Google on Thursday updated the audio summarization feature of its AI note-taking and research assistant NotebookLM, which recently gained widespread attention for its podcast-like audio conversations based on content shared by users, with the ability to direct those conversations and focus on specific ones instead To focus on topics, simply create holistic audio summaries.

Today, audio overviews in NotebookLM allow users to process and understand the information in long documents or videos through AI-generated audio conversations. Shortly after launching last month, the feature helped NotebookLM gain attention as many began sharing audio summaries of their content on social media, including those created using their diaries or journals.

While Google hasn't revealed what traction this has given NotebookLM, data from website traffic analytics platform SimilarWeb suggests that NotebookLM saw traffic increase by over 371% in September to 3.07 million monthly visits, up from 652,181 a month ago.

Previously, audio digests automatically generated AI conversations from users' sources. However, since conversations sometimes revolve around unimportant content, Google is introducing an update that allows you to adapt the overviews to your needs. This allows users to focus the tone more on a specific topic in their content.

A dedicated Adjust control is available before the existing Generate button, allowing you to direct the AI ​​presenters in the audio to focus on a specific point.

Raiza Martin, product lead for NotebookLM and senior product manager for AI at Google Labs, told TechCrunch that the update gives users the ability to move the AI ​​in the direction they want.

“The entire team worked hard to listen to and analyze all the feedback we received. And the main feature that people wanted was just to give AI a little nudge,” she said.

Customizing audio summaries can also help reduce hallucinations to some extent, that is, the cases in which the AI ​​cooks content itself. Still, Martin said, the NotebookLM team is monitoring user feedback and trying to detect hallucinations as quickly as possible.

She also emphasized that customizing audio summaries does not mean using user instructions to train the AI ​​model.

“In general, we don't train on user data. No matter how you use it or what questions you enter, whatever answers you enter, we don't train the models on it,” she said, adding: “We get a lot of feedback from our users.”

Besides the customization option, users get background listening in audio overviews. This allows you to continue working in NotebookLM, querying your sources, getting citations, and exploring relevant quotes while the audio plays in the background.

NotebookLM was originally unveiled as a project at Google's I/O developer conference last year and was first made publicly available in the US in December. In June, the company expanded to markets including India, the UK and more than 200 countries. Although the product initially found some acceptance in education and research use cases, companies and organizations only started trying it after Google expanded its support to more sources and added new features.

According to Google, over 80,000 organizations now use NotebookLM and see it as an opportunity to explore monetization. Hoping to capitalize on this momentum, the company launched the NotebookLM Business pilot program on Thursday.

Companies can apply for the pilot project. If accepted, Google says they will receive early access to product features, training and email support.

Martin told TechCrunch that as part of the business pilot, her team is training organizations interested in using NotebookLM on how other companies use it.

“We (also) want companies to tell us what features we want to provide,” she said.

NotebookLM Business general availability and pricing will be announced later this year. However, Google has not yet announced the exact schedule and pricing tier details.

According to SimilarWeb, NotebookLM currently receives 4.17 million monthly visits, with 2.5 million coming from desktop and 1.6 million from mobile.

The Assistant currently does not have its own mobile app and is available on all screens via its website. However, Martin told TechCrunch that the team is actively exploring a native mobile experience to expand NotebookLM's presence among smartphone users. Additional voices, languages, and audio overview controls are also being explored.

Additionally, the team explored and prototyped a different number of speakers – to go beyond the existing two speakers for AI audio discussions – although it likely won't be available soon, as Martin said it wasn't the most requested feature by users.

Last month, NotebookLM added YouTube videos and audio files as sources for creating summaries, alongside existing sources such as Google Drive, URLs, PDFs, and text.

According to Martin, NotebookLM sees PDFs and YouTube videos as the two most important sources. The team also observed a “very high percentage” of users listening to an audio overview and using chat. The next largest group is made up of users who only use the chat without generating an audio overview.