Our buyer guide publishes six questions we suggest sending to any agency in this category. Publishing questions is easy. Answering them in writing, where anyone can hold you to the answer, is the part most agencies skip.
This page is our answer to all six, for our own work. If our answer to any of them is weaker than another agency you are considering, that is useful information and you should act on it.
A citation is counted when a generative engine returns a response that either names the brand or draws text from a page on the brand's domain, for a query inside the frozen panel. A link and a bare mention are recorded separately and never merged into one figure.
Each panel question is run on each platform on a fixed schedule. Repeated appearances of the same page for the same question within one measurement cycle count once. Appearances of different pages for the same question count separately, because they show breadth of retrieval rather than repetition.
What this rules out: counting the same answer twice because it was run twice, and counting a mention of the category as a mention of the brand.
The panel holds 75 to 100 questions written in the words a real buyer would type, not keyword phrases. It is built in the first two months from the questions your sales team actually receives, the questions competitors are currently winning, and the questions that decide a purchase in your category.
It is frozen at month three and handed to you in full. From that point it does not change for the duration of the engagement.
Why freezing matters: a panel that changes between reports produces movement that reflects the panel, not the brand. If a report shows improvement and the questions were swapped, the improvement is an artefact. You can run the frozen panel yourself at any time and check our numbers without us present.
Six: ChatGPT, Google Gemini, Perplexity, Anthropic Claude, Microsoft Copilot and Grok. Google AI Overviews is tracked as a separate index because it behaves differently from Google AI Mode.
Every platform is reported on its own line. We do not publish a blended visibility score.
Why: an average hides the platforms where a brand does not appear at all. A brand strong in Perplexity and absent from ChatGPT will show a respectable average and a serious commercial problem. If a platform is paused or unavailable during a cycle, it is reported as paused, not as zero, because those are different findings.
Both, and they are separate lines in the report. Description accuracy covers how the engine characterises scale, service area, sector, capability and ownership.
A citation that describes a business inaccurately is not a win. An engine that names you as a small regional supplier when you operate nationally has cited you and damaged you in the same sentence.
Inaccurate descriptions are logged with the platform and the query that produced them, and correction becomes work in the following cycle. This is measured separately because the remedy is different: absence is fixed by publishing more, inaccuracy is fixed by publishing more precisely.
Yes. Three research reports carry registered DOIs and are indexed on Google Scholar and Zenodo: the AI Citation Rate Report 2026, State of SEO Indonesia 2026, and Digital Marketing Benchmark Indonesia 2026.
The benchmark report is built on a primary survey of 127 Indonesian businesses, cross validated against four research sources. 124 data points were validated and five were rejected for lack of a verifiable source, and the rejected five are named in the report.
This page is itself part of that answer. A methodology that exists only inside a proposal cannot be checked by anyone.
It will name them. The closing audit lists every milestone commitment made before signature, marks each as met or missed, and states the reason for each miss.
Milestone commitments are published before contracts are signed and are split into two groups: targets we accept being held to, and targets where measurement is not yet mature enough to promise a number. The second group is stated as such rather than quietly omitted.
What this costs us: a document that lists only achievements is easier to sell. It is also worth less to a buyer, because a report that has never recorded a failure is not a record, it is marketing.
RoGEO is the framework these answers sit inside. It has three pillars, and each is reported as its own figure rather than combined into a single score.
How often the brand is retrieved and cited, counted per platform against the frozen panel.
How much of the answer space the brand occupies against named competitors, expressed as share of voice. Published category leadership sits at 15 to 25 per cent.
What reaches the business from AI-referred traffic, tracked where analytics permits attribution.
| 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. |
| 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. |
| Description accuracy | Whether the engine describes the business correctly. Almost nobody measures this. |
| Trust Flow | Measures the quality of linking sources rather than their number, which makes it considerably harder to game than a raw count. |
| Number of backlinks | An activity count, not an outcome. Easy to inflate and weakly correlated with AI citation. |
AI citation has a crawl latency of two to three months. The month three baseline is usually close to zero, and that is expected rather than a failure.
Movement typically becomes visible around month six. Category position, meaning consistent appearance against named competitors, typically arrives around month twelve.
In contested categories the first six months are usually spent building the entity and the evidence base rather than chasing citations, because citations follow the material rather than the other way round. We say this before contracts are signed, not when a client asks why nothing has moved.
It does not guarantee citation in ChatGPT or any other engine. It does not guarantee ranking, lead volume, pipeline value, or revenue. No vendor can guarantee these, because generative engines change their retrieval behaviour without notice.
It does not measure by article count or backlink volume. Both are activity measures, and both can be inflated without moving a single citation.
It does not produce a single blended score. Every attempt we have seen to compress six platforms and three pillars into one number has hidden something the client needed to see.
We use cookies to ensure the website runs optimally and to help us understand how you use our services. You can choose which categories to allow. Read our Privacy Policy.
Required for basic website functionality. Cannot be disabled.
Help us understand how visitors interact with the website. Data used anonymously.
Used to display relevant ads and measure campaign effectiveness.
Enables live chat, social media integrations, and language preferences.
