Disclosure: This article is published by Upsolve AI. Where our product appears alongside others, the comparison is based on publicly available information as of November 2026.

Looker Conversational Analytics is the chat interface inside Looker: a user asks a question in plain English, Gemini picks the LookML fields and filters, and Looker writes the SQL. Usage carries no token charges today, but that promotion ends on January 31, 2027. From February 1, questions draw on a metered token pool, and archived copies of Google’s pricing page show three different versions of that pool since May.

Two things need checking before then: how much of your LookML the feature can use, and how many tokens your questions consume. The first matters more.

Looker Conversational Analytics vs the Rest of Gemini in Looker

“Gemini in Looker” is the umbrella. Conversational Analytics is one feature under it, and it comes in three forms that behave differently:

• Explore conversations: typically one user, one Explore, no custom context. The Start a conversation button isn’t available in the new Explore experience.

• Explore data agents: up to five Explores, plus instructions, glossaries and verified queries. They can be shared, and they’re the only form where verified queries work.

• Dashboard agents: one shared agent per dashboard, scoped to it and its query-linked tiles. No Advanced Analytics, and agents on LookML dashboards can’t take custom instructions.

The other Gemini features are mostly analyst tools, each with its own admin setting and permission:

Insight Assistant is a separate feature, yet System Activity lists it as a surface on the token usage tab, so include it when you baseline usage.

The mode setting also affects cost. Google says Thinking mode isn’t more accurate than Fast mode; it spends extra tokens to show the reasoning, and those thoughts count as output tokens. Fast became the default in October 2026.

How It Depends on LookML, and Where That Breaks

Because Looker composes the query, joins, access grants and user attributes apply as they do in an Explore. The price is that the agent only reaches what your model exposes. These limits surprise LookML developers:

• Custom fields and table calculations are invisible. Google says business logic kept in them is inaccessible. Promote the common ones into LookML.

• No pivots. Multi-dimension results come back in long format, and verified queries can’t contain pivots or custom fields either.

• Filter-only fields can’t be set. The docs say the feature cannot set the value of a field defined with the parameter or filter parameters. In practice, expect trouble with Explores built around metric selectors or templated filters, where Liquid needs a user-chosen value. Test these first.

• Single-chart questions. Trends, breakdowns, top values and single metrics work. Forecasting, correlation and anomaly detection are outside the standard path. Advanced Analytics covers some of this with Python, but only in data agents, and not on Google Cloud core instances using CMEK or VPC Service Controls.

• 50,000 rows per query. That’s the documented cap.

Your LookML is now prompt context with a price on it. Every exposed field description is input the model reads, so hiding fields improves answers and cuts the bill at once.

The feature fetches the Explore schema at the start of a request, and LookML metadata counts as input tokens. Google’s advice on field bloat is to hide keys and technical fields, and to consider an agent-specific Explore built with extends. It also warns that free-form instructions go stale, and recommends verified queries, no more than 30 to 50 per agent.

Governance has two gaps worth raising with security early. By default, raw prompts and responses aren’t recorded in System Activity or Cloud Logging; admins see them only through a preview feature, for users who consent. Admins also can’t bulk-delete history or set a retention period. Separately, the feature isn’t yet in the FedRAMP High or Medium boundaries.

Looker vs Looker Studio: Conversational Analytics in Data Studio

Looker Studio went back to the name Data Studio in April 2026, and Google’s docs now say “Gemini in Data Studio (née Gemini in Looker)”. So an older page about “Gemini in Looker” may describe a different product. The two share a feature name and little else:

• Looker is the paid platform. Its chat is grounded in LookML, on both Google Cloud core and Looker (original). Looker (original) must be Looker-hosted for Gemini features.

• Data Studio has two experiences. The new one is open to all users, but only talks to data agents built in BigQuery and shared to Data Studio.

• Legacy Data Studio chat handles CSV, Sheets, Looker and BigQuery sources, and needs a Data Studio Pro subscription at $9 per user per project per month. Legacy agents and conversations can’t be migrated, and Google says the switch stays available only “for a time”.

The trap: Looker data sources aren’t supported in the new experience. A team chatting with Explores from Data Studio today should plan to rebuild those agents in Looker before legacy goes away. Also, complimentary Data Studio Pro licenses aren’t available for Google Cloud core instances created after August 1, 2026.

