Stop Paying for One AI Score: Measure Visibility Like This
TL;DR: To measure AI visibility properly, track each priority prompt separately for each engine and language, and record two things every time: whether your brand is named and whether your own page is cited. A single blended score is a useful headline, but it should never be the number you pay against. Five columns turn a score into a measurement.
A blended score can rise while the engine your buyers use stays empty. Here is what a report needs before it is worth a retainer.
1. One row per prompt, with the wording frozen
AI visibility can only be compared over time if the exact prompt wording stays the same. That is why serious studies freeze wording: Discoverable submitted each of its 1,000 questions with identical wording across six AI services and three rounds (Discoverable, 16 Sep 2026). Agree the prompt list in writing, then measure the same list every cycle.
2. A column for every engine your buyers use
An average across engines hides the engine that matters most to your buyers. In Indonesia, ChatGPT held 84.91% of AI chatbot share in September 2026, with Perplexity at 8.08% and Gemini at 6.13% (Statcounter). Google’s AI Overviews and AI Mode sit inside Search and need their own columns, and Google’s own Search Console counts them inside overall “Web” traffic rather than reporting them separately (Google Search Central).
3. Named and cited, recorded separately
Being named in an AI answer and having your page cited as a source are two different results. In Discoverable’s UK study of 1,000 questions across six AI services, Perplexity did not name the business ranked first on Google in 74.9% of eligible answers; counting citations as presence brought that down to 55.9% (Discoverable, 16 Sep 2026). A score that merges the two cannot tell you which problem to fix. We covered this gap in cited is not recommended.
4. A language label on every prompt
Indonesian and English prompts behave differently, so they should never be averaged together. Google reports that 84% of Gemini prompts in Indonesia use the local language (The Gemini Report, Google internal data, January to March 2026). A brand that wins in English and is absent in Indonesian can still show a healthy blended score.
5. A dated change log next to the numbers
AI results move even when your pages do not, so every score needs a date and a note of what changed. In early September 2026, Google’s AI Mode briefly showed almost no links before they were restored the next morning (Search Engine Roundtable). Without a log, a vendor can claim credit for an engine change, or blame one.
Blended score vs prompt ledger
| Blended score | Prompt ledger | |
|---|---|---|
| Shows the headline trend | Yes | Yes, as a summary row |
| Shows which engine is empty | No | Yes |
| Separates named from cited | Usually no | Yes |
| Separates Indonesian from English | Usually no | Yes |
| Safe to pay against | No | Yes, on an agreed prompt set |
Where White Wood stands
White Wood measures AI visibility with SOLEDAD, our own engine, which records where a brand is named and where it is missing across AI engines and rolls that up into a Visibility Index. We publish an index too, and we think it is a fine headline. Our position is simply that the index should sit on top of the prompt rows, not replace them, and that fees should be tied to the rows. Read more on reading AI visibility scores by engine.
FAQ
How do I measure AI visibility? Freeze a list of priority prompts, check each one per engine and per language on a fixed schedule, and record whether your brand is named and whether your own page is cited. Summarise into a score only after that.
Are AI visibility scores useless? No. They are summaries. They mislead only when they are the only number in the contract.
Can Google Search Console show my AI Overview traffic? Not separately. Google says AI Overviews and AI Mode traffic is included in the Performance report under the “Web” search type.