Designing Around Apple Intelligence Writing Tools Beta
Quick Answer: When testing the iOS 27 beta, we realized Apple Intelligence was sanitizing users' raw emotional logs. We disabled system-level text rewriting in our core fields for three reasons: it increased logging time by 42%, it sterilized authentic negative emotions into "friendly" corporate-speak, and it distracted from our core 3-second emoji-based tracking. While we offer optional AI insights using cloud-stored, de-identified data via Supabase, we believe the moment of logging should remain raw and unfiltered.
*
During our early internal tests of the iOS 27 beta, we noticed a strange anomaly in our testing cohort's data. The frequency of "Rough" and "Low" mood check-ins remained stable, but the text memos attached to them suddenly sounded like PR statements.
A beta tester who previously wrote, "exhausted, boss is a nightmare, want to quit," was suddenly logging, "I am experiencing a challenging professional environment today and feeling quite fatigued."
We hadn't changed our prompt. The users hadn't suddenly enrolled in corporate communications training. They were just playing with the new system-wide Apple Intelligence Writing Tools, tapping the "Professional" or "Friendly" rewrite buttons out of curiosity.
As the Head of Product at ViviDiary, this was a massive red flag. Our product is designed to be a lightweight modular mood and life tracker. We want your raw, unedited reality. Designing around Apple Intelligence Writing Tools beta became our immediate priority, not to integrate it deeper, but to figure out how to intentionally block it from ruining the authenticity of our users' diaries.
The Sterilization of Emotion: Our iOS 27 Beta Discovery
When a platform vendor introduces a system-level feature, the default product instinct is to support it immediately. When we worked on our iOS 27 integration for Apple Health, we adopted the new APIs without hesitation because they passively synced data without altering the user's truth.
Apple Intelligence's Writing Tools are different. They actively mediate the user's thoughts.
In our testing, we found that when users are presented with a button that promises to "fix" or "rewrite" their text, they click it. It's a digital pacifier. But a journal entry isn't an email to your manager. It doesn't need to be grammatically perfect, concise, or friendly.
!User testing session showing a before and after of a sterilized journal entry
When we looked at the test logs, the sterilization of emotion was glaring. The raw edges of human frustration, sadness, and even overwhelming joy were being sanded down into a homogenous, polite AI voice. We realized that if we allowed system-level generative text tools to operate unchecked inside ViviDiary, we would be complicit in erasing our users' actual emotional fingerprints.
What We Rejected: Auto-Correcting the Four Emotion Journaling Method
Before deciding to disable the feature, we debated leaning into it. We asked ourselves: Could we use Apple Intelligence to help users articulate their feelings better?
We considered building a custom integration that mapped the AI's rewriting capabilities to the four emotion journaling method UX we initially studied when we abandoned the Likert scale. The idea was that if a user typed a messy rant, the AI could automatically categorize it into core emotional buckets and rewrite the entry for clarity.
We built a prototype and put it in front of 15 beta testers. The results were decisively negative:
- Loss of Ownership: 11 out of 15 users reported feeling disconnected from their own entries after the AI rewrote them. One user noted, "It looks like someone else's diary now."
- The "Correction" Implication: Offering to rewrite a diary entry subtly implies that the user's original expression was "wrong." When you are logging a "Rough" day, the last thing you need is a machine correcting your grammar or tone.
- Friction: Our core value proposition is speed. You can log your day in ViviDiary in under 30 seconds using just a mood selection (Great/Good/Okay/Low/Rough) and a few emojis. Adding a step where users review an AI-generated rewrite pushed the average logging time up by 42%.
We firmly believe in designing against AI slop in wellness apps. We rejected the auto-correcting approach because a diary must remain a safe space to be messy.
3 Reasons We Disabled Apple Intelligence Writing Tools
Ultimately, we made the call to disable the `UITextView` properties that allow system-level Apple Intelligence Writing Tools within our core text input modules. Here is exactly why we did it:
1. Protecting the 3-Second Check-In ViviDiary is modular. The only required input is your Mood. Everything else—memos, voice, photos, the Focus module (Routines + Todos), and the 22 manual emoji modules—is strictly opt-in. We rely heavily on anti-attention design to get you in and out of the app quickly. The Writing Tools beta introduced a distraction layer that kept users lingering on the input screen, tweaking sentences instead of logging their mood and moving on with their lives.
2. Preventing Prescriptive AI We have a strict rule: we sit beside you, we don't coach you. We don't use pressure-style streaks, traffic-light progress UIs, or guilt-inducing "you missed today" notifications. Allowing an AI to rewrite your feelings crosses the line from observation to prescription. We've seen the damage caused by an algorithm dictating their feelings, and we refuse to build that into our product.
3. Maintaining Our Single "Warm" AI Tone ViviDiary does have an AI feature, but its role is highly constrained. It is an optional supporting tool for days you want to record more deeply. It operates with a single "Warm (따뜻하게)" tone and never saves or confirms without explicit user review. Apple's system tools offer "Professional" and "Concise" tones, which directly conflict with the empathetic, non-judgmental environment we've cultivated.
Privacy Reality Check: Cloud-Stored and De-Identified
With the launch of Apple Intelligence, there is a massive industry push toward privacy-first AI. The marketing narrative is that if data is de-identified before processing, it's inherently safe.
I want to be completely transparent about ViviDiary's architecture here, because we take a fundamentally different approach to privacy.
ViviDiary protects your privacy through data minimization and de-identification. Our data layer is cloud-stored using Supabase. We do this because users expect seamless cross-platform syncing, and they trust us to maintain a reliable archive (like Day One) that survives if they drop their phone in a lake.
So, how do we protect your most intimate thoughts if we aren't storing them locally?
Our privacy model is built entirely on data minimization and de-identification. Before any text ever touches our optional AI processing layer, it is stripped of identifying markers. We keep your data cloud-stored and de-identified. The privacy comes from the fact that the system processing the text has no idea who you are, not from the hardware it's running on.
When we looked at Apple's AI tools, we realized that relying on hardware-based processing wasn't a silver bullet for privacy if it simultaneously degraded the quality and authenticity of the user's data. We prefer our method: secure cloud storage, strict de-identification, and keeping the AI strictly opt-in.
(Note for new users: ViviDiary is Free for all input modules, unlimited mood logging, a 3-month calendar archive, and up to 3 Routines / 5 Todos. For users who want deeper historical pattern discovery, Premium is $2.99/mo or $11.99/yr.)
Embracing the Rawdogging Reality Trend in UX
There is a growing cultural shift, particularly among Gen Z, often referred to as "rawdogging reality." It's a rejection of constant digital mediation. It's choosing to sit on a flight without headphones, or choosing to feel a negative emotion without immediately opening TikTok to numb it.
When we are designing for the rawdogging reality trend, we are designing for friction-free honesty.
Users don't want an app to fix them. They want a blank space that accepts them. By disabling system-level AI rewrites, we are protecting that blank space. If you had a "Rough" day and your Routine fell apart, ViviDiary's Mirror feature will quietly observe that pattern for you on Sunday morning. It won't scold you, it won't break a streak, and it certainly won't rewrite your frustration to sound more "professional."
We build software to help you notice your life, not to edit it for you. Sometimes, the best product decision you can make with a new technology is choosing exactly where not to use it.
