NotebookLM is a generative AI tool by Google that allows users to input and engage with multiple sources at once. In my previous post, I outlined some updates to NotebookLM in the last year. These updates included an updated interface, new interactive features, and more suggested content types in the NotebookLM Studio.
One of the affordances that NotebookLM offers users is the ability to easily generate various types of content based on the input without needing to engineer their own prompts. This type of interaction really lowers the bar for users who might not be confident or familiar with writing AI prompts.
Let’s explore!
In this post, I want to explore a couple of the content types in a bit more depth. First, we’ll take a look at the options available for creating reports based on sources. Then, we’ll explore audio overviews. With each type of content, we’ll look at the different ways you can customize output as well as discuss a couple of ways these types of content might be useful for language teachers and learners.
For the sample content I’ll work with in this post, I wanted to go back to my roots. I grew up in Minnesota, USA, and I spent the first chunk of my career teaching international students at the University of Minnesota. During the summer term, we spent a lot of time exploring themes related to tourism in Minnesota and Minnesota history. For this post, the input I’ll use in NotebookLM will be a Youtube video about tourist attractions in Minnesota.
Without further ado, let’s explore making reports in NotebookLM.
Reports
In my first blog post about NotebookLM from September 2024, I wrote a bit about the types of content you could generate in the Notebook Guide feature of the application. Basically, NotebookLM would generate suggested content types based on your sources.
This functionality is the same after the updates over the last year, but each content type is a bit more fleshed out and robust. With regard to reports, in the previous version, you could generate summaries, study guides, FAQs, timelines, and a table of contents at the click of a button. For any content more substantial, you would need to draft your own prompts in the chat section of the application.
With the current version of NotebookLM, there are multiple options for generating various types of reports based on your sources. For instance, you can easily create a briefing document, study guide, blog post, as well as other suggested content derived from the topic of your sources. You can also create your own type of report by specifying the structure, tone, and style.
With my content exploring Minnesota tourist attractions, the reports feature suggests I create a destination marketing brief, content strategy memo, location profile, and a thematic guide.
To start, let’s explore how you can create your own report by customizing style, structure, and tone specifications.
Create your own report
The first thing to note when creating your own report, is that you’re able to select the language of the report despite the language your sources might be in. For language learners, this could be an interesting feature to explore language features and understanding of a topic. We’ll explore this a bit more in a minute.
But just looking at creating your own report, after selecting the language, you need to write a prompt to generate the report. I used the prompt: “Create a summary paragraph about the source that includes language appropriate for A1 English learners.” This is a prompt that instructors might use to specify the difficulty of the output.
The summary paragraph generated looks like:
“Minnesota is a place in the United States. It has many lakes and is called the “Land of 10,000 Lakes”. You can do fun things outside in nature and also find interesting places in cities. For example, you can visit big cities like Minneapolis, see big parks, and explore the very big Lake Superior.”
At this point, you’re able to give feedback about the generated report by clicking the “Good report” or “Bad report” buttons.
You can see that the summary paragraph generated is simple, includes details from the source video, and would be more or less appropriate for an A1 English learner. When using the same prompt to generate content for B1 English learners, the following paragraph was generated:
“Minnesota is a state in the Midwestern United States known for its beautiful nature and lively cities, making it an interesting place for travelers. People often call it the “Land of 10,000 Lakes,” but it really has many more lakes than that. In Minnesota, you can enjoy activities outside in nature and also explore exciting cities. For example, you can visit Gooseberry Falls State Park to see powerful waterfalls. You can also go to big cities like Minneapolis and St. Paul, known as the Twin Cities, where you can watch sports games or listen to live music. This mix of quiet nature and busy city life makes Minnesota a great place for a trip.”
