Changes we're making to Google Assistant
Google is stripping a wide range of capabilities from Google Assistant — from sending emails and managing calendar events by voice to third‑party integrations like AnyList and Calm — while framing the move as part of “improving” the product. Commenters see it instead as cost‑cutting and enshittification: narrowing Assistant to basic searches, timers and media, likely to push users toward Google’s own services or a future paid, LLM‑based assistant. The changes also revive broader doubts about the long‑term reliability of cloud assistants, their poor feature discoverability, and the wisdom of tying everyday workflows and home devices to products Google can and does abruptly degrade.
Feature removals and perceived regressions
- Many commenters see the removed capabilities (emailing, rescheduling calendar events, payments/reservations, media alarms, cookbook/recipes, AnyList-style shopping workflows, etc.) as the “assistant” part of Assistant.
- Users are baffled by cuts to seemingly trivial features like stopwatch control and caller ID on speakers, and by changes that push them back to manual app usage or generic search/YouTube results.
- Several people report that Assistant and related Google features have already degraded over the last few years (timers failing, reminders misinterpreted, car parking/travel integration gone, worse hands‑free usability).
Business, layoffs, and internal politics
- Many infer this is driven by cost-cutting and codebase reduction after recent layoffs in the Assistant team.
- Some suggest org politics and a shift toward driving Search and YouTube metrics: defaulting to search/YouTube and weakening hands‑free actions makes the product more “measurable” and ad‑adjacent.
- Others see it as part of a broader “enshittification” pattern across Google products, where features are removed or bent toward monetization.
Third‑party integrations and lock‑in
- Removal of integrations (e.g., AnyList, Calm, broader Conversational Actions) is widely viewed as user-hostile and possibly anti‑competitive, forcing people into Google’s own shopping lists and services.
- Some note Google has already shut down the broader conversational ecosystem; others link to docs showing certain “built‑in intents” remain, but only on phones and not speakers.
LLMs and future direction
- A common hypothesis: hard‑coded features are being removed in preparation for a generic LLM-based assistant (e.g., Gemini/Bard-style), but timing and communication are seen as poor.
- Some are optimistic LLMs plus function-calling could finally make assistants truly flexible and discoverable.
- Others argue LLMs won’t fix missing business logic, discoverability, latency, or the fundamental friction of conversational UIs for transactional tasks.
UX, discoverability, and reliability
- Major themes: zero discoverability of commands, inconsistent behavior over time, and assistants that are too limited to learn yet too unreliable to trust.
- Several prefer strict, predictable voice grammars over “smart” but flaky natural language systems.
Privacy, trust, and alternatives
- Some reject cloud-based assistants entirely over privacy concerns and product instability.
- Others are migrating to Alexa, Siri, Home Assistant, or DIY/local voice solutions, or simply using devices as dumb speakers/photo frames.