Personalization & memory
ChatGPT separates AI behavior customization from technical settings. Users pick named tone presets, fine-tune characteristics, declare profile fields, and manage inferred memories with search, deprioritize, and version restore.

Personalization tab: base style and tone, characteristics, custom instructions, and About you fields.
What's happening
ChatGPT treats personalization as a primary product surface: a dedicated settings tab with named tone presets, characteristic fine-tuning, explicit profile fields, and a full memory management stack — list, deprioritize, and version restore.
Patterns
Named tone presets and characteristic sliders let users shape voice without writing system prompts
Searchable memory rows with prioritize, delete, deprioritize states, and point-in-time restore
Master memory toggle separates declared profile from inferred chat memories
UX Insights
- •Personalization tab sits beside Data controls — behavior prefs are not buried
- •Tone presets use outcome labels (Professional, Friendly) not model parameters
- •Auto-save with toast confirmation removes friction for global preferences
- •Memory list shows system deprioritization with timestamps, not silent deletion
- •Version history lowers risk of turning memory on
- •Bing/search disclosure on memory use builds trust for retrieval personalization
Design Decisions
By giving Personalization its own tab and layering explicit prefs over inferred memory, ChatGPT positions itself as a persistent assistant. Named presets onboard casual users; characteristic sliders and custom instructions capture power-user nuance; memory list plus version restore address the creepiness problem without turning memory off entirely.
More case studies

Personalization & memory
Claude shapes behavior through natural-language profile instructions, opt-in chat memory under Capabilities, cross-platform memory import, and named voice style presets.

Personalization & memory
Perplexity personalizes search relevance: occupation and custom instructions, demographic modals, response length prefs, Health and Finance vertical profiles, and search-history memory with Brain upsell.

Incognito Mode
Claude offers an incognito (private) mode that lets users have conversations without saving them to history or using them to improve the model. The mode is toggled via a ghost icon in the header.

Settings
Claude offers a comprehensive settings interface with a strong focus on AI personalization, allowing users to customize their profile, work context, personal preferences, appearance (including dyslexic-friendly fonts), and notification controls.