And C1:
“Known as the “Land of 10,000 Lakes,” Minnesota offers a compelling blend of vast natural adventures and vibrant urban experiences. The state’s diverse landscapes are anchored by the majestic Lake Superior, whose North Shore invites exploration along extensive hiking trails and rewards visitors with sights like the powerful waterfalls of Gooseberry Falls State Park, the picturesque Split Rock Lighthouse—a National Historic Landmark—and the blossoming artistic charm of the small village of Grand Marais. Minnesota’s natural significance is further highlighted by Itasca State Park, which contains the headwaters of the mighty Mississippi River, and the pristine, water-rich environment of Voyageurs National Park, a perfect destination for escaping crowds and viewing local wildlife. In contrast to this wilderness, the metropolitan “Twin Cities” provide a dynamic cultural heart. Minneapolis, the state’s largest city, is a hub for professional sports and a celebrated music scene, famously centered around the First Avenue venue. Adjacent to it, the capital city of St. Paul showcases historic, castle-like architecture such as the Landmark Center and offers popular family-friendly attractions like the Como Park Zoo & Conservatory. Ultimately, Minnesota presents a rich and varied travel experience, catering to nature lovers and city explorers alike.”
Okay. So it seems like there’s potential for differentiating content based on a single source through the use of NotebookLM. I’ll write more about possible uses for NotebookLM with learners and for teachers, but one obvious advantage of this tool is that it’s really easy and quick to differentiate content based on learners’ proficiency levels.
Creating reports in different languages
As a German learner myself, I’m excited by the idea of being able to generate content in German based on English sources that I’m familiar with.
“Minnesota ist ein Bundesstaat in den USA. Es ist das “Land der 10.000 Seen”, denn es hat sehr viele Seen. In Minnesota kann man die Natur und große Städte sehen. Die größte Stadt ist Minneapolis und die Hauptstadt ist St. Paul. In der Natur gibt es viele Parks und den großen See Superior. Minnesota ist ein schöner Ort für ein Abenteuer oder einen Urlaub.”
With this type of feature, I could use a source in my first language, write a summary paragraph based on the content in my target language, then generate an example paragraph in NotebookLM to compare to my own writing. I could also use the chat feature to ask NotebookLM to give me feedback on my paragraph if something isn’t clear.
These types of activities can be done with most generative AI chatbots (ChatGPT, Claude AI, etc). But a benefit of NotebookLM is that all of the content generated based on a source can be kept in one notebook that I can go back to later.
Traveler’s narrative
As stated previously, there are various types of reports you can generate based on your source. Another type of report suggested for my video source is a traveler’s narrative. Without giving any prompts, the this narrative was generated in NotebookLM using my source.
With regard to functionality, creating reports in NotebookLM is simple and straightforward. You’re able to customize the types of reports you create by writing your own prompts. You’re also able to generate content based on suggested content types.
All of these reports are quite polished, but with anything generated or created, it is important to review the content before using it with learners. That being said, the content generated can be a great place to start when creating materials for learners or your own language learning.
You can check out this notebook and everything I generated here.
Audio overview
Alright, let’s take a look at the audio overview content type. Written content has long been a feature of AI chatbots, so one of the more interesting capabilities of NotebookLM is the audio overview. Initially, this feature created podcasts about the sources. When this feature rolled out in 2024, people went wild with the possibilities. You can read more about that in my post from last year.
Since then, the capabilities remain largely the same, but the audio output has been repackaged as “audio overviews” instead of podcasts.
You have the option to generate audio overviews based on your sources with or without customization. If you simply click “Audio Overview” in the studio, an audio overview is generated and kept in your studio. The default settings are to generate a deep dive into the content, and the audio is usually between 10 and 12 minutes long.
In this screenshot, you can see the audio file is playable right in the studio. You also have the option to download the audio and engage with it in interactive mode. We’ll look at this a bit more in a minute. You can listen to the generated audio overview without customization here.
One thing to note before we look at the customization options, you’re also able to use what is generated in NotebookLM as sources. In an Inception-like move, you can download the audio overview, for example, and upload it as a source in your notebook.
