Skip to main content

Overview

Glean's usage analytics log provides a structured, easy-to-analyze record of key user actions such as searches, clicks, chats, and feedback. These logs enable clients to measure engagement, improve adoption, and build custom dashboards for internal analysis.

By default, Glean Customer Event logging is enabled and events are written as a unified stream called glean-customer-event to a log bucket in object storage (for example, Google Cloud Storage or Amazon S3) in your Glean project. For analytics and ad‑hoc queries, these logs are typically loaded into a data warehouse (such as BigQuery or Athena).

Glean delivers GCE logs as a single unified stream. Filtering by event type, connector, user, or content is not supported at the export layer. All events flow to the configured destination. Downstream systems (BigQuery, Athena, Splunk, and similar) must apply filters after ingestion. Plan destination index sizing, retention, and license consumption accordingly.

AWS Athena schema

For AWS deployments, the raw glean_customer_event_external Athena table uses a schema-on-read definition over the JSON event stream. The table declares a broad set of payload fields, but its columns do not guarantee that every deployment emits every event type or field. A declared field is NULL when the corresponding event does not populate it. The table definition and event producers evolve independently, so a field documented here might not yet be available in every deployment's table. Filter on jsonPayload.type before reading the matching payload struct.

Exports of these logs to customer-managed analytics destinations (for example, a BigQuery dataset in your own project) are not enabled by default. The exact project, dataset, and table names are set your organization during onboarding and feature enablement and may differ from the examples shown here.

If you would like to understand the options to consume these logs, or are unsure where your glean-customer-event logs are stored or exactly which project/dataset/table to query, contact your Glean Account team.

Glean Customer Event Types

The table below summarizes the key event types captured in the Glean customer event log, along with a brief description and examples to help you understand what each event represents and how it can be used in your analytics.

Event TypeDescriptionExample Action Logged
SEARCHMetadata about user search actions, including queries, filters, and surfaces.A user types a search query in the Glean New Tab Page.
SEARCH_CLICKCaptures clicks on search results.A user clicks the second result on the full-page web app.
SEARCH_FEEDBACKUser-submitted feedback on the quality or helpfulness of search results.A user rates search results as 3/5 and leaves a comment.
AUTOCOMPLETELogs interactions with autocomplete suggestions while typing.Autocomplete suggestions appear as a user types "onboard…".
CHATRepresents a chat turn sent to Glean.A user sends a question on the Glean homepage.
HUMAN_ONLY_CHAT_MESSAGECaptures metadata for a user-authored chat message, including its type, length, and whether it included a file upload.A user sends a message with a file attached.
VOICE_CHATCaptures real-time voice chat metadata, transcripts, and response latency.A user asks Glean a question using real-time voice.
CHAT_CITATIONSLists source documents referenced in the AI response to a chat.AI cites a Google Doc and a Confluence page in response.
CHAT_CITATION_CLICKLogs when a user clicks on a cited source in a chat response.A user clicks on the linked Google Doc cited by the AI.
CHAT_FEEDBACKCaptures feedback left by users on a chat message or assistant response.A user downvotes a Glean reply and adds a comment.
AI_ANSWERLogs the AI-generated response returned for a search query.

Note: Starting July 2025, AI Answers are now available in Workflow runs instead. See the Insights production queries for how to query both event types.
The AI generates a short answer summarizing multiple docs.
AI_ANSWER_VOTETracks upvotes/downvotes on AI-generated answers.A user upvotes an AI answer in a search result panel.
AI_SUMMARYRecords when an AI summary is generated.A user requests a summary of a long Slack thread.
AI_SUMMARY_VOTECaptures user votes on AI-generated summaries.A user clicks thumbs-down on a doc summary.
SHORTCUTLogs use of Glean shortcuts (for example, go/alias).A user visits go/standups which redirects to a calendar.
CLIENT_EVENTFrontend interactions like button clicks, page loads, and other UI activity.A user clicks "See more" on a suggested docs carousel.
WORKFLOW_RUNA single invocation of a Workflow, the underlying execution unit for Agent and Glean requests.A user runs an Agent.
LLM_CALLLogs LLM (Large Language Model) API calls made by Glean, including token usage, model information, provider details, audio usage, and request dimensions.An AI workflow makes a call to GPT-4 consuming 1000 input tokens.
MCP_USAGELogs MCP tool usage, including the tool, server, client application, authentication method, status, and duration.An agent calls a built-in MCP search tool.
ACTIONLogs action execution events, including status, timing, and associated workflow or session identifiers.An agent executes an action through a workflow.
ARTIFACTLogs artifact lifecycle actions and metadata, such as artifact type, version, surface, and export format.A user creates or exports an artifact.
WORKFLOW_CONVERSATIONCaptures the full conversation history within a workflow run, including messages and workflow input fields.Records all user and agent messages in a workflow session.
WORKFLOWDetailed logs providing execution details for a Workflow. DEPRECATED as of 2025-06-13: use WORKFLOW_RUN events instead.A user initiates a content review workflow.
GLEAN_BOT_ACTIVITYBackground actions or proactive behaviors by Glean bot, such as reminders or nudges.The Gleanbot nudges a user to follow up on an unread file.
PRODUCT_SNAPSHOTThis event type contains once-daily snapshots (not event logs) of entities in Glean, across seven types: USER, WORKFLOW, WORKFLOW_STEP, SUBSCRIPTION, ACTION_PACK, ACTION, and ACTION_INSTANCE. Each snapshot captures the state of the object at a point in time and is useful for understanding historical configuration and metadata for users, workflows, and tools.Captures user attributes such as departmentID and startdate.
Was this page helpful?