Claude + GA4: How to Ask Your Analytics Questions in Plain English

By Emily Redmond, Data Analyst at Emilytics · April 2026

TL;DR: Claude can connect to your GA4 data via Model Context Protocol (MCP) and answer analytics questions in plain English. You ask "What drove the traffic spike this week?" and Claude fetches the data, analyzes it, and explains what happened—all in 30 seconds.

The Setup (It's Easier Than You Think)

Here's the truth: connecting Claude to GA4 used to require engineering skills. You'd need API credentials, custom code, error handling—the whole deal.

Now it takes five minutes.

Claude has a built-in MCP integration that speaks to GA4 directly. You authenticate once with Google OAuth, and Claude becomes your analytics agent.

Here's the real workflow:

Open Claude (web or Cursor)
Tell Claude: "Connect to my GA4 account"
Claude walks you through Google OAuth
You authorize GA4 read access
Start asking questions

That's it. You're done.

💡 Emily's take: The first time I asked Claude a GA4 question and got an instant answer, I was stunned at how frictionless it was. No API docs. No debugging. Just conversational analytics. This is what I'd been waiting for.

What Questions Can You Actually Ask?

Anything GA4 can answer, Claude can answer faster.

Here are real examples I use every week:

Traffic & Audience:

"What were my top 5 landing pages last week by session count?"
"How many new users did I get this month vs. last month?"
"Which countries drive the most revenue?"
"Are my bounces increasing or decreasing?"

Conversion & Goals:

"What's my conversion rate by traffic source?"
"Which landing pages convert best?"
"Show me the user journey for people who converted."

Content Performance:

"Which blog posts are driving the most traffic?"
"What's the most popular content in my niche?"
"Which pages are getting organic traffic but not converting?"

Anomalies & Changes:

"Did anything unusual happen to my traffic last Tuesday?"
"Why did mobile traffic spike Thursday?"
"Which metrics have changed most since last month?"

Comparisons:

"How does this week compare to last week?"
"Is this month better or worse than March?"
"Which campaigns are driving ROI?"

Claude doesn't just fetch numbers. It interprets them. You get context, comparison, and explanation.

Example response:

Your organic traffic increased 34% week-over-week. This is driven by two factors: (1) Your homepage ranking improved 5 positions for 'AI analytics automation' (+22 sessions), and (2) A new article about natural language analytics is already ranking page 2 for its target keyword (+18 sessions). Overall, your organic strategy is working well. The main opportunity is the article on page 2—small optimization could push it to page 1 and double its traffic.

That's not just data. That's actionable insight.

How Claude Accesses GA4

When you ask Claude a question, here's what happens behind the scenes:

Step 1: Understanding Your Question

Claude reads what you asked and maps it to GA4 concepts. "Top landing pages" → landing_page dimension, sorted by session count.

Step 2: Fetching via MCP

Claude doesn't query GA4 directly. It uses MCP tools exposed by the GA4 MCP server. These tools are pre-built and tested.

Available tools include:

get_traffic_overview – Total sessions, users, pageviews
get_top_pages – Top pages by traffic
get_top_search_queries – Top organic keywords
get_traffic_sources – Traffic breakdown (organic, direct, social, referral)
compare_traffic_periods – Compare week vs. week, month vs. month
detect_anomalies – Flag unusual changes

Step 3: Interpreting Results

Claude gets back raw numbers, then explains what they mean. It compares to baselines, calculates growth rates, and flags what's unusual.

Step 4: Explaining to You

Claude writes a conversational response with context, comparison, and actionable next steps.

Total time: 15–30 seconds. No authentication needed after setup.

💡 Emily's take: I've used GA4's web UI for eight years. The difference between clicking through filters and just asking Claude is like the difference between a calculator and a spreadsheet. One is a tool; the other is a conversation. I'll never go back to dashboards.

Real Workflows: Before & After

Before (Dashboard Era)

You: Wonder if your campaign drove any ROI Action: Log into GA4 → Find the campaign source → Create a custom date range → Wait for the chart to load → Compare to last campaign manually → Find the conversion metric → Cross-reference with cost → Do math in your head Time: 8 minutes

After (Claude Era)

You: "Did the Q2 campaign drive ROI?" Claude: Fetches campaign data, compares to baseline, calculates ROI automatically Time: 30 seconds

The difference is speed. But it's not just speed—it's repeatability. You can ask Claude 20 questions in the time it used to take to answer 1.

What Makes This Work So Well?

1. Conversational Context

Claude remembers context across messages. You can say "Compare this week to last week" and Claude knows you meant "this week" in the current conversation. Natural conversation flow.

2. Explanation Built In

Claude doesn't just return data. It interprets and explains. "Traffic spiked 45% because..." is way more useful than "Traffic: 1,234 (↑ 45%)".

3. Follow-Up Questions

You can drill deeper without starting over. "That looks good. Now show me which pages contributed most" → Claude already has context and gives you the answer immediately.

4. No Authentication Overhead

You authenticate once, at setup. After that, every question just works. No token refreshes, no error handling.

The Limitations (Be Honest)

Claude + GA4 isn't perfect:

GA4 limitations are Claude limitations. If your GA4 setup is broken, Claude can't fix it. Garbage in, garbage out.
You need smart questions. "Tell me about my data" is too vague. "What drove the spike in organic traffic?" is specific and answerable.
Some advanced analysis requires custom code. Claude can handle most queries, but complex custom dimensions or very specific calculations might need SQL or a custom MCP server.
Rate limits apply. Google has API quotas. If you ask too many questions in a short time, you might hit them. It's rare, but possible.

These are all manageable. They're constraints, not blockers.

Setting Up Claude + GA4

(Detailed setup in MCP for Analytics: How to Connect Any AI Agent to GA4 and GSC)

Quick walkthrough:

Open Claude (claude.ai or Cursor)
Go to Settings → MCP Servers
Add GA4 MCP server (Emilytics or Anthropic's official server)
Authenticate with Google (OAuth)
Test with a simple question: "How many sessions did I get last week?"
Start asking real questions

Total setup time: 5 minutes.

Why This Changes Everything

Before Claude + GA4, accessing your analytics required:

Opening a website
Clicking through menus
Waiting for dashboards to load
Exporting data
Doing manual analysis in spreadsheets

Now it requires:

One sentence

That shift—from procedural to conversational—is profound. It's the difference between a tool you use and a tool that works with you.

Next Steps

If you want to try this right now, set up Claude + GA4.

If you want to understand the technical foundation (Model Context Protocol), read about MCP.

If you want to see how this fits into the broader shift away from dashboards, read about how AI agents are replacing dashboards.

The Bottom Line

Claude + GA4 proves that the future of analytics is conversational. You don't need to learn a tool. The tool learns your language.

Set it up, ask a question, and watch what happens. It's that good.

Emily Redmond is a data analyst at Emilytics — the AI analytics agent watching your GA4, Search Console, and Bing data around the clock. 8 years experience. Say hi →