Top 10 Business Prompts for Google Analytics

The best AI prompts for Google Analytics give the AI a specific GA4 report, name the metrics, date range and comparison in GA4’s own terms, ask one business question and say what format you want back. You then check the answer against GA4 before you act on it.

Below are 10 copy-ready prompts for the questions founders ask most: why traffic dropped, which channels and campaigns bring customers, which landing pages convert, where the signup funnel leaks, how devices and returning users compare, how much traffic comes from AI assistants, what changed this week, and a monthly summary. Each one says which data to paste and how to check the result. They work in ChatGPT, Claude or Gemini with an exported report, and most of them also work in Ask Advisor, the Gemini assistant built into Google Analytics.

Key Takeaways

• Ask Advisor is a Gemini-powered assistant inside Google Analytics, in beta for eligible properties set to English, and it answers questions about your data only from that property (Google Analytics Help, October 2026).
• A GA4 report exports to CSV or Google Sheets with up to 100,000 rows, which is how you give ChatGPT or Claude the data (Google Analytics Help).
• Google’s Analytics MCP server lets AI assistants such as Gemini and Claude query GA4 directly, for read requests only (Google for Developers).
• Google renamed GA4 conversions to key events on March 27, 2024, so prompts that say “conversions” can mix up GA4 key events with Google Ads conversions (Google Analytics release notes).
• An engaged session lasts longer than 10 seconds, has a key event or has at least 2 page or screen views (Google Analytics Help).
• Since May 13, 2026, GA4’s default channel group has an AI Assistant channel for visits from ChatGPT, Gemini, Claude and similar chatbots (Google Analytics release notes).

How to Give an AI Your Google Analytics Data

An AI can only analyze the numbers it can see. A general chatbot has no access to your Google Analytics account unless you connect it, and without data it will answer from general knowledge or make numbers up. There are three ways to give it the data:

Route	How it works	Best for	Limits
Ask Advisor in Google Analytics	Click the Ask Advisor icon in your property, or the icon on a chart card, and type a question.	Quick questions about your own property, and “how do I” questions about GA4.	Beta, eligible properties set to English only. Google notes it can make mistakes.
Export a report and upload it	In a report, click Share this report, then Download File: PDF, CSV or Google Sheets.	ChatGPT, Claude or Gemini, with your own prompt and full control over the data you share.	Up to 100,000 rows per CSV or Sheets export. The data is a snapshot, so re-export for a new period.
Google Analytics MCP server	Google’s open-source server connects GA4’s Admin and Data APIs to an AI assistant such as Gemini, Claude or Cursor.	Repeat analysis and custom agents that query GA4 directly.	Read requests only. Setup needs Python, a Google Cloud project and API access.

Third-party connectors that pipe GA4 data into ChatGPT also exist. Google says its terms broadly prohibit collecting raw personal data in Google Analytics, and exported reports are aggregated counts. Still, page URLs can carry emails or names in query strings, so scan a Landing page export first, and check your AI tool’s data settings before you upload company data. For Ask Advisor, Google states that your prompts and answers are not used to train its models and that chat logs are deleted after 55 days (How Ask Advisor uses data).

What a Good Google Analytics Prompt Includes

Every prompt below follows the same five parts:

• The data. Name the GA4 report and attach it, or tell Ask Advisor which report to use.
• GA4’s own terms. Say sessions, engaged sessions, engagement rate, key events and session default channel group. A vague word such as “conversions” or “traffic quality” lets the AI pick its own definition.
• A date range and a comparison. “Last 28 days compared with the previous 28 days” gives the AI a baseline. A number with nothing to compare it to tells you little.
• One question. Ask about one decision at a time, then follow up.
• The output. Ask for a table, a ranked list or a short summary, and ask it to show the numbers and filters it used.

Replace the [brackets] with your own details. If your key event is a signup, a demo request or a purchase, name it, because GA4 counts each key event separately.

Top 10 Business Prompts for Google Analytics

1. Why Did Traffic Drop?

Use this when sessions fall and you need to know where the drop came from before you react.

Paste: the Traffic acquisition report by session default channel group, by day, for the last 28 days and the previous 28 days, plus a list of anything you changed on the site or in campaigns in that time.

