Spotify plans to scale AI into podcast summaries and audio ads

Finding a new podcast can feel like homework, and creating audio ads has always been too expensive for most businesses. Spotify thinks generative AI can solve both problems at once.

During Spotify’s second-quarter earnings call, CEO Daniel Ek teased several upcoming AI features. The streaming giant plans to expand personalized voice features across its entire catalog, auto-summarize podcasts, and let advertisers spin up thousands of tailored audio ads in seconds.

Automated summaries and thousand-variant audio ads

Earlier this year, Spotify rolled out its AI DJ feature, which serves up music alongside an AI voice that adds commentary about your favorite tracks and artists. Ek called it one of his personal favorite launches in recent years and confirmed that Spotify intends to use similar contextualization across the rest of the app.

One obvious target is podcasts. Ek pointed out that diving into an unfamiliar show is tough for listeners. By using generative AI to summarize episodes, Spotify wants to make podcasts easier to discover, which Ek believes will drive creator growth and listener engagement.

The real moneymaker, however, is advertising. Ek explained that generative tools will dramatically cut the production cost of audio ads.

Instead of paying for studio time and voice talent to cut a single ad, brands could theoretically generate thousands of hyper-targeted audio variations and A/B test them across Spotify’s ad network. Ek noted this kind of rapid testing is already standard for text-based online ads, but has never been practical for audio.

Whispers, shouts, and sarcastic synthetic voices

Spotify’s AI ambitions aren't just talk. On July 20, a patent filed by Spotify in February was officially published, revealing a text-to-speech synthesis engine designed to inject emotion and intent into synthetic speech.

According to the patent, the system can generate realistic voices that convey anger, happiness, sadness, and even sarcasm. It can also produce speech in different accents, or alter its delivery to whisper or shout.

This technology builds directly on Spotify’s acquisition of London-based voice startup Sonantic last year, which originally powered the DJ feature. The patent suggests Spotify could use natural-sounding synthetic voices for full audiobook narration, directly taking on Apple Books, which launched AI narration for select titles earlier this year.

The push into AI comes as Spotify works to fix its bottom line. The company reported 551 million monthly active users and 220 million paying subscribers, up 17% year over year. Second-quarter revenue hit roughly €3.2 billion, but operating losses still totaled €247 million. Just a day before the earnings call, Spotify announced a price hike in the U.S., bumping its individual Premium tier from $9.99 to $10.99 per month.

Why it matters

For regular users, AI summaries should make browsing long-form podcasts much less tedious, and realistic voice synthesis could rapidly expand the catalog of narrated audiobooks. The catch is your ad experience: free users will likely hear far more personalized, programmatic audio ads designed specifically for their listening habits.

For the digital audio industry, we think Ek’s strategy is remarkably sharp. Rather than trying to generate synthetic music and alienating record labels, Spotify is using AI where the business actually leaks money: discovery and ad production. If audio ads become as cheap and modular to test as Google search ads, Spotify could attract a massive wave of small-business advertisers that previously couldn't afford radio or streaming spots.

The main risk is user fatigue. When voice synthesis becomes good enough to simulate emotion, whisper, or use sarcasm, the line between helpful host and manipulative ad delivery gets uncomfortably thin.