Gracenote flags AI errors in content discovery as streaming evolves

MUMBAI: When AI starts making up movie plots, the credits roll a little too early. As streaming platforms increasingly lean on artificial intelligence to help viewers find their next binge, a new study by Gracenote, Nielsen’s content intelligence business, suggests the technology still has a script problem when it isn’t backed by trusted data.

In its report, Plot Holes in AI: Why Ungrounded LLMs Can’t Fix Content Discovery, Gracenote analysed 2,600 film and television titles across 13 global markets and found that AI models operating without access to verified entertainment databases frequently generated inaccurate or entirely fabricated information.

The study revealed that 506 out of 2,600 titles contained completely fabricated metadata. Less than one-third of responses generated by ungrounded AI models were rated as high quality, while between 77 per cent and 91 per cent fell into zero-, low- or medium-quality categories, raising concerns about the reliability of AI-powered content discovery.

According to the report, AI models struggled particularly with newly released films and series because of knowledge cut-off dates, making it difficult to identify or describe recent content accurately. They also confused productions with similar titles, blending together plots, genres and cast details from entirely different shows and films.

The problem became even more pronounced in markets where non-English content is underrepresented in AI training datasets. Gracenote found significantly higher hallucination rates in such regions, highlighting the need for AI systems to be grounded in accurate, localised entertainment metadata.

The findings carry particular significance for India, where audiences consume content across multiple languages and streaming platforms. As OTT services increasingly experiment with AI-powered search, recommendations and conversational assistants, the report suggests that continuously updated, trusted metadata will be essential to deliver accurate viewing experiences.

The research also underscores why content discovery has become a strategic priority for streaming services. Gracenote’s consumer survey found that 64 per cent of viewers already know what they want to watch before switching on their television, while 84 per cent consider user experience an important factor when choosing a streaming platform.

Despite AI’s current shortcomings, consumer optimism remains strong. The study found that 66 per cent of viewers believe artificial intelligence will play an important role in improving entertainment experiences in the future, provided it delivers recommendations that are relevant, accurate and rooted in reliable data.