One output depth rarely fits both skim and deep work.
← Back to patterns
Google
Magic Ink: Information Software and the Graphical Interface Sensecape: Enabling Multilevel Exploration and Sensemaking with Large Language Models Accordion Editing and Apple Picking: Early UX Patterns of AI
Inline & Contextual
EmergingSemantic Zoom
Complexity on a Single Axis
The same AI output rendered at three levels of abstraction — Headline, Summary, Full Detail — controlled by a single slider. The chart is a zoom level too: the series is resampled per depth rather than pinned at full fidelity beside a one-line summary. Depth can be applied to the whole document or to one section at a time, so two sections can be skimmed while a third is read in full. No shipped product resamples the visualization alongside the prose; closest neighbours are length controls (Granola, Apple Intelligence), and Sensecape (UIST 2023) is where the three-level move is designed and measured.
Metaphor
Map
Interaction
Slider, Level Select, Per-Section Dial
Use case
Research Reports, Documentation, Data Summaries
Failure mode
Hidden State
Control mechanisms
3-Level Zoom Slider · Per-Section Scope · Resampled Chart · Level Labels
Reference examples
Gemini Notebook Audio Overview · Granola · Apple Intelligence Summaries
Interactive Example
The live side-by-side is the comparison.
This prototype already contains the contrast between chat and the pattern response. It stands as the primary comparison surface for this page.
One depth for skimming and deep work fits neither.
A single output length is always wrong for someone. The skimmer drowns in detail; the deep reader is starved of it. In chat the only fix is to keep asking for a shorter or longer version, and each request breaks continuity and reorders the content, so you lose your place in the very answer you were navigating.
Semantic Zoom binds one answer to three abstraction levels — Headline, Summary, Full Detail — under a single slider. The content stays anchored while depth changes, so the reader moves along complexity without losing context, zooming in only where they need it. The same object, seen at the resolution the moment calls for.
Let the reader change the resolution without changing their place.
DOPE Lens
How this pattern improves AI interaction quality.
Observability
A persistent indicator states the position exactly — "Level 2 of 3 — Summary · ≈ 53 words · 0.4% of source" — with the count derived from the copy on screen, so it cannot drift
Directability
One slider moves the whole report; per-section scope gives each section its own dial, in both directions, with no regeneration and no lost position
Predictability
Every level declares a contract for both channels — 1 sentence + 1 sparkline, 4 findings + 6 annual marks, 3 cited sections + 12 observations — so the reader knows what a zoom step returns before dragging
Explainability
Zooming changes the representation, not the crop — prose from sentence to cited argument, chart from sparkline to fitted trend with its scenario band — and each level names its method
Fit
Where it earns its place.
Use it
when
when
One answer needs to serve both skim and deep-read modes without rerunning generation.
Avoid
when
when
Meaning does not survive well across summary and detail layers.
In the wild
Where this pattern already ships.
The pattern is shipping but still maturing across products. Entries below show how different teams have approached it; expect variation as the space settles.
Gemini Notebook Audio Overview
Length control: Shorter / Default / Longer for audio overviews, plus formats (Brief / Deep Dive / Critique / Debate). One-control-multiple-depths, exactly the pattern.
Visit Google ↗
Granola
After a meeting, switch between TL;DR / Full / Action Items tabs on the same source content — semantic zoom as tab switcher.
Visit Granola ↗
Apple Intelligence Summaries
Apple
Caveat — closer to format presets (Summary / Key Points / List / Table) than abstraction-depth on a single axis. Still useful as a zoom-adjacent shipping reference.
Visit Apple ↗
Research Frame
Why this pattern exists, and where it breaks.
Bind one answer to multiple abstraction levels on a single control.
Users must repeatedly ask for shorter or longer versions and lose continuity.
Compression levels can hide nuance or imply false completeness.
Common questions
What people ask about this pattern.
When should I use the Semantic Zoom pattern?
+
One answer needs to serve both skim and deep-read modes without rerunning generation.
When should I avoid the Semantic Zoom pattern?
+
Meaning does not survive well across summary and detail layers.
What problem does Semantic Zoom solve?
+
One output depth rarely fits both skim and deep work.
Why is chat the wrong fit for this?
+
Users must repeatedly ask for shorter or longer versions and lose continuity.
Grounded in Research
Not just opinion.
The papers, guidelines, and usability studies this pattern rests on.
Bret Victor
worrydream.com · long-form essay · 2006
Distinguishes information software (you read) from manipulation software (you operate) and argues most products misclassify themselves. The case for generated answers that are interfaces, not paragraphs.
Sangho Suh, Bryan Min, Srishti Palani, Haijun Xia
University of California, San Diego · UIST '23 · 2023
Names and evaluates semantic zoom as an LLM interaction: the response is fed back to the model and re-rendered at three levels — Lines / Summary / Keywords — rather than cropped, with depth either bound to canvas zoom or pinned per node. The empirical case that a single depth axis over one answer is usable, not merely appealing (12.58 uses per session, SD 8.92). Pairs with the NNGroup study below: that one names the failure mode, this one designs and measures the fix.
Sarah Gibbons, Tarun Mugunthan, Jakob Nielsen
Nielsen Norman Group · 2023
Empirical usability study naming the exact chat failure modes — expanding/condensing whole answers when only a span needs work, manually harvesting fragments across regenerations — that selection-scoped transforms and single-axis depth controls exist to fix.
Related Patterns
Curated connections, not just neighbors.
Hand-picked relationships — what this pattern pairs with in practice, what it replaces, what it builds on, and what builds on it.
Often paired with
Progressive Disclosure
Semantic Zoom sets how deep the answer reads; Progressive Disclosure does the same for the controls around it.
Often paired with
Reasoning Trace
Semantic Zoom sets how deep you read the answer; Reasoning Trace opens the assumptions behind any claim at that depth.
Often paired with
Citation Trail
Semantic Zoom's shorter levels can hide nuance; Citation Trail lets you hover any sentence to check the source behind it.