Google teased the product at Google I/O in May, launched Audio Overviews in September and on Thursday moved NotebookLM out of experimental mode and introduced new features to make it more useful.

Updates to the audio summary feature for its AI notetaking and research have driven attention for its podcast-like audio conversations with the ability to guide and focus messages on specific topics instead of just generating audio summaries.

The two new features of Audio Overviews — Guide the Conversation, and Background Listening — further integrate the podcast-generating experience into NotebookLM.

The goal is to create a more seamless user experience. 

Users can now guide the conversation by giving instructions before generating a custom podcast.

NotebookLM is built with Gemini 1.5. After uploading the audio sources, it instantly becomes an expert, grounding responses to the material and providing ways to transform information, according to the blog post.

The tool — which began as an experiment, like so many of Google’s products based on AI, to help research assistants make sense of all their notes, material and more — transforms the text into audio podcasts. One thing not talked about is measurement and analytic tools.

Oxford Road, an independent audio agency, released a report titled What Brands Want: How Audio Publishers Can Win Over Advertisers, offers insights and data on how audio advertisers feel about the state of podcasting, radio and streaming. Many believe measurement tools and innovative applications related to AI are failing to live up to their potential.

Brands largely agreed the tools available to measure audio are less effective than those used for other channels.

Only 14% of chief audio officers (CAOs) said they find the tools available for measuring audio campaigns “somewhat more” or “much more useful” than those for other channels.

Eighty-three percent said they find the tools used to measure audio advertising “much less” or “somewhat less useful” than those used for measuring campaigns in other channels.

Oxford Road polled 49 chief audio officers. Although it is a small sample, brand marketers who own audio advertising responsibilities within their organizations tell an interesting story.

Based on Magellan data, five of the top 15 podcast advertising spenders were represented among these CAOs, all from key brands investing in audio, such as DraftKings, BetterHelp, Indeed, Stamps.com and dozens of others.

Brands also showed significant skepticism over the rise of AI — specifically around the topic of audio advertising automation, which they perceive as undercutting genuine partnerships with creators. With YouTube’s focus on creators, this seems like an interesting situation.

For example, when asked how comfortable they are with AI-generated host reads or cloned voices for personal endorsements, with ratings on a scale of 0-5, 80% of CAOs rated their comfort level between 0 and 2. 

In the study, the topic of “political extremism” has emerged as CAOs’ top brand-safety concern. AI verification, programmatic advertising, synthetic voice and brand safety are all examples of areas where technology is making the marketplace better, according to the report.

For many uses of audio for lower-funnel marketing — indicating the channel’s value in driving performance and for audio-specific performance measurement — CAOs prefer Media Mix Modeling over survey-based or more streaming audio and podcast-friendly pixel-based tools.

Advertisers working with audio believe measurement tools and innovative applications related to AI are failing to live up to their potential for podcasting, radio and streaming.