I then used the chat feature to generate a transcript of the audio overview, and it seems pretty accurate. I think the most notable benefit of all of this for me as a language teacher is that it works astoundingly quickly. I was able to generate the audio overview and the transcript in about 5 minutes.
Customization of audio overviews
Okay, let’s look at the options we have to customize the audio overviews. In the screenshot below, you can see the options available to customize this type of content.
As you can see, though, you’re able to customize the audio output using the application interface. You can choose the format of the audio: deep dive, brief, critique, or debate. You can select the language of the output as well as the length. Finally, you can specify what to focus on in an additional prompt box.
In this audio overview sample, I used the prompt: “Focus on tourist attractions and things visitors could do in Minnesota. The language in the audio should be appropriate for A1 English learners.” Again, I wanted to create content appropriate for the level of my learners.
As with the reports, you can create content in different languages as well. Again, as a language learner, this is an interesting feature because I can generate audio content based on written sources in my target language. My sources can be in my first language, the target language, or another language. I think the ability to process and consume input in different modalities (in writing and audio) like this is an interesting affordance that language learners and teachers could take advantage of.
Here’s an audio overview I generated in German based on the same source.
Interactive mode
I think one of the most interesting NotebookLM updates is related to the ways you can interact with the content. I wrote a little bit about some of the new interactions available in NotebookLM in a previous post. And now I’d like to explore the interactive mode within the audio overview content type.
In addition to being able to generate audio content based on your sources, you’re now able to interact with the audio content. Basically, the audio overview is similar to a podcast discussion where two hosts talk about the content. By clicking the “interactive mode” hand, you can interact with the hosts.
In the audio overview about my Minnesota source, for example, the hosts start talking about Minnesota being the land of 10,000 lakes. Then they talk about the duality of Minnesota having outdoor and urban attractions. I joined the discussion and asked the hosts to define “duality,” and they explained what the word means for me before continuing on with the overview.
Spoken interaction
I think this type of interaction is novel and mimics real spoken interaction in a way that could be useful for language teachers and learners. Interacting with the content in this way uses both speaking and listening skills, and if you follow along with the transcript while listening, it can also incorporate reading. It just feels like a really interesting way to engage with the content of your sources as a language learner.
The audio produced includes language that is quite natural sounding, although I’m sure there are opportunities for strange usages to pop up. The speakers use fillers like “uh” and “um” appropriately, and their interaction sounds natural in English. I’d be curious to hear from speakers of other languages, though, if the output in those languages also sounds natural.
Something to note is that you can only generate 3 audio overviews per day without upgrading the version of NotebookLM.
Another limitation with the audio overview feature is that you can’t input written content and ask for a direct audio version of the content. For example, if you input a script, you can’t generate a direct reading of the script with the current version of NotebookLM.
Final thoughts & considerations
Ultimately, NotebookLM stands out as a powerful generative AI tool, providing users with the ability to easily input and engage with multiple sources at once. The tool significantly lowers the bar for users who may not be familiar or confident with writing detailed AI prompts.
The features for generating reports are robust, allowing users to easily create content like briefing documents, study guides, and blog posts. A key affordance for educators is the simplicity of differentiating content—whether creating summary paragraphs for A1 English learners or detailed reports for those at a C1 proficiency level. Furthermore, the option to create content in different languages makes it a valuable resource for comparing writing and enhancing language learning.
Beyond written content, the audio overview feature offers a particularly interesting capability, allowing content to be consumed in different modalities. These overviews can be customized by format, language, and length. The interactive mode is also a novel update that could support language learners’ speaking and listening skills.
That being said, it is important to remember that all generated content should be reviewed before using it with learners. Users should also be aware of current limitations, such as the inability to generate a direct audio version of an input script and the limit of three audio overviews per day without an upgrade. Despite these considerations, the content generated by NotebookLM serves as a great starting place when creating materials for learners or supporting your own language learning goals.
Again, you can check out this notebook and everything I generated here.
Leave a Reply