Here is our GA4 Traffic acquisition data by session default channel group and day for the last 28 days and the previous 28 days [attach], and a list of the site and campaign changes we made in that time [paste]. Tell me which channels account for most of the change in sessions, on which dates the drop started, and whether engaged sessions fell by the same share. Then list the changes on our list that happened within two days of the drop. Show a table of channel, sessions in each period, change and share of the total change.

Keep that change list in GA4 itself: annotations mark a date on any line graph, and Google may add its own when a significant event affects your data. John Doherty, who ran Getcredo with three W-2 employees when he gave this talk, annotates every change for the same reason:

When I make any change to my conversion forms especially, I’m annotating it in Google Analytics, so if something drops and I don’t have it annotated, I’m in trouble. I don’t know what I changed … so I annotate every single thing so I can roll stuff back as quickly as possible.

John Doherty, Founder, Getcredo ($500K revenue). Watch at 19:50

Check the output: the channel changes should add up to the total change in sessions. A drop on one date across every channel often points to tracking, such as a broken tag or a consent change, rather than to demand.

2. Which Channels Bring Engaged Visitors and Key Events?

Use this to see which channels send people who do something, not only the most visits.

Paste: the Traffic acquisition report by session default channel group for the last 90 days, with sessions, engaged sessions, engagement rate, key events for [your key event] and session key event rate.

Here is our GA4 Traffic acquisition report by session default channel group for the last 90 days [attach]. Rank the channels by key events for [signup], then show each channel's sessions, engagement rate and session key event rate. Point out channels with many sessions but a low key event rate, and channels with few sessions but a high one. Ignore channels with fewer than [100] sessions. Finish with three questions I should investigate, not conclusions.

Check the output: recompute one channel’s engagement rate (engaged sessions divided by sessions) yourself. Don’t divide key events by sessions to check the key event rate: key events counts every signup, while session key event rate counts sessions with at least one. If the Unassigned channel is large, fix the UTM tagging before you trust the comparison.

3. Which Campaigns Bring the Customers We Want?

Use this before you move budget between campaigns.

Paste: the Traffic acquisition report with session campaign as the dimension (filter to paid channels), with sessions, key events and total revenue if you track purchases, plus your spend per campaign from the ad platforms.

Here is our GA4 data by session campaign for the last [60] days [attach] and our spend per campaign [paste]. For each campaign, calculate cost per key event for [demo request] and, where revenue is tracked, revenue per session. Rank the campaigns, and flag any campaign whose cost per key event is more than twice the average. Tell me what extra data would show which campaigns bring our largest customers.

The cheapest lead isn’t always the best one. Jeremy Goillot, founder and CEO of the phone system Allo, tracks which campaigns bring his largest accounts rather than only the lowest cost:

For every ads, both in my Google Analytics … I can see each campaign how many percentage of whales that bring me back. You know? And so today, I know my whales is above like two k or three k a year on my product in term of ACV. So I don’t really care about the CAC for those people because I know they will make it.

Jeremy Goillot, Founder & CEO, Allo. Watch at 10:46

To do the same in GA4, send the account size or plan as an event parameter when someone signs up, and add it to this prompt.

Check the output: cost per key event is a lead cost, not a customer cost. Compare it with what a customer is worth using a CAC calculator before you cut a campaign that looks expensive.

4. Which Landing Pages Convert, and Which Leak?

Use this to choose which pages to fix first.

Paste: the Landing page report for the last 90 days with sessions, engagement rate, key events and session key event rate (add the last two under Customize report if they are missing), ideally split by session default channel group.

Here is our GA4 Landing page report for the last 90 days [attach]. List the 10 landing pages with the most sessions, and for each show engagement rate and session key event rate for [signup] compared with the site average. Flag pages with above-average sessions and below-average key event rate as fix-first candidates. Only include pages with at least [200] sessions. For the top three, suggest what to test, based on what the page is for: [describe each page].

This is the third of three checks for LinkedIn ad campaigns on a slide by Brian Yam, head of marketing at Paragon: click-through rate tells you whether the audience and message are right, cost per lead or conversion compares campaigns with each other, and website activity tells you whether the offer and the page are good enough. See slide 41.

Check the output: the AI only sees numbers, not your pages. Open each flagged page yourself before you accept a suggested fix, and compare pages within the same channel, because a page that gets mostly paid traffic converts differently from a blog post.

5. Where Does the Signup Funnel Lose People?

Use this when traffic is fine but signups or purchases are not.

