My first 24 hours with Siri AI on the Mac

The trajectory: does the new Siri stick?

Picture 100,000 Mac users who, like the author, switched Siri off long ago. The model moves them through four stages — still-off skeptics, a fresh trial, daily use, and switched-off-again — with the macOS 27 launch buzz pulling skeptics into a new trial that either hardens into a daily habit or disappoints them back into the off pile.

The first weeks look like a win: daily use climbs fastest and peaks near 40,500 users around day 31. But it doesn't hold. Around day 18 the "switched it off again" curve overtakes daily use, and by the end of the run the lapsed pile (about 47,800) is larger than the daily-user base (about 39,100). The new Siri earns a genuine second look — it just doesn't convert a majority of those looks into a habit.

The most revealing part is which lever matters. Re-running the model shows the finish is far more sensitive to the win-back rate — lapsed users tempted back by later updates — and to churn than to first-day conversion. The first 24 hours set the mood; the updates that follow set the outcome.

What's differentBaseline run — the model at its default settings, no interventions yet. Every later run is measured against this.

The field: where Siri AI lands in the assistant race

Weighted the way a Mac user actually decides — privacy and OS integration carry the most weight, raw reasoning less — the new Siri AI vaults from the middle of the pack in 2025 (0.44, neck-and-neck with Gemini and behind Copilot) to the front of the field in 2026 (0.84), ahead of Copilot (0.55), Gemini (0.54), ChatGPT (0.50) and Alexa+ (0.32).

But the lead is conditional. It rests on the two factors Apple is strongest on — on-device privacy (0.90) and integration into the Mac (0.85). On raw reasoning quality Siri still sits near the back at 0.55, behind every rival except Alexa+. Reweight the index toward intelligence and the lead evaporates.

This is a snapshot built from editorial scores, not benchmarks — an argument about which factors matter, not a measurement. That is exactly why it's interactive: change the weights to match what you care about and see who wins.

What's differentComposite index: 5 entities across 2 periods. siri ranks highest; weighted across 5 factors.

What it would take to make it stick

So what flips the picture? Introduce one disciplined intervention: Apple keeps shipping the kind of capability updates that pull lapsed users back and keep daily users from drifting away. In the model that means roughly halving churn (0.04→0.02), more than doubling the win-back rate (0.06→0.15), and lifting first-day conversion (0.12→0.18).

The outcome inverts. Daily users finish at about 75,800 instead of 39,100, the lapsed pile shrinks to roughly 15,700 from 47,800, and the day-18 crossover never happens — daily use leads from the first week to the last. The takeaway for the next macOS point release: the battle for Siri isn't won in anyone's first 24 hours. It's won in the updates that follow.

How to read this

Two caveats so you can weigh the charts honestly. The trajectory is a behavioural model, not adoption telemetry — a day-one review has no user data, so the rates are plausible illustrative values chosen to expose the dynamics, and the levers are there for you to disagree with. The assistant index is an editorial composite: the scores are a defensible read of each assistant's strengths in 2026, not benchmark results, and the ranking moves with the weights. Both are arguments you can interrogate by dragging the controls — not measurements to take on faith.