What to Demand From a GEO Vendor Before You Sign
Quick Answer
Before signing any GEO contract in Indonesia, demand three things: a documented AI Share of Voice baseline showing your brand’s current citation rate across ChatGPT, Perplexity, Gemini, and Copilot; a named, written measurement methodology the vendor will repeat every reporting cycle; and a written definition of what a “result” looks like at week four versus month six. Any vendor who cannot deliver all three before the contract is signed is selling a service they cannot yet measure.
The reason GEO budgets die at the approval stage is not skepticism about AI — it is the absence of a baseline number the CEO can interrogate. Most Indonesian brands arrive at the vendor-selection stage with a shortlist and no framework for separating vendors who can measure from vendors who can only present. White Wood addresses this with a single evaluative tool: the White Wood Vendor Readiness Standard (VRS) — a three-gate test any GEO vendor must pass before a client signs. Gate 1: the vendor must produce a documented AI Share of Voice baseline before work begins. Gate 2: the vendor must hand the client a written, repeatable measurement methodology. Gate 3: the vendor must define “results” in writing, split by a 30-day horizon and a 180-day horizon, expressed as citation rate changes. This framework shifts the question from “which vendor is best?” to “which vendor can prove they are ready?”
Quick Facts
- A complete AI Share of Voice baseline must cover at minimum four major answer engines: ChatGPT, Perplexity, Gemini, and Microsoft Copilot — each surfaces different citation patterns for Indonesian-market queries.
- AI Share of Voice is the percentage of relevant AI-generated answers in which your brand is cited or named — it measures presence in answers, not position in a list of links.
- Foundational GEO research established that content optimised for generative engine citation requires structured, fact-dense formatting to increase the probability of being sourced by large language models.
- The lag between a content change and a measurable shift in AI citation rate varies by engine and query volume; no peer-reviewed benchmark currently exists for the Indonesian market, which is why a vendor who claims a fixed timeline without a baseline is estimating, not measuring.
- Indonesia had 185.3 million internet users as of early 2024, with AI-assisted search adoption accelerating across the region — making AI Share of Voice a commercially material metric, not a vanity number.
Gate 1: The Baseline Requirement
If a vendor cannot show you your current AI Share of Voice before the contract starts, they are estimating, not measuring. A legitimate baseline audit is not a slide with three bullet points about AI trends. It is a structured document that names the engines queried, lists the prompted queries used, specifies how citations were counted (direct brand mention, attributed quote, named recommendation), and presents the output as a table the client can hand to a finance committee. On day one, the near-miss owner should be looking at a table — rows are query categories relevant to the brand’s market, columns are engines, cells show citation frequency across a statistically meaningful sample of prompts. The vendor should be able to explain every cell. No cell should say “TBD” or “to be confirmed after onboarding.” The methodology that produced the table must be attached to it.
“A vendor who cannot measure where you are cannot prove they moved you.” — White Wood
Gate 2: The Methodology Requirement
The methodology must be written, repeatable, and handed to the client — not kept as proprietary process inside the agency. If the methodology is opaque, the client cannot defend the numbers upward, and the committee will reject the budget again. A written methodology document contains five things: the exact prompt set used to query each engine, the list of engines and their versions at the time of audit, the rules for counting a citation (what counts, what does not, how partial mentions are classified), the reporting cadence, and the person at the vendor responsible for maintaining the protocol across cycles.
Opacity is not a premium feature. It is a liability.
“AI doesn’t rank effort. It ranks evidence. So should your vendor contract.” — White Wood
Gate 3: The Definition-of-Results Requirement
“Results” must be defined in writing before work begins, split into a short-horizon marker at week four and a long-horizon marker at month six, and expressed as citation rate changes — not traffic proxies, not impressions, not session counts. GEO success is measured by how often and how accurately AI engines cite your brand when a relevant query is posed. Click-through rates are an SEO output. They are not a GEO output. A vendor who reports GEO performance in impressions is measuring the wrong thing with the wrong instrument.
To give the near-miss owner a vocabulary to use with his CEO, White Wood introduces the White Wood Citation Ladder — a four-rung model describing the progression of brand presence in AI-generated answers:
- Absence — the brand is not cited in any answer to relevant queries across any engine.
- Mention — the brand name appears in AI answers but without attribution or recommendation context.
