Google Messages Smart Reply Gets Major Upgrade

For years, Google Messages users have lived with the same small anxiety: hovering a thumb near a Smart Reply chip and knowing that one errant tap could fire off an AI-generated "Sounds good!" to the wrong conversation at the worst possible moment. That era is finally ending.

In early March 2026, Google began rolling out a "Tap to Draft" toggle in the beta version of Google Messages, fundamentally changing how Smart Replies work. Instead of instantly sending a suggested response the moment you tap it, the feature drops the text into the compose box, giving you a chance to review, edit, or simply delete it before hitting send [1]. It is a seemingly small tweak — but one that reveals a great deal about where AI-assisted communication is heading, and the tensions between automation and human agency that come with it.

The Feature: What Changed and Why It Matters

The new setting lives under Settings > Suggestions in Google Messages version 20260303_00_RC00, the latest beta build [2]. Users now see a toggle labeled "Tap to Send," which when turned off switches the behavior to "Tap to Draft." The default remains the legacy instant-send mode, meaning users must actively opt in to the new behavior [3].

The mechanics are straightforward. When a message arrives and Smart Reply generates chips like "Thanks!" or "On my way," tapping one no longer fires it into the conversation. Instead, the text populates the compose field. From there, users can modify the wording, append additional context, or simply tap the send button to confirm [4].

"This addresses two key frustrations: accidental message sends from unintended taps, and the inability to customize AI-generated responses before delivery," noted Android Authority in its analysis of the feature [5].

The upgrade was first spotted in development by 9to5Google in January 2026, when code references to a "Tap to Edit" setting surfaced in an earlier beta build [6]. Two months later, the feature has materialized in a functional form and is now progressively deploying to beta testers. Google has not announced a timeline for a stable-channel rollout.

A Long Time Coming: The History of Smart Reply Frustrations

Smart Reply is not new. Google introduced the concept in Gmail back in 2017 and later brought it to Google Messages as part of its push to make Android's default messaging app competitive with iMessage and third-party alternatives. The premise was simple: use machine learning to analyze incoming messages and suggest short, contextually appropriate responses that users could send with a single tap [7].

The problem was equally simple — that single tap was too easy. Online forums and social media have been littered with complaints from users who accidentally sent a breezy "Haha" to a serious message, or an enthusiastic "Yes!" to an invitation they meant to decline. The instant-send behavior, designed for speed, became a liability for anyone who valued precision over velocity.

The frustration was compounded by the fact that competitors had already solved this problem. Samsung's Galaxy S25 series, launched in early 2025, introduced AI-powered suggested replies that populate the compose field by default, requiring users to confirm before sending [8]. Apple's approach with iOS has similarly kept suggested text in the draft state. Google Messages was, in this regard, playing catch-up — a point several tech publications emphasized in their coverage.

"Google Messages adopts a smart feature from Samsung Messages and Apple Messages," headlined Tech Edu Byte, framing the update as Google finally matching its rivals' design philosophy [9].

The Bigger Picture: AI's Expanding Role in Google Messages

The Tap to Draft feature does not exist in isolation. It is part of a broader, accelerating effort by Google to infuse its messaging platform with artificial intelligence — primarily through its Gemini large language model.

Gemini arrived in Google Messages in 2024, appearing as a conversational AI accessible from within the app. Users with an Android device sporting at least 6 GB of RAM, a personal Google account, and RCS chats enabled can interact with Gemini directly inside their messaging interface — drafting messages, brainstorming ideas, planning events, or asking questions without leaving the app [10]. The feature is now available in over 165 countries.

Alongside Gemini, Google has rolled out a series of AI-powered features over the past two years:

• Magic Compose (2024): Allows users to rewrite drafted messages in different tones and styles, with on-device processing via Gemini Nano on Pixel 8 Pro and newer devices [11].
• Photomoji Reactions (2024): Uses on-device AI to transform photos into message reactions [11].
• Remix Photo Editing (2025): Brings AI-assisted photo editing directly into the Messages interface [12].
• Find Hub Integration (2026): Connects Google Messages with airlines and other services for real-world utility, like tracking lost luggage [13].

Smart Reply's Tap to Draft upgrade fits into this trajectory as a refinement of the human-AI interaction model. Rather than simply adding more AI capabilities, Google is also reconsidering how those capabilities interact with users — giving them more control rather than less.

The Scale: Google Messages and the RCS Boom

To understand why this feature matters beyond individual convenience, consider the scale at which Google Messages now operates.

Google announced that its Messages app surpassed 1 billion monthly active users on RCS by late 2023, up from 500 million just 18 months earlier [14]. In the United States alone, mobile phone users send more than 1 billion RCS messages per day [15]. The platform is not a niche product — it is the default communication layer for a significant portion of the world's smartphone users.

Source: Mordor Intelligence

Data as of Mar 9, 2026CSV

The broader RCS market is experiencing explosive growth. According to Mordor Intelligence, the RCS market was valued at $2.87 billion in 2025 and is projected to reach $10.93 billion by 2031, growing at a compound annual rate of 24.95% [15]. Juniper Research found that RCS business messaging traffic reached 50 billion messages globally in 2025, up from 33 billion the year prior — a 50% single-year surge [16].

Apple's adoption of RCