Things to know

Long definition

Things to Know surfaces for queries Google reads as multi-faceted, where users likely want to explore the topic from several angles before refining. Examples: "acrylic painting", "starting a business", "type 2 diabetes". The block typically shows 5-10 subtopic cards or chevrons — "Getting started", "Common challenges", "Best techniques" — that expand inline to reveal short summaries.

The feature was announced as one of the first user-visible products of MUM (Multitask Unified Model), Google's multimodal model trained across 75 languages. MUM clusters subtopics by reading how concepts co-occur across the corpus, not just by query co-occurrence. That lets it group "stretching", "hydration", and "cool-down" under "running recovery" even when those exact phrases never appear in the same query.

Implementation varies. Some Things to Know expansions are clickable links to refined queries — clicking "Getting started" runs a sub-query. Others are display-only summaries pulled from top organic results, with the source URL cited but not the primary CTA. Mobile layouts emphasize the inline-expansion pattern; desktop sometimes renders the block as a horizontal rail.

Optimizing for Things to Know overlaps heavily with featured-snippet optimization. Subtopics get filled from passages in top-ranking pages, structured around clear subheadings. A long-form guide with H2s mapping to natural subtopics ("Equipment needed", "Common mistakes", "Cost breakdown") gives Google extractable units. Pages built as one undifferentiated wall of prose rarely supply Things to Know answers even when they rank organically.

The feature is volatile and rotated heavily by Google. A query showing Things to Know in March may not show it in June. Treat it as a bonus surface, not a primary investment.

Common misconceptions

• "Things to Know is the same as PAA." Different feature. PAA is question-shaped and click-driven. Things to Know is subtopic-shaped and often expansion-only. The MUM-driven clustering also makes Things to Know more conceptual than PAA's query-co-occurrence questions.
• "There's a schema for Things to Know." No schema triggers it. Subtopic markup (when it exists) targets other features. Things to Know is purely algorithmic, sourced from passage extraction.
• "It only shows for medical and how-to queries." It shows across many verticals — hobbies, education, business, lifestyle, products. The common thread is multi-faceted intent, not topic.
• "Clicks from Things to Know show up as a separate Search Console category." They're folded into normal Web search clicks. There's no breakout report for Things to Know specifically.