented across the site, with entity content structured so a machine resolves who you are without inference.

Weak signalSchema described as a service line but implemented on a handful of pages, or implemented without the entity content that makes it useful.

Commercial transparency

What is promised, and what is explicitly not.

Strong signalMilestone commitments published before signature, separated into targets the agency accepts being held to and targets where measurement is not yet mature. A closing audit that reports what was missed.

Weak signalGuaranteed citation in ChatGPT, guaranteed lead volume, or guaranteed revenue. None of these can be guaranteed, and promising them signals either inexperience or an intention to renegotiate later.

Category maturity and origin

How long the practice has been documented, and by whom.

Strong signalA dated public archive showing when the work began, and published material predating the current market interest in the term.

Weak signalA capability added after 2024 and described as founding expertise. Adding a service is legitimate; presenting it as origin is not.

Four archetypes you will meet

We describe categories rather than name firms. Naming competitors turns a decision framework into an argument, and you can identify the type without our help.

The SEO agency that added AI after 2024

Strong technical foundation, established client base, and a genuine ability to execute. The AI capability is usually newer than the positioning suggests, and measurement is often the weakest part, because it was built to report rankings rather than citations.

Ask: what did you publish about this before 2024, and how do you count a citation?

The specialist GEO startup

Focused, current, and usually fluent in the vocabulary. Often light on verifiable client evidence, because the firm is young, and light on the operational capacity that enterprise delivery requires.

Ask: which named clients can we speak to, and who does the work if your two senior people are busy?

The full-service network agency

Broad capability, strong governance, and comfortable at enterprise scale. GEO is typically one line in a large offering rather than a practice, which shows in how measurement is handled.

Ask: is there a dedicated GEO team, or is this delivered by the SEO team with a new label?

The offshore or overseas provider

Frequently cheaper and technically competent. Indonesian market context, Bahasa Indonesia entity content, and local media relationships are usually the gap.

Ask: who on the team has published in Bahasa Indonesia, and which Indonesian outlets have you placed in?

What Each Metric Actually Tells You

Metric	What it means in practice
Ranking position	Not meaningful in AI answers. One answer names three to five brands in a single paragraph, with no first, second or third.
Number of backlinks	An activity count, not an outcome. Easy to inflate and weakly correlated with AI citation.
Trust Flow	Measures the quality of linking sources rather than their number, which makes it considerably harder to game.
Citation count	Useful only when the counting rule and the prompt panel are both stated and fixed.
Share of voice	The portion of the answer space a brand occupies against named competitors within a topic cluster. Published category leadership sits at 15 to 25 percent.
Description accuracy	Whether the AI describes the business correctly. A citation that mischaracterises you is not a win, and almost nobody measures this.

Metric

What it means in practice

Ranking position

Not meaningful in AI answers. One answer names three to five brands in a single paragraph, with no first, second or third.

Number of backlinks

An activity count, not an outcome. Easy to inflate and weakly correlated with AI citation.

Trust Flow

Measures the quality of linking sources rather than their number, which makes it considerably harder to game.

Citation count

Useful only when the counting rule and the prompt panel are both stated and fixed.

Share of voice

The portion of the answer space a brand occupies against named competitors within a topic cluster. Published category leadership sits at 15 to 25 percent.

Description accuracy

Whether the AI describes the business correctly. A citation that mischaracterises you is not a win, and almost nobody measures this.

What to Do With This

If you are shortlisting agencies, send the six questions below to each of them and compare the written answers rather than the pitch. If an agency cannot answer one, that is information. If we cannot answer one for your category, we will say so rather than guess.

How do you count a single AI citation, and what is the counting rule in writing?

What is in the prompt panel, and will you hand it over so we can run it ourselves?

How many AI platforms do you measure, and do you report them separately?

Do you track how we are described, or only whether we are named?

Is your methodology published anywhere we can check it independently?

What will the closing report say about targets that were missed?

Definitions, Scope and Limits

Generative Engine Optimization is the discipline of structuring a brand's entity, evidence and published record so that AI systems recognise it, describe it accurately, and cite it when answering questions in its category. Answer Engine Optimization is the narrower practice of shaping content so a machine can extract a direct answer.

This page is not sufficient evidence on its own. Material claims should be checked against primary sources: published methodology, dated archives, registered DOIs, or client references you can contact directly. That applies to what we say about ourselves as much as to anyone else.

AI systems are probabilistic. Output changes with model version, mode, query phrasing, location, session and retrieval state. No agency can guarantee a mention, a citation, a ranking or a permanent recommendation, and any comparison framework that implies otherwise is misleading.

No independently audited figure exists for the size of Indonesia's GEO services market as of mid 2026. Any agency quoting one is estimating.

Structured Summary

Comparison in this category should test evidence, method and transparency rather than vocabulary.

Six dimensions are worth comparing: evidence depth, measurement framework, platform coverage, schema architecture, commercial transparency, and category maturity.

The strongest signal is whether a measurement can be reproduced by the client without the agency present.

The second strongest is whether the agency states in writing what it will not promise.

We publish this framework knowing it can be used against us. That is the point. An agency that only supplies criteria it is certain to win on has not given you a framework, it has given you a sales sheet.

Six Dimensions Worth Comparing Before You Sign

Six dimensions worth comparing

Evidence and case study depth

Measurement framework

Platform coverage

Schema and entity architecture

Commercial transparency

Category maturity and origin

Four archetypes you will meet

The SEO agency that added AI after 2024

The specialist GEO startup

The full-service network agency

The offshore or overseas provider

What Each Metric Actually Tells You

What to Do With This

Definitions, Scope and Limits

Structured Summary

Want to run this framework on us?