# Global Gravity: SEO and AI visibility measurement

Method note: 2026-10-05. This is a measurement definition, not a fresh visibility score or customer-result claim.

## Three manifolds, separate evidence and observations

Demand Manifold describes real questions by market, language, role and buying stage. Brand Asset Manifold records applicable brand information and its supporting sources. Target Positioning Manifold defines the contexts and descriptions the brand seeks to earn. Evidence supports those descriptions; observed platform answers test them. A vector similarity is a diagnostic input, not a probability of recommendation or a permanent Top-3 guarantee.

Use existing assets before adding new material. Distinguish missing detail, valid contextual differences, verified contradictions and unresolved claims. Missing repository evidence does not establish that a business claim is false.

## Denominators must remain explicit

- Search Console Web clicks and impressions describe property-level search performance. Disclosed query rows can omit private queries and must not replace property totals.
- A dedicated Generative AI report, where available to the property, describes its documented AI surface. Do not add overlapping AI impressions to Web totals.
- GA4 sessions and events describe visits after arrival, subject to consent, referrer and tagging limits. Direct traffic does not establish AI provenance.
- Bot access, indexing eligibility, a brand mention, a recommendation, an exact URL citation and a qualified enquiry are distinct observations.
- A Top-3 rate is applicable only to explicit ordered recommendations under a fixed panel. Unordered mentions have no invented rank.
- A network failure or budget skip is missing data, not a negative answer. A successful answer without the brand is a valid negative observation.
- Preserve platform, model/surface, locale, region, prompt, time, source answer and exact citation URL. Repeated prompts and platform/time correlation are not automatically independent samples.
- A server-confirmed submission, non-test deduplicated contact and business-qualified enquiry are separate states. They must not be counted as the same customer or treated as incremental revenue without evidence.

## Historical snapshots

The [June GSC opportunity map](https://www.global-gravity.com/ai/gsc-opportunity-map.md) and [June intent/asset map](https://www.global-gravity.com/ai/semantic-coverage-map.md) retain their original dates and context. They are historical snapshots, not current search performance or a reproduced vector-coverage benchmark. No new October AI recommendation score is reported here.

## Official references

- [Google AI search and websites](https://developers.google.com/search/docs/appearance/ai-features)
- [Generative AI performance report](https://support.google.com/webmasters/answer/16984139)
- [Google AI inclusion](https://support.google.com/webmasters/answer/16908024)
- [OpenAI bot purposes](https://developers.openai.com/api/docs/bots)
- [Perplexity crawler purposes](https://docs.perplexity.ai/guides/bots)
- [GEO original paper](https://arxiv.org/abs/2311.09735): its experiment is not a universal present-day uplift guarantee.

## Buyer paths

[GEO scope](https://www.global-gravity.com/en/services/geo) · [Method and inputs](https://www.global-gravity.com/en/technology) · [Products](https://www.global-gravity.com/en/products) · [Request an audit](https://www.global-gravity.com/en/get-audit)

The corresponding buyer modules are available in zh, en, ja, ko, de, fr, es, pt and ar. This machine-readable note does not replace those localized pages.