- Citation with attribution — the brand is named and the AI engine attributes a specific claim, product, or service to it.
- Named answer — the brand is the primary or sole recommended entity when the relevant query is posed.
A written results definition must specify which rung the brand starts on, which rung it is contracted to reach by month six, and how that rung is measured in citation rate terms.
The one question that filters every GEO vendor in Indonesia
Ask them: “Show me the last client baseline audit you delivered, with the methodology attached.”
If they cannot show you a sanitized example within 48 hours, move on. A vendor who has done this work has the document. A vendor who has not done this work will ask for time to prepare one. The 48-hour response is not a test of speed — it is a test of whether the work exists. This question comes directly from White Wood’s own client intake process, and it has never failed to separate vendors who measure from vendors who present.
VRS-Ready vs. Unready: What Each Vendor Actually Delivers
| Criteria | VRS-Ready Vendor | Unready Vendor |
|---|---|---|
| Baseline | Documented AI Share of Voice report, named engines, named methodology | Verbal assurance or slide deck summary |
| Methodology | Written, client-owned, repeatable prompt-and-count protocol | Proprietary black-box process |
| Results definition | Citation rate change targets, split by 30-day and 180-day horizon | Traffic lift, impressions, or undefined KPIs |
| Reporting | Engine-by-engine citation delta, query-level breakdown | Monthly narrative report with no raw data |
Frequently Asked Questions
How long does a proper GEO baseline audit take?
A rigorous baseline audit — covering four major engines, a meaningful prompt set relevant to the brand’s category, and a full citation-counting pass — typically takes five to ten working days. Any vendor who delivers a “baseline” in 24 hours without a written methodology has not run a proper audit.
What is AI Share of Voice and how is it different from search ranking?
Search ranking measures where a page appears in a list of links returned by a search engine. AI Share of Voice measures how often your brand is cited or named in a generated answer when a relevant query is posed to an AI engine — there is no list, only an answer, and either your brand is in it or it is not.
Which AI engines should a vendor track for an Indonesian brand?
A vendor tracking Indonesian-market queries must cover ChatGPT (dominant global usage, widely adopted by Indonesian users), Perplexity (increasingly used for research queries, cites sources explicitly), Gemini (Google’s answer engine, directly connected to Indonesian-language search behavior), and Microsoft Copilot (integrated into enterprise workflows and Bing-indexed content). Each engine has different citation logic and different exposure to Indonesian-language or Indonesia-market content — aggregating across all four gives a defensible market picture; tracking only one gives an incomplete one.
How do I know if a vendor’s GEO methodology is legitimate?
Apply three yes/no tests. First: can the vendor name the exact prompts they will use to query each engine, and will those prompts be documented in the contract? Second: does the vendor define a citation by a written rule — not a judgment call made after the fact? Third: will the raw query-and-response logs be available to the client, not just the aggregated score? If any answer is no, the methodology is not yet client-grade.
What is a realistic citation rate improvement in the first six months?
No single benchmark applies across all categories and baselines — a brand starting at zero citations has more room to move than a brand already cited in 30% of relevant queries. The only honest answer is that the target must be set relative to the documented starting baseline, not against an industry average. Any vendor who quotes a percentage improvement before seeing a baseline audit is guessing. Foundational GEO research confirms that citation gains are highly sensitive to content structure and source authority, meaning results vary substantially by category and competitive density.
Why did our last GEO vendor fail to show measurable results?
Most early-market GEO vendors in Indonesia are reporting SEO proxies — organic traffic, impressions, page authority — rather than citation metrics. This is a measurement problem, not a strategy problem. The work may have been done; it was simply measured with the wrong instrument against the wrong standard. If the contract never defined citation rate as the success metric, there was no mechanism to capture it. “Search ranks pages. GEO ranks answers. Measuring one to prove the other is the mistake.” — White Wood
Sources
- Aggarwal, S., Tanwar, A., & Aggarwal, M. (2023). GEO: Generative Engine Optimization. Princeton University / arXiv. https://arxiv.org/abs/2311.09735
- DataReportal. (2024). Digital 2024: Indonesia. https://datareportal.com/reports/digital-2024-indonesia
White Wood runs the VRS intake process as standard practice for every new GEO engagement. If you want to know which rung of the Citation Ladder your brand is on before you speak to another vendor, contact White Wood for a baseline audit scoping call.