Our Biggest Accuracy Upgrade Yet: Podcast AdBlock 1.5
In early 2026, we shipped Podcast AdBlock 1.5 — the biggest update to our ad detection system since launch. Here's what changed and why it mattered.
What Improved
The core of the update was a completely retrained detection model. Version 1.5 used a larger training dataset and a refined approach to identifying ad segments in podcast audio. The result was noticeably better accuracy, especially on shows with host-read ads that blend naturally into the content.
Key Numbers
- Ad detection accuracy improved significantly across our test library of popular shows
- False positives (content incorrectly flagged as an ad) dropped considerably
- Processing speed improved, with most episodes analysed faster than before
How It Worked
The detection system analysed audio patterns, speech cadence, and contextual cues to identify likely ad segments. Version 1.5 improved on all three fronts:
Better audio fingerprinting — the model got better at recognising the subtle audio differences between editorial content and ad reads, even when the same host delivers both.
Improved context awareness — by looking at longer windows of audio, the model could better understand transitions between content and ads.
Faster inference — optimisations to the model architecture meant the analysis ran quicker on-device without sacrificing quality.
Community Feedback
The response from users was positive. Many reported that shows which previously had inconsistent detection were now handled much more reliably. The improvement was particularly noticeable on interview-format podcasts and shows with mid-roll ad breaks.
Looking Back
Podcast AdBlock 1.5 represented a significant step forward in what was possible with on-device audio analysis. The techniques developed for this update — better fingerprinting, wider context windows, and efficient on-device processing — became foundational for everything that followed.
The podcast listening experience continues to evolve, and tools that help listeners get the most from their time remain as relevant as ever.