Paste: a Funnel exploration with your steps (for example landing page view, pricing page view, signup started, signup completed), with device category as the breakdown, exported with the completion and abandonment rate per step, for two periods.

Here is our GA4 funnel exploration for [landing page, pricing page, signup started, signup completed] for the last 30 days and the previous 30 days [attach]. Tell me which step loses the most users, whether that step got better or worse between the periods, and how the drop differs between mobile and desktop. Give me three hypotheses for the biggest drop, each with the GA4 data that would confirm or rule it out.

Check the output: the funnel only shows steps you track as events. If a step is missing, such as trial to paid, the AI can’t see it; ask it to say which steps it could not measure.

6. How Do Mobile and Desktop Compare?

Use this to decide whether the mobile experience needs work.

Paste: the Traffic acquisition report with device category added as a dimension, and the Landing page report with device category as a secondary dimension, both with sessions, engagement rate, key events and session key event rate for the last 90 days.

Here is our GA4 data by device category, overall and by landing page, for the last 90 days [attach]. Compare mobile, desktop and tablet on sessions, engagement rate and session key event rate for [signup]. If mobile converts at less than half the desktop rate, tell me which landing pages have the largest gap. Present it as a table and a three-sentence summary.

Check the output: a lower mobile rate is normal for B2B products that people research on a phone and buy at a desk. Compare the gap with your own history, not with a general benchmark.

7. Do New Users Come Back?

Use this to see whether the people you acquire keep returning, by acquisition channel.

Paste: a Cohort exploration by first-visit week for the last 8 weeks, with first user default channel group as a breakdown, exported.

Here is our GA4 cohort exploration by first-visit week for the last 8 weeks, split by first user default channel group [attach]. Tell me which channel's users return most in weeks 1 to 4, whether recent cohorts return more or less than older ones, and which channel brings users who never return. Keep the answer to a table and five bullets.

Check the output: website return visits are not product retention. If you sell software, measure customer retention from your billing data, for example with a SaaS cohort analysis, and use this prompt only for the marketing site.

8. How Much Traffic Comes From AI Assistants?

Use this to track visits from ChatGPT, Gemini, Copilot and other chatbots.

Paste: the Traffic acquisition report filtered to the AI Assistant channel, by session source with landing page as a secondary dimension, plus the channel’s sessions by week, for the last 90 days, alongside the Organic Search totals.

Here is our GA4 AI Assistant channel data by session source and landing page for the last 90 days, and our Organic Search totals for the same period [attach]. Show sessions, engagement rate and key events by AI source, the landing pages AI assistants send people to most, and the trend by week. Compare the AI Assistant channel's engagement rate and key event rate with Organic Search.

Check the output: Google added this channel on May 13, 2026, so compare periods after that date: earlier visits from AI assistants may sit in other channels such as Referral, so don’t read the start of the series as growth. Visits from Google’s AI Overviews and AI Mode count as Organic Search, not AI Assistant.

9. What Changed This Week That I Should Know About?

Use this as a weekly check, so problems surface before the monthly review.

Paste: the Traffic acquisition and Landing page reports for the last 7 days and the previous 7 days, plus key events by name.

Here are our GA4 Traffic acquisition and Landing page reports and our key events for the last 7 days and the previous 7 days [attach]. List anything that changed by more than [20%] in sessions, engaged sessions or key events, by channel, landing page and key event. Flag any key event that fell to zero or close to it, because that usually means tracking broke. Rank the changes by size, and say for each whether it looks like a tracking issue or a real change.

Check the output: week-over-week numbers swing on small sites, so set the threshold for your traffic. GA4’s AI overviews on the Home page also summarize the biggest changes since your last visit; compare the two lists.

10. Write the Monthly Marketing Summary

Use this to turn the month’s reports into a short update for your team, investors or board.

Paste: the month’s Traffic acquisition, Landing page and key event data, with the previous month and the same month last year if you have it, and your targets.

Act as a head of marketing writing for a founder. Using this GA4 data for [month] with the previous month [attach] and our targets [paste], write a one-page summary: five headline numbers with the change from last month (sessions, engaged sessions, key events for [signup], session key event rate, top channel by key events), what worked, what didn't and the three decisions we need to make. Use only the numbers in the data, and no adjective without a number behind it.

