Reading vsScanning

Jakob Nielsen's research showed it decades ago: users don't read web pages. They scan them. Eyes jump from headline to headline, from bold word to bold word, picking up fragments and assembling meaning. Yet AI products still serve walls of unformatted prose as if we're reading novels.

The F-Pattern

Eye-tracking studies consistently show an F-shaped reading pattern on the web. Users read the first line, scan partway through the second, then skim down the left edge. Most of the content on the right side of a paragraph is never seen. This isn't laziness. It's efficiency. The brain triages information ruthlessly.

Chat interfaces fight this pattern. Every message is the same width, same weight, same visual priority. The eye has no landmarks. Scanning fails, so users either read everything slowly or give up and skim randomly.

Design for scanners, and readers will thank you too. Design for readers, and scanners will leave.

Scannable Trees

• Node titles are scannable headlines, bold and concise
• Color coding provides instant categorization without reading
• Spatial layout creates landmarks the eye can anchor to
• Hierarchy is visual, not buried in paragraph structure

Respecting Behavior

Persephonie doesn't try to change how people consume information. It designs for reality. Nodes are scannable by default. Titles are bold. Descriptions are short. Color tells you the sentiment before you read a word. We meet users where they are, not where we wish they were.