Exporting charts from Looker Studio to Google Slides with AI-generated insights is now a reality thanks to Gemini.
In this post, I’ll explain this functionality, which streamlines the report creation process, allowing you to transform your data into presentations with just a few clicks.
Although what truly sets it apart is its ability to include AI-generated insights within your slides. Instead of just displaying numbers and charts, Gemini analyses the data for you and adds comments to your slides, providing an analysis of the trends and patterns shown in your charts.
About the data we will be working with
Before diving into how to generate slides with Gemeni in Google Slides, let’s review the report we’ll be working with.
The report contains data from Google’s e-commerce, where it sells the company merchandise. This data comes from Google Analytics 4. You can see it on this link.

The report consists of two pages. On the first page, we have four scorecards displaying some of the site’s most important metrics. These metrics are:
- Total Users, that represents the total number of users who visited the site.
- New Users, which refers to users who visited the site for the first time during the analysed period.
- Engagement Rate, which is a specific metric from Google Analytics 4, and it shows the percentage of sessions where users interacted with the site. Interactions are counted for visits lasting at least 10 seconds or those that included a key event interaction.
- Conversion Rate, that shows the percentage of users who completed a purchase.
If we scroll down a bit, we can see, on one side, a bubble chart about marketing channels, and, on the other side, a table on landing pages.

The marketing channels chart shows the relationship between the number of users and the conversion rate for each channel. The landing pages table lists the pages that users entered the site through. It displays the session metrics to identify the main access points and the engagement rate to help assess the quality of the traffic.
On the second page of the report, we see the four main metrics from scorecards again, but this time, we’re looking at their trends. Each metric has its own time series chart, showing data from January to September 2024.

How to export your Looker Studio report to Google Slides with Gemini
Let’s dive into how to generate the slides.
You’ll notice that when you’re in edit mode on the report, on the right-hand side, there’s a column where you can see Data, Properties, Filter, and Gemini.

This last option (Gemini) is the one we need to click on to see the option to generate slides.

If you don’t see Gemini, it might be because you don’t have it enabled. You need Looker Studio Pro to activate Gemini.
Once you click on Gemini, you’ll see the option to generate slides from all the visualisations in the report, or to choose specific ones.

In my case, I’m going to select all, and it will begin the process of generating the presentation, and when it’s done, it provides a link where I can access the slides.

If I click on it, I can see that it has created a presentation with one slide per visualisation, except for the scorecards.

The first slide is about marketing channels and has four bullet points in the comments.

Let’s remember that in the graph, we are analysing users and the conversion rate by marketing channel.
The first comment tells us that the direct traffic channel has generated the majority of users, at 52.54%, and that Display has brought in an insignificant percentage of users, less than 0.01%.
So far, this is great because it highlights which channels are performing best and worst for one of the analysed metrics, which is total users.
The next point addresses the conversion rate, indicating that the Referral channel has the highest rate at 3.35%. This comment is also very useful, as it shows us which channel performs best for the other analysed metric, the conversion rate.
Finally, it mentions that the average number of users per channel is 8,339.
I personally believe the last point lacks much value, as the average number of users per channel doesn’t really tell us anything useful. Understanding how marketing channels work, we wouldn’t expect all of them to bring in a similar amount of traffic, so knowing the average doesn’t serve as a benchmark. For instance, if the average is 8,339 and Display is generating 100, I can’t say that Display is not performing well just because it’s far from the average, as that wouldn’t be a target for all channels. Each channel’s target depends on the actions the eCommerce takes and the behaviour of the channel itself.
After reviewing these comments, I find them great as a starting point because they indicate where I need to dig deeper. For example, in the case of the Referral channel, the next step would be to look at which sources are contributing to that strong conversion rate.
Regarding the last point, while the comment itself isn’t bad, understanding what to highlight often requires knowledge of the industry or business, which, in this specific case, the AI model doesn’t seem to possess.
This is where the analyst’s role comes in, as they need to ensure the analysis is accurate. However, these comments serve as nearly perfect starting points.
Let’s move onto the second slide. Here, we’re analysing the landing pages, meaning the first page users see during their session, along with the engagement rate for each.

The first point states that the most successful landing page is the homepage, which accounts for 47% of the sessions. It also adds that this indicates a very high concentration of traffic, which is true; as we can see, nearly half of the sessions entered the site through the same page, the homepage.
The second point mentions that there are 157 landing pages with minimal engagement, representing only 0.1% of the total sessions. Looking at the graph, we can see that there are a total of 680 landing pages, and these 157 have generated just 0.1% of all sessions. This highlights that a few landing pages are receiving the majority of the traffic.
Next, it states that the average number of sessions per landing page is 220, indicating a very broad distribution of traffic across different pages. This means that while a few pages receive the most sessions, there are still many landing pages in total—680 in fact, each with at least one visit.
The comment continues by noting that 384 landing pages have achieved maximum engagement, which is 100%, while 56 landing pages show no engagement, at 0%. It also suggests that we should investigate and optimise those pages performing poorly.
In this instance, we see that Gemini not only comments on the data but also provides recommendations on what steps to take—specifically, to investigate and optimise the underperforming pages.
Moving on to the next slide, we begin to see the time series charts.

The first one shows total users, stating that total users have experienced a growth of 20.6% from January to September 2024. It also mentions that despite the overall growth throughout the year, there was a significant drop in users of nearly 47% from August to September. This is accurate and reflected in the graph; however, it doesn’t clarify that the data for August includes the entire month (31 days), while September only accounts for half the month (15 days).
This example further illustrates that the comments generated by Gemini are a great starting point but need to be validated by the analyst, who ideally has a deeper understanding of the context.
I’ll jump directly to the last slide since the two previous ones are similar to the one we just reviewed, and here I want to focus on the first point.

It mentions that the conversion rate was consistently low over the nine months analysed. Looking at the graph, we can see that, except for January and April, it stayed between 1% and 1.40%. What really caught my attention is that Gemini has determined that a conversion rate between 1% and 1.40% is considered low. It’s not just providing the numbers; it’s also interpreting their significance. In the last point, it even states that these data indicate potential issues in the customer journey leading to purchase.
In the second point, it mentions that in month 4, the conversion was almost zero. An analyst, upon seeing this, would likely think that there was an implementation issue, but Gemini does not mention this possibility. Once again, we see how crucial the analyst’s role is in validating these comments.
As we’ve seen, using Artificial Intelligence is a very good starting point for analysis. Sometimes, it can even show you insights that you might not have stopped to analyze yourself. However, it’s important that these insights are validated by an analyst.
If you’d like to see the presentation yourself and read the comments, you can do so by following this link.
Leave me a comment if you have any questions! 🙂