- Publish useful answers in accessible, indexable HTML.
- Connect each important claim to identifiable evidence.
- Separate visibility, visits, enquiries and confirmed appointments.
An evolving measurement framework
In May 2026, Google published a resource on optimisation for generative Search experiences. Its June announcement about Search Generative AI reports was updated on 31 August to indicate global rollout. These reports provide visibility observations; they are not a record of your sales appointments.
GEO aims to make expertise easy to find, understand and cite in answer systems. No format guarantees a citation. We recommend working on SEO foundations, then observing actual answers and visits using a stable protocol.
Build a page around a decision
Choose a precise question: do you need multiple agents, when should you use RAG, or how should a pilot be measured? Answer in the opening paragraphs. Then explain conditions, steps and limits. A decision table often adds more than a long series of similar phrases.
Link the page to the relevant service and two or three analyses that extend the subject. Readers should be able to move from a question to a method and then a discovery conversation. Use descriptive links and a coherent journey in each language.
Make information provenance visible
Distinguish INKWAY analysis, vendor announcements, independent studies and client cases. For a figure, explain who measured it, in what context and when. Do not attribute a vendor’s results to your company. Label fictional examples before readers encounter them.
Include a responsible author or organisation, publication date, genuine update date and bibliography. Structured data should describe visible content. Article markup does not turn an opinion into scientific research, and an extra file does not compensate for weak content.
Check access, then measure the journey
Check HTTP status, canonical URLs, languages, sitemap and robot directives. Useful text should exist in the HTML response. Search crawlers and crawlers used for other purposes can have separate settings; OpenAI documentation distinguishes OAI-SearchBot from GPTBot.
Maintain a dated list of representative questions. Record cited sources, wording and the limits of each observation. Compare findings with measured visits and then qualified enquiries in the CRM. Variation between answers requires repeated observations: a single screenshot cannot establish commercial impact.
| Level | Measure | Interpretation |
|---|---|---|
| Visibility | Observed impressions and citations | Presence measured in a specific context |
| Acquisition | Visits from search services | Arrival on site, subject to available attribution |
| Conversion | Enquiries, then confirmed appointments | Separate stages of the commercial journey |
Sources & methodology
This insight combines cited publications with editorial analysis. Illustrative examples are not client results. Vendor features and terms may change.