Looker 26.20 makes the new Explore default and Visualization Assistant GA
Google has turned on the new Explore and merge query experience as the default interface in Looker 26.20, and the same release moves several Gemini in Looker features and chart types to general availability. According to the Looker release notes entry of October 29, 2026, the changes are enabled automatically from October 26, 2026 on Looker (original) and Looker (Google Cloud core) instances running Looker 26.20, the version whose final deployment Google scheduled for October 25, 2026.
The practical change for most users is the Explore screen itself. Analysts who open an Explore on an upgraded instance now land in the new layout with its Configure panel, while administrators get two legacy switches that can send everyone back to the classic Explore for a limited period. The release also makes the Gemini-based Visualization Assistant and Explore summaries generally available, adds bullet charts, a KPI chart that replaces Single Value, table row grouping and a histogram chart, and renames one value in the Conversational Analytics System Activity dashboard in a way Google flags as breaking.
Google Cloud core customers on the Regular release channel will see none of this yet. The same release notes state that the Rapid channel and No Channel are on Looker 26.22, while the Regular channel remains on Looker 26.18.
··········
WHAT CHANGES IN THE LOOKER 26.20 EXPLORE EXPERIENCE.
The release notes describe the new Explore and merge query experience as generally available and the default. In the documentation for creating and editing Explores, the new layout is organized around a Configure panel with three tabs. All Fields lists every available field grouped by view, Fields in use shows the active fields and the selected chart type, and Style holds the visualization options.
The panel includes a search bar that can match field labels, descriptions, group labels and view labels, plus an Explore summary with a field count and a Go to LookML link for users who hold the see_lookml permission. Each field carries icons for pivoting, filtering and field information, and a More menu covers custom fields and dimension group timeframes. The data columns have their own menu for removing, filtering, pivoting, sorting, binning, grouping and copying values.
Merged results get a new entry point: users click Merge queries (join data), which opens a Join data page, so internal training material for the classic merge workflow will need new screenshots.
The documentation also lists limits that apply to anyone building reports in the new Explore. Pivoting supports up to 200 pivoted values, row totals and table calculations cannot be sorted once a row limit is reached, some sorting happens in the browser and can differ from other Looker content, and sorting applied in the Visualization pane is temporary and does not persist when the query is saved as a Look or dashboard tile.
String filters also behave differently. Clicking Done or clicking outside the filter no longer turns uncommitted text into a filter chip, so users have to press Enter, type a comma or paste values.
··········
LEGACY EXPLORE SWITCHES AND THE 27.6 REMOVAL DATE.
Looker's legacy feature schedule lists two switches introduced in 26.20. Use Legacy Explore and Merge Query Experience opens Explores and merged queries in the classic experience for all users, who then cannot switch to the new experience themselves. Use Legacy Explore and Merge Query Experience for Embed does the same for embedded Explores and merged queries.
The schedule sets the same path for both switches: default-off in Looker 27.2, then disallowed and removed in Looker 27.6. Two details decide what a given instance sees today. Instances that disabled the New Looker Explore and Merge Query Experience preview before 26.18 keep the legacy experience enabled by default. The embed switch is enabled by default until 27.2, and the documentation states that turning on the new experience in embedded contexts requires disabling both the embed switch and the non-embed switch.
........
Legacy switch | Introduced | Default-off | Removed |
Use Legacy Explore and Merge Query Experience | 26.20 | 27.2 | 27.6 |
Use Legacy Explore and Merge Query Experience for Embed | 26.20 | 27.2 | 27.6 |
Use Legacy Single Value Visualization | 26.20 | 27.2 | 27.6 |
........
For teams that embed Looker in customer-facing applications, embedded users stay on the classic Explore until an administrator changes the setting or the instance reaches 27.2, and the classic option disappears at 27.6. Testing embedded workflows against the new interface before 27.2 is the safer sequence.
··········
GEMINI IN LOOKER: VISUALIZATION ASSISTANT AND EXPLORE SUMMARIES.
The Visualization Assistant is now generally available on Looker (original) and Looker (Google Cloud core). The documentation describes it as a Gemini in Looker feature that turns text prompts into JSON formatting options and applies them to a visualization. It works on charts built on the HighCharts API, which covers most Cartesian charts such as column, bar and line charts.
Users open it from a visualization's Edit menu, in an Explore or while editing a Look or dashboard. They can type a request in the Ask anything field, apply generated suggestions, or open Edit Chart Config to change the JSON by hand. Clear Chart Overrides removes all custom changes after a confirmation, and an Autofix code option can repair invalid JSON. Google's prompting advice is to request one change per prompt, refer to series by name rather than position, and give specific value ranges and hex color codes for conditional formatting.
The documented requirements are specific. On Looker (original), Gemini in Looker must be enabled in the Admin settings and the instance must run Looker 25.2 or later; on Looker (Google Cloud core), Gemini in Looker is enabled in the Google Cloud console. The user's Looker role must include both the gemini_in_looker and can_override_vis_config permissions. The page also warns that editing default visualization options after using the assistant can produce unexpected behavior, including blank visualizations.
