How to Make a Chart in Looker with AI

Creating charts in Looker used to require a fair amount of clicking and dragging, navigating fields, and understanding measures and dimensions. Now, you can build many of those same visualizations just by asking for them in plain English. This guide will walk you through how to use Looker's AI capabilities to create charts, helping you get from data to insights faster than ever before.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

What is Gemini in Looker and How Does It Work?

Gemini in Looker is Google's generative AI feature layered directly into the Looker platform. Instead of manually selecting fields and configuring chart settings, you can now type a description of what you want to see, and Looker will try to build it for you. This functionality transforms the user experience from a manual, click-based process to a conversational one.

The core purpose is to lower the barrier to entry for data analysis. You no longer need to be a grizzled data veteran or a LookML expert to explore your organization's data. If you can describe a business question in words, you have a solid starting point for getting an answer visualized as a chart or table.

For example, instead of searching for the ‘Sales’ measure and the ‘Order Date’ dimension, you can simply ask: “What were our total sales by month for the past year?” Looker's AI parses this request, identifies the relevant data fields, applies the necessary filters, and generates a visualization - often a line chart by default for time-series data - that answers your question.

Before You Start: Prerequisites for Using AI

Before you jump in, there are a few things that need to be in place. AI chart-building isn't magic, it relies on a well-structured data foundation set up by your data team.

• Gemini in Looker is Enabled: The most important piece is that your Looker administrator must have enabled the "Gemini in Looker" feature for your instance. If you don't see the AI prompt options, it likely hasn't been turned on.
• Proper Permissions: You’ll need the appropriate permissions to access ‘Explores’ in Looker - this is where self-serve analysis happens.
• A Well-Defined LookML Model: Looker's AI is only as smart as the data model it's built upon. For it to understand a term like "revenue," your company’s LookML developers must have already created a measure named revenue and provided descriptive labels. Clear and organized LookML is the backbone of successful AI-driven analysis.

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

How to Make a Chart in Looker with AI: A Step-by-Step Guide

Ready to build your first chart through conversation? Let’s walk through the process, from asking your initial question to refining the result.

Step 1: Navigate to an Explore

Everything starts in an ‘Explore.’ An Explore is a starting point for a query, designed around a specific business subject like "Sales," "Users," or "Web traffic." In the left-hand navigation pane, select the Explore that contains the data you're curious about. For this example, let's assume we're working in a "Sales Transactions" Explore.

Step 2: Use the "Look with AI" Prompt Bar

Once you're inside an Explore, you'll see a prompt bar either at the top or in the main panel prompting you to ask a question. This is your primary interface for interacting with the AI. You might also find AI-generated query suggestions, which can be great starting points if you're not sure what to ask.

Click into the prompt bar to begin typing your request.

Step 3: Formulate and Ask Your Question

This is the most critical part. The way you phrase your question will directly impact the output. Your goal is to be specific but to use natural, conversational language. Stick to the business terms you and your team normally use.

Examples of Good Prompts:

• "Show me total revenue by country for last quarter"
• "What are our top 10 best selling products by total units sold?"
• "Create a line chart of weekly active users over the past 6 months"
• "Show the number of support tickets created this month, grouped by product category"

Notice that these prompts are direct and clearly state the metric (revenue, users), the dimension (country, product), and often a timeframe (last quarter, last 6 months).

Tips for Writing Effective Prompts:

• Be Specific on Timeframes: Words like "last month," "this year," or "the past 90 days" are much clearer than vague terms like "recently."
• Use Correct Lingo: Use terms that align with the measures and dimensions in your LookML model. If the revenue field is named "Total Gross Revenue," using that exact phrase might get better results. If you don't know the exact names, sticking to common business verbiage is fine - the AI's goal is to translate it correctly.
• Start Simple: Begin with a straightforward question. You can follow up with more complex requests to iterate on your initial result.

After typing your question, press Enter or click the "Generate" button. Looker's AI will get to work, selecting what it believes are the best-suited metrics, dimensions, and filters.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

Step 4: Review and Refine the Chart

The AI will generate both a visualization and a data table for your review. Now it's time to check if it understood your request correctly.

Ask yourself:

• Are the correct Measures and Dimensions selected? Look at the panel and the data table. Did it correctly pull "Total Revenue," "Country," and filter for the "last quarter"?
• Is the Chart Type Appropriate? The AI usually does a good job of choosing a common chart type (like a line chart for time-series data or a bar chart for categorical comparisons). If its choice wasn’t what you wanted, you can easily change it either manually or with a follow-up question.
• Were the Correct Filters Applied? Verify that the appropriate timeframe filters or categorical filters were applied.

You can refine the result conversationally. After seeing the initial chart for product sales, you might type:

"Change it to a bar chart"

"Now, only show me the top 5"

"Filter for the 'Apparel' category"

Each new prompt acts on the result that came before it, letting you slice and drill into your data without manually searching for filters or settings.

Step 5: Save Your Visualization

Once you are happy with the chart, it's time to save it so you can refer to it later. In Looker, a saved visualization is known as a "Look."

Click the gear icon menu on the visualization and select 'Save...' From here, save your chart to a new or existing Looker Board (its term for a dashboard).

Common Challenges and How to Troubleshoot Them

AI is a powerful assistant, but it’s not infallible. You may encounter a few common hurdles.

Challenge: The AI Misunderstands Your Prompt

Sometimes, the AI interprets your question incorrectly, grabbing the wrong metric. For instance, you ask for "sales," and it returns "number of orders."

Solution: Rephrase your question for greater clarity. If your first try failed, be more explicit and try using a field name like 'Total Sales' rather than just 'sales'. Don't be afraid to try two or three different variations to see what connects.

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

Challenge: The Generated Chart Looks Nothing Like What You Asked

You might ask a simple question and get a completely unexpected or broken result.

Solution: This almost always indicates a deeper, structural issue with the data model, not the AI itself. The naming conventions might be inconsistent, or the relationships between data tables might not be well-defined. In these cases, it's best to reach out to your data team. Let them know what you were trying to accomplish and the result you got. This type of feedback helps them improve the underlying model, making it more user-friendly for everyone.

Challenge: You Require a More Complex Calculation

Maybe you need to see 'profit margin' (price minus cost), but that’s not a saved metric. You can also prompt the AI to 'Create calculations...' from a text prompt. Ask the AI Assistant:

"Create a calculation for profit divided by revenue"

It will then suggest a Looker Expression that you can use as a field. This is an advanced technique, but it showcases the power of language to perform an originally-complicated analysis.

Final Thoughts

Getting comfortable using Looker's AI tools shifts your focus from the mechanics of building to the actual analysis. By translating your queries to actionable charts, Looker accelerates your ability to analyze your company's data, uncover trends, and arrive at critical business decisions faster.

While tools like Looker pack a lot of power for business intelligence and allow teams to dive deep into their own data, we built Graphed for teams who want to move even faster without deep technical expertise. You can integrate all your data - Google Analytics, ad platforms (Facebook, Google Ads), CRM tools like HubSpot - and then simply convert your queries to dashboards in seconds. It’s less about building and more about getting immediate answers across all the systems that run your business, making analytics intuitive for your entire team.