Check the output: read the summary against the reports, not on its own: AI writing makes weak numbers sound confident. Pair it with the revenue figures in our list of SaaS KPIs so traffic is not the only measure of the month.

Follow-up prompt: after any of these, ask “Which of these findings would change what we do next week, and what would we measure to know it worked?” It turns a report into a decision.

For campaign copy and channel plans, see our AI prompts for marketing; for forecasts and board reports, see our AI prompts for finance teams.

How to Check What the AI Tells You

Even Google says Ask Advisor “may have some capability limitations and make mistakes”, and a general chatbot reading an export can misread columns or invent a number that isn’t in the file. Four checks catch most problems:

• Match one number to GA4. Open the same report for the same dates and confirm at least one figure the AI quotes. If it doesn’t match, check the date range and filters first.
• Check the metric. Users, sessions and engaged sessions are different counts, and key events are not Google Ads conversions. Ask the AI which metric each number is.
• Check the sample size. A page with 30 sessions can show a 10% key event rate from three signups. Set a minimum number of sessions in every prompt.
• Look for a tracking cause. A sudden change on one date, across every channel, is more often a tag, consent or filter change than a change in demand.

Then decide what to do yourself. The AI is good at sorting and summarizing a report quickly; it doesn’t know your product, your pricing or what changed in your market last week.

Put the Numbers to Work

The same traffic, key event and campaign data these prompts analyze is part of how a funder reads your growth. Founderpath’s Google Analytics integration is a read-only connection that, as its page states as of October 2026, adds traffic and conversion context to your funding profile alongside your revenue data.

Founderpath provides non-dilutive capital for bootstrapped software founders. Connect your billing, banking and accounting data, and Founderpath shows you what you qualify for, with every figure disclosed up front. There’s no pitch deck, no equity and no board seats.

Check what you qualify for. It takes less than 5 minutes.

Google Analytics AI Prompts FAQ

What are the best AI prompts for Google Analytics?

The best ones attach a specific GA4 report, name the metrics in GA4’s terms (sessions, engaged sessions, key events, session default channel group), give a date range with a comparison, ask one question and set the output format. The most useful jobs are explaining a traffic drop, comparing channels and campaigns, finding landing pages that leak, checking the signup funnel and writing a monthly summary.

Can ChatGPT connect directly to Google Analytics?

Not on its own. You either export a GA4 report as CSV or Google Sheets and upload it, use a third-party connector, or connect an AI assistant through Google’s Analytics MCP server, which handles read requests only. Inside Google Analytics, Ask Advisor answers questions about your property directly.

Does Google Analytics have its own AI assistant?

Yes. Ask Advisor is a conversational assistant in Google Analytics powered by Gemini. As of October 2026 it is in beta, available in eligible properties set to English, and it answers questions about your data only from that property. Google also shows AI overviews on the Home page that summarize the biggest changes since your last visit.

Is it safe to upload Google Analytics data to ChatGPT?

GA4 reports are aggregated counts, and Google’s terms broadly prohibit collecting raw personal data in Google Analytics, so a standard report export holds aggregated counts, not individual users’ details, unless your page URLs carry personal data in query strings. Still, use the AI tool your company has approved for business data and check whether it uses uploads for training. For Ask Advisor, Google states prompts and answers are not used for model training and chat logs are deleted after 55 days.

How much GA4 data can I export for an AI to analyze?

A CSV or Google Sheets export of a GA4 report includes up to 100,000 rows. For more than that, or for repeat analysis, use the Google Analytics Data API or the Analytics MCP server, which reads the same data through the API.

What are key events in GA4?

Key events are events you mark as important to your business, such as a signup or a purchase. Google renamed GA4 conversions to key events on March 27, 2024; a conversion now means an action used to measure ad campaigns, as in Google Ads. Use “key events” in your prompts so the AI reads the right GA4 metric.

How do I track ChatGPT and other AI assistant traffic in GA4?

Since May 13, 2026, GA4 assigns visits from recognized AI assistants such as ChatGPT, Gemini and Copilot to an AI Assistant channel in the default channel group, with the medium ai-assistant. Visits from Google’s AI Overviews and AI Mode are counted under Organic Search instead.

Can AI replace Google Analytics?

No. Google Analytics collects and stores the data; an AI assistant reads it and helps you sort, compare and summarize it. The AI’s answer is only as good as the report you give it, so check its numbers against GA4 before you act.