Change the chart to a column chart.
Add data labels to the Revenue series.
Color Revenue values below 10000 in #D93025 and values above 50000 in #188038.The release notes also list Gemini-generated Explore summaries as generally available. A summary describes the suggested purpose of an Explore and the types of data used in the query, and it appears only for Explores where the LookML developer has not already defined a summary. In the new Explore experience, users click the Explore Summary button on the Configure panel. The feature page, as fetched on November 1, 2026, still carries a pre-GA notice and lists Trusted Tester features, a role with the develop permission for at least one model, and the Looker Gemini role as requirements, so administrators should confirm in their own instance which of those conditions still apply after the GA announcement.
Teams already following Google's Gemini rollout across its data products will recognize the pattern seen in conversational analytics in Connected Sheets and in Gemini's chart generation in Google Sheets: the AI features sit behind an admin switch and role permissions, and the output is meant to be reviewed before it is shared.
··········
NEW LOOKER CHART TYPES: BULLET, KPI AND HISTOGRAM.
Bullet charts are now generally available. According to the documentation, a bullet chart needs exactly one dimension and one measure in the Explore. The target is set in the visualization options, either as a fixed value or from the measure itself, and up to three range thresholds can be defined, typically to mark poor, average and good performance. Looker sorts the range limits from lowest to highest automatically.
The KPI (Single Value) chart is generally available and replaces the Single Value option in Explores. It shows a primary value, can compare it with another measure displayed as a value, a change, a percentage change or a progress bar, and can add an area or column spark chart to show a trend. A matching legacy switch, Use Legacy Single Value Visualization, follows the same 26.20, 27.2 and 27.6 schedule as the Explore switches, and it is enabled by default on instances that never turned on the KPI preview.
The release also adds a histogram visualization that bins values to show how an attribute is distributed, and makes Table Row Grouping and the Grouping menu generally available so table data can be displayed in hierarchical groups.
........
Feature in Looker 26.20 | Status in release notes | Who needs to act |
New Explore and merge query experience | GA, default | All users; admins for legacy switches |
Visualization Assistant | GA | Admins enable Gemini in Looker and grant permissions |
Explore summaries with Gemini | GA | Admins check role and setting requirements |
Bullet charts | GA | Report builders |
KPI (Single Value) chart | GA, replaces Single Value | Report builders; admins for legacy switch |
Table Row Grouping | GA | Report builders |
Histogram chart | New feature | Report builders |
Dashboard Tile Groups | Preview | Admins enable preview feature |
Release Changelog | Preview | Admins and users |
........
··········
ADMIN CHANGES: SYSTEM ACTIVITY RENAME, CONNECTIONS AND CI.
Google marks one change as breaking. In the Conversational Analytics System Activity dashboard, on the Token usage tab, the value Conversational Analytics API has been renamed Looker API. Custom dashboards or alerts that filter on the old value may stop returning data, so administrators who track Gemini token consumption should update those filters.
On Looker (original), the Use Legacy Connections Page legacy feature and the legacy database connection workflow have been removed, and all connections now use the current workflow. A release note added on October 30, 2026 describes network diagnostics: when a database connection test fails, Looker now checks DNS and TCP connectivity automatically, and administrators can test reachability to any host and port from the Connections page.
For LookML developers, Continuous Integration now supports hub-and-spoke projects through local project imports, so a pull request on an upstream hub project can trigger tests on downstream spoke projects. This fits teams that keep shared definitions in a central project, the same kind of governed semantic layer that Databricks is building with metric view relationships.
Dashboard Tile Groups, which organizes related tiles into collapsible groups on newspaper-layout dashboards with LookML support, and the Release Changelog arrive in preview. Google Cloud core instances also gain a Legacy features page in the Admin panel.
··········
LOOKER 26.20 ROLLOUT: WHAT DEPENDS ON CHANNEL AND SETTINGS.
Google's release notes confirm that the new Explore and merge query experience is generally available and the default on Looker 26.20, that the Visualization Assistant, Explore summaries, bullet charts, the KPI chart and table row grouping are generally available, and that the legacy Explore, embed and Single Value switches become default-off in 27.2 and are removed in 27.6.
What a team actually sees depends on its platform and settings. Google Cloud core instances on the Regular channel were still on 26.18 at the time of the October 29 release notes, instances that disabled the new Explore preview before 26.18 keep the classic experience by default, and embedded Explores stay classic until both legacy switches are turned off or the instance reaches 27.2. The Gemini features require Gemini in Looker to be enabled and specific role permissions, and the Explore summary page still lists pre-GA conditions that administrators should verify.
Google has not published dates for the 27.2 and 27.6 releases in the entries reviewed, and the release notes do not describe any pricing change tied to these features. Data teams that rely on embedded analytics or on the Gemini token usage dashboard have the most concrete work to do now: test the new Explore with embedded users, update System Activity filters to the Looker API value, and review which roles carry gemini_in_looker and can_override_vis_config. Organizations comparing AI-assisted BI across vendors can set these requirements beside the connectors described in OpenAI's Data agent for BigQuery, Snowflake and Databricks.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org
