The codebook defines the fields the pipeline codes and the values each can take. These pages list the public fields of release 2.7 whose answers it defines, with what each answer means, in the codebook’s own words. Fields it describes only in its coding rules are left out.
The records on the site today were classified before this release came out, and will be reclassified using it.
The full codebook is in the G3O pipeline repository.
Verdict
page_verdict
Level: page
Values
confirms_activity- The page documents a specific generative-AI activity AT THIS institution: its own GenAI tool, pilot, policy or guidance on GenAI, procurement, deployment, or institution-directed staff GenAI training. The text must be explicitly generative AI (LLMs, ChatGPT, Copilot, Gemini, AI chatbots, AI-generated content or equivalents) and tied to this institution, not only to a national programme or another body. The activity must be carried out or adopted by the named institution itself. Exclude activity by a parent, higher-level or neighbouring body (e.g. a district, county, regency, region, consortium or city department when the named institution is a village, ward or sub-unit), by a similarly named organisation or company, and pages by vendors that only use the place name. Generic 'AI', 'artificial intelligence', 'smart' systems, automation, analytics, prediction, transcription, translation, image or video recognition, or traditional machine-learning tools do not count unless the page says the tool generates text, images, audio, code or other content, or names a generative product or model (e.g. ChatGPT, Copilot, Gemini, Claude, an LLM, an AI chatbot or AI assistant).
ambiguous_genai- The page mentions AI at this institution, but it cannot be settled whether it is generative AI (e.g. a generic 'AI strategy', 'AI tools', analytics, traditional machine learning).
ambiguous_attribution- The page mentions generative AI, but it cannot be settled whether THIS institution is involved (e.g. a country-wide or regional GenAI initiative without confirmation that this institution takes part, or a page about a similarly named body).
other_body- The page documents a generative-AI activity, but it clearly belongs to a different body, not the named institution: a parent, higher-level or neighbouring body (e.g. a district, county, regency, region, consortium or city department when the named institution is a village, ward or sub-unit), a similarly named organisation or company, or a vendor or marketing page that only uses the place name.
event_only- The only GenAI content is a public-facing event, workshop, seminar, showcase or educational programme, with no documented adoption, procurement, deployment or governance by the institution itself (not a qualifying activity).
background_only- The page gives only general context (national AI strategy, vendor overview, news about GenAI in general) and documents no activity at this institution.
confirms_absence- The page states that this institution has no generative-AI activity, for example that it does not use GenAI tools. Different from no_genai, where the page simply has no relevant content.
no_genai- The page contains no generative-AI content relevant to this institution.
not_assessablenot_assessable, set mechanically when a page was not reviewed, never a classifier answer, not blocking institutionno, and left out of the base for institution-level rates
genai_mechanism_basis
Level: activity
Values
named_genai_product_or_model- The page names a generative product or model, such as ChatGPT, Copilot, Gemini, Claude or a named language model.
explicit_genai_label- The page calls the tool generative AI, a large language model or the like, without naming a product.
generative_capability_described- The page says the tool generates text, images, audio, code or other content, without naming a product or using a generative-AI label.
ai_chatbot_or_assistant_label_only- The only signal is that the page calls the tool an 'AI chatbot' or 'AI assistant'; under the GenAI boundary rule this counts as generative.