Looker Conversational Analytics Pricing From February 2027

You need Looker first. Google’s pricing page says call sales, though its SKU list shows Standard edition at $5,000 and users at $30, $60 and $125. The list states no billing period. Our Looker pricing guide treats Standard as about $5,000 a month, and the detailed breakdown covers the rest of the platform.

Tokens are metered on top. The terms as of November 2026:

Tokens pool across the instance, reset monthly and don’t roll over. Standard edition is for fewer than 50 users, so it always sits on the floor. The floor equals 60 internal users, or 240 embed users.

The meter has already changed twice

Pages written a few weeks ago can be wrong:

• May 2026: per license type (Viewer 1M, Standard 2M, Developer 4M input), billing from October 1, 2026.

• August to October 2026: per subscription tier (Standard 60M, Enterprise 300M, Embed 1.2B input). By late September the October date had gone, replaced by a promise of 90 days’ notice.

• November 2026: per activated user with a floor, billing from February 1, 2027.

None of these allowances has been billed. But the current model shifts allowance from small instances to large ones, and hits small embedded deployments hardest.

An Embed instance with 500 embed users and 20 internal users went from 1.2B input tokens to 145M, about 88% less. An Enterprise instance with 100 users went from 300M to 100M. With 500 users it rose from 300M to 500M.

A worked example

Google publishes no tokens-per-question figure, and it will vary with schema size and conversation length, so these numbers are illustrative.

Assumptions: 80,000 input and 3,000 output tokens per question, averaged across follow-ups, with input and output allowances as separate buckets. A question above the allowance costs $0.24 + $0.06 = $0.30.

At modest token use, overage is small next to the platform fee. On these assumptions output runs out first for internal teams, because the allowance ratio is 50:1. And the result swings hard with tokens per question, so that number matters more than the rate.

Don’t guess it. The token usage tab in System Activity shows estimated tokens by agent, user and conversation, refreshed every 24 hours. You have until January 31 to measure a real number.

Two other meters sit beside this one:

• Your warehouse. Answers run as real queries, and sampling values and fuzzy search can add database load. Hard-code suggestions or point suggest_dimension at something cheap.

• API quotas. The Looker-managed MCP server costs nothing extra, but agent tool calls consume API quota, and Standard edition includes 1,000 query-based API calls a month. An MCP-connected agent could use that up quickly.

When a Dedicated Data Agent Fits Better

Stay native if your important data is already modeled in LookML, your users work in Looker, and most questions fit a single chart. Governance is already in place, and buying another tool to avoid some model cleanup rarely makes sense.

A dedicated agent such as Upsolve fits better in four situations:

• The data isn’t in LookML. Modeling comes first with Looker, and each model can use only one connection. Google’s API docs describe multi-Explore questions as separate queries compared afterward, not joins. Upsolve needs no existing semantic layer and works across Snowflake, BigQuery, Databricks, Redshift and Postgres at once.

• People want answers outside Looker. Google documents Looker, iframes, its APIs and Gemini Enterprise (Preview). A chat agent in Slack or Teams means building on those APIs. Upsolve ships a Slack agent, Teams, an SDK and MCP. MCP alone isn’t the deciding reason, since Looker’s managed MCP server went GA on October 26, 2026.

• The chat is part of your product. Looker supports signed embedding, but it’s iframe-based, requires a Looker-hosted instance, and applies only General theme options on the Conversations page. Upsolve’s Team tier adds an embedding SDK, multi-tenancy and row-level security.

• You want testing and traces in your own hands. Looker offers verified queries, a preview pane and thumbs feedback. Upsolve’s golden assets double as test cases (tests are in beta), scored with LLM-as-judge, with step-by-step traces for each answer. Judge the results on your own questions.

Upsolve Team is $2,000 a month with 10,000 credits, about 1,000 analytical questions, and overage is $0.05 per credit. If your data and users already sit inside Looker, its chat is the natural default. A dedicated agent is worth paying for when it reaches data and users Looker doesn’t. For the wider field, see our AI analytics tools roundup and Looker alternatives.

Looker Conversational Analytics vs Data Studio vs Upsolve

As of November 2026.

Frequently Asked Questions

What is conversational analytics in Looker?

It’s a Gemini-powered chat feature that answers plain-English questions from your Looker Explores and dashboards. Gemini chooses the fields and filters, and Looker builds the query from LookML.

Is Looker Gemini the same as Conversational Analytics?

