MUM

Definition

MUM (Multitask Unified Model) is a large language model introduced by Google in May 2021 for Google Search. Google described it as "1,000 times more powerful than BERT." Key MUM capabilities: (1) Multimodal — MUM can process both text and images (and potentially audio and video in future applications), enabling visual search and combined text+image query understanding; (2) Multilingual — trained on 75 languages simultaneously, allowing cross-language knowledge transfer (information learned from content in one language benefits understanding in others); (3) Multi-step reasoning — can understand complex, multi-step queries that would previously have required multiple searches (e.g., "I want to hike Mt. Adams — what should I prepare differently than when I hiked Mt. Fuji?" requires understanding hiking, both mountains, differences in terrain/climate/altitude, and preparation changes). MUM is used in specific Google Search features: visual search (Google Lens), "Things to know" in search results, and some featured snippet improvements. MUM is not the primary ranking algorithm — it is applied to specific complex query types and feature generation rather than general web document ranking.

Why it matters for SEO

MUM\'s practical SEO implication is the continuing evolution away from keyword-matching and toward genuine content comprehensiveness. Content that deeply covers a topic with accurate, nuanced information — that could answer the follow-up questions a user would have after their initial query — aligns with what MUM-era search systems reward. For multilingual site operators, MUM\'s cross-language capability means that a site with strong content in one language may benefit from Google\'s understanding of its topic area in other languages.

How DeepSEOAnalysis checks this

MUM\'s effects on rankings are diffuse and not directly auditable through technical SEO checks. DeepSEOAnalysis checks the structured and technical signals (structured data, canonical, AI visibility) that support content discoverability and citability in all algorithm contexts. The AI visibility scoring — particularly FAQPage schema and content chunkability — aligns with the kind of structured, complete content that benefits from advanced semantic AI systems like MUM and its successors.