NotebookLM Smashes Its Biggest Bottleneck: Custom Chat Prompts Jump 20x

Google has quietly but significantly upgraded NotebookLM’s custom chat capabilities, expanding the instruction limit from a meager 500 characters to a massive 10,000 characters. Confirmed by the NotebookLM team on X, this 20-fold increase transforms the feature from a simple tweak into a robust customization engine.

Why This Limit Mattered (and Why It's Gone)

Until now, the "Configure notebook" settings—where users define the AI's role, tone, and specific rules—were incredibly restrictive. Users trying to create complex personas or strict output formats often found themselves fighting against the character count, forced to truncate their instructions or rely on vague prompts.

This bottleneck was particularly jarring because it existed alongside recent backend improvements. In late October 2025, Google polished the underlying engine to improve context understanding and reasoning. However, as noted in reports, those capabilities were often leashed by the inability to give the model detailed instructions.

With the limit lifted to 10,000 characters, the model’s reasoning capabilities can finally stretch their legs. This change is the latest in a rapid upgrade cycle for the platform, which has recently improved conversation retention, context handling, and response coherence. The character expansion appears to be the "missing piece" that allows users to fully leverage these underlying improvements for long-term projects.

What You Can Build With 10,000 Characters

This isn't just about longer sentences; it's about structural complexity. The expansion allows for granular control over how the AI processes information. Google demonstrated this potential by releasing specific prompt examples that were previously impossible to implement fully.

The Bigger Picture

This update fundamentally changes the utility of the "Custom mode." Before, it was too limited for anyone trying to build a consistent editorial assistant or a rigid research partner. Now, users can set comprehensive context, goals, constraints, and personality traits without butchering their instructions.

It is important to note that while the instruction limit has skyrocketed, the platform’s broader framework remains the same. The update does not change upload limits or daily usage caps. However, by removing the instruction bottleneck, Google has effectively unlocked the existing power of the model, allowing for richer personas and far fewer generic replies.