No. Gemini in Looker is the full set of AI features, including the Visualization Assistant, LookML Assistant and Insight Assistant. Conversational Analytics is the chat feature within that set, and it’s the one with published token pricing.

How much does Looker Conversational Analytics cost?

There are no token charges through January 31, 2027, within fair usage limits. From February 1, 2027, usage above your allowance costs $3 per 1M input tokens and $20 per 1M output tokens, on top of your Looker subscription. Check the page before budgeting, since the terms have changed this year.

Does Conversational Analytics work without LookML?

Not in Looker. It’s built on Explores, so someone has to model the data first. In Data Studio, the new experience uses BigQuery data agents instead, with no LookML involved.

Can I use Conversational Analytics in Looker Studio?

Yes, under its new name, Data Studio. But the new experience doesn’t support Looker data sources. To chat with Explores, use Looker itself.

Want to see how a data agent handles questions outside your LookML? Book a demo or compare plans.

Sources

1. Google Cloud: Looker pricing, including data token allowances and billing dates, read November 4, 2026. https://cloud.google.com/looker/pricing

2. Google Cloud: Conversational Analytics in Looker overview. https://docs.cloud.google.com/looker/docs/conversational-analytics-overview

3. Google Cloud: Gemini in Looker overview. https://docs.cloud.google.com/looker/docs/gemini-overview-looker

4. Google Cloud: Query an Explore in natural language (Fast and Thinking modes). https://docs.cloud.google.com/looker/docs/conversational-analytics-looker-data

5. Google Cloud: Create and manage Explore data agents. https://docs.cloud.google.com/looker/docs/conversational-analytics-looker-data-agents

6. Google Cloud: Best practices for configuring Conversational Analytics. https://docs.cloud.google.com/looker/docs/conversational-analytics-looker-best-practices

7. Google Cloud: System Activity dashboards, token usage tab. https://docs.cloud.google.com/looker/docs/system-activity-dashboards

8. Google Cloud: Looker release notes. https://docs.cloud.google.com/looker/docs/release-notes

9. Google Cloud: Looker (Google Cloud core) feature differences. https://docs.cloud.google.com/looker/docs/looker-core-feature-differences

10. Google Cloud: Embedding Conversational Analytics. https://docs.cloud.google.com/looker/docs/conversational-analytics-looker-embedding

11. Google Cloud: Looker-managed MCP server. https://docs.cloud.google.com/looker/docs/mcp

12. Google Cloud: LookML connection parameter for models. https://docs.cloud.google.com/looker/docs/reference/param-model-connection

13. Google Cloud: Conversational Analytics API known limitations. https://docs.cloud.google.com/gemini/data-agents/conversational-analytics-api/known-limitations

14. Google Cloud: Platform SKU list, filtered to Looker. https://cloud.google.com/skus?filter=looker&currency=USD

15. Google Cloud: Conversational Analytics in Data Studio overview. https://docs.cloud.google.com/data-studio/conversational-analytics-overview

16. Google Cloud: Converse with Data Studio data (Looker source limit). https://docs.cloud.google.com/looker/docs/studio/conversational-analytics-looker-studio

17. Google Cloud: Data Studio welcome page, on the rename from Looker Studio. https://docs.cloud.google.com/data-studio/welcome

18. Google Cloud: Data Studio product and pricing page. https://cloud.google.com/looker-studio

19. Internet Archive: Looker pricing page as captured May 16, 2026. https://web.archive.org/web/20260516215114/https://cloud.google.com/looker/pricing

20. Internet Archive: Looker pricing page as captured October 1, 2026. https://web.archive.org/web/20261001061147/https://cloud.google.com/looker/pricing

21. Upsolve AI: Pricing. https://upsolve.ai/pricing

22. Upsolve AI: Homepage and product overview. https://upsolve.ai/

23. Upsolve AI docs: Upsolve for Slack. https://docs.upsolve.ai/slack/overview

24. Upsolve AI docs: Golden Assets. https://docs.upsolve.ai/ai-agent-builder/golden-assets

25. Upsolve AI docs: Agent Observability. https://docs.upsolve.ai/ai-agent-builder/agent-observability

26. Upsolve AI: Looker pricing guide. https://upsolve.ai/blog/looker-pricing

27. Upsolve AI: Looker pricing detailed breakdown. https://upsolve.ai/blog/looker-pricing-detailed-breakdown