Using Generative AI in BigQuery to Analyze GA4 Data at Scale

Unlock deeper insights from GA4 with Generative AI in BigQuery. Learn how to analyze multilingual feedback, summarize page performance, and apply NLP tasks using SQL—no external tools required.

Shreya Banker

Data Scientist

Data Analyst enthusiast. More than 7 years of exposure in Data Analysis and Software programming. I am a highly motivated, versatile IT professional with experience in Data Analysis, Visualization and Database Management. I look for the hardest problem to solve and where I can learn and develop the most. I love a challenge and never run from a difficult task. I'm determined to succeed, and I look forward to what life has to offer.

Google Analytics (GA4) has revolutionized digital analytics with its event-based data model, flexible schema, and seamless integration with BigQuery. But for many teams, one major challenge remains:

How do you analyze unstructured data like customer feedback, search terms, or multilingual inputs at scale?

You can analyze unstructured GA4 data directly in BigQuery using generative AI. With AI.GENERATE() and Vertex AI models, teams can translate, classify, score sentiment, and summarize text-based data using SQL, without external tools or pipelines.

Why This Was Difficult Before

Traditionally, this required:

Exporting data from BigQuery
Using external NLP libraries or APIs (like Google Cloud Translation or sentiment scoring)
Writing complex Python scripts or custom pipelines

This created three key challenges:

Slow analysis cycles
Heavy reliance on engineering teams
Disconnected analytics workflows

What Changed: Generative AI in BigQuery

With BigQuery’s built-in support for Vertex AI models via AI.GENERATE(), you can now analyze unstructured GA4 data directly within SQL.

This includes:

Translation
Topic classification
Sentiment analysis
Summarization

All within your existing GA4 → BigQuery pipeline.

Use Case: Analyzing Multilingual Feedback from GA4

Let’s say you’ve implemented a GA4 custom event called nps_feedback, which captures open-text user feedback via an event parameter named feedback.

The challenges:

Feedback is unstructured and difficult to analyze
Responses come in multiple languages
Data cannot be easily used for reporting or segmentation

Step 1: Export GA4 to BigQuery

Exporting GA4 data to BigQuery provides access to raw, event-level data, including feedback across all users and sessions. (Prerequisite)

Step 2: Structure the Feedback Data

Before applying AI, feedback data should be organized into a structured table where each row represents a feedback event.

This enables:

Consistent querying
Scalable enrichment
Reliable downstream analysis

Step 3: Use AI.GENERATE() to Enrich the Feedback

This is where generative AI transforms GA4 data.

Using Gemini models inside BigQuery, you can convert unstructured feedback into structured fields that can be used across analytics and reporting.

You are turning text into measurable data.

WITH ai_function_feedback AS (

SELECT

feedback AS original_feedback,

AI.GENERATE(

CONCAT(

'Translate this feedback to English if not already in English. ',

'Identify if it is about customer service and provide a positivity score. Text: ', feedback

),

connection_id => 'us-central1.vertex_ai_conn',

endpoint => 'gemini-2.0-flash',

model_params => JSON '{"temperature": 0}',

output_schema => 'englishreview STRING, is_about_service BOOL, positivity_score FLOAT64'

) AS Englishreview

FROM `your_project.your_dataset.feedback_table`

)

SELECT

original_feedback,

Englishreview.englishreview,

Englishreview.is_about_service,

Englishreview.positivity_score

FROM ai_function_feedback;

Step 4: Output Example

You now have structured outputs:

Translated feedback
Topic classification (e.g., customer service)
Sentiment score (positivity_score)

Now you can use these columns to:

Identify negative user experiences at scale
Segment users based on sentiment
Track customer experience trends over time
Feed enriched data into dashboards and models

How Generative AI Improves GA4 Analysis

GA4 tells you what users are doing (page views, clicks, sessions) but without context, behavior is just numbers. By integrating Generative AI, you can transform unstructured inputs into meaningful, structured insights.

Challenge	How Generative AI Helps
Unstructured, multilingual feedback	Translate and normalize directly within SQL
Subjective sentiment	Generate a standardized `positivity_score` field
Hard to categorize feedback	Automatically classify using an `is_about_service` boolean
Manual analysis is slow and labor-intensive	Automate at scale

From Data to Insight

Generative AI bridges the gap between behavioral data and customer context, enabling a shift from:

Reporting → Insight generation
Data collection → Decision-making

This is a key step toward AI-driven measurement strategies.

Use Case: AI-Summarized Page Performance in GA4

Another high-impact use case is summarizing page-level performance. In this use case, we utilized AI.GENERATE() to automatically summarize user engagement for each GA4 page_path. By combining quantitative metrics like pageviews and average session duration, the model generates concise, human-readable descriptions of page performance. Additionally, it assigns a performance label (e.g., High, Low) to help quickly identify content that requires attention or optimization

Using AI.GENERATE(), you can combine metrics such as:

Pageviews
Engagement time
Session behavior

And generate:

Human-readable summaries
Performance labels (High, Medium, Low)

This approach enables:

Faster content performance audits
Scalable page-level insights
Smarter dashboard annotations - all with SQL and no external processing.

All of this is generated directly within BigQuery using SQL. No external tools. No additional pipelines.

Data Quality, Validation, and Workflow Integration

AI outputs are only as reliable as the underlying GA4 data. To generate meaningful and trustworthy insights, organizations need a strong data foundation built on clean event tracking, consistent naming conventions, and well-structured, validated BigQuery pipelines.

At the same time, generative AI should enhance analysis, not replace it. Outputs still require human oversight to ensure they are accurate, aligned with business logic, and consistent with internal definitions and governance standards.

The real value comes when AI is embedded into existing workflows. This includes integrating AI-driven insights into reporting environments, aligning them with marketing decision-making processes, and incorporating them directly into data pipelines so insights can be generated and acted on at scale.

Conclusion

Generative AI in BigQuery transforms GA4 from raw data into rich, human-readable insights: all with SQL. Whether you're analyzing multilingual feedback or summarizing page performance, AI.GENERATE() removes the need for external tools, speeds up analysis, and unlocks smarter decision-making directly within your data pipeline.

More Insights

Why Poor GA4 Data Is Hurting Your Google Ads Performance

Skylar van Dalen-Flude

Senior Data Analyst

May 20, 2026

How to Track CTA Clicks in Google Tag Manager

Aiswarya Nair

Senior Implementation Specialist

May 6, 2026

Four ways Google Cloud helps build data strength in Google Marketing Platform

Colin Temple

SVP, Data Solutions

Apr 21, 2026

More Insights

Sign Up For Our Newsletter

Napkyn Inc.
204-78 George Street, Ottawa, Ontario, K1N 5W1, Canada

Napkyn US
6 East 32nd Street, 9th Floor, New York, NY 10016, USA

212-247-0800 | info@napkyn.com

About

Services

Our Work

Insights

Webinars

Careers

Kepler

Kyu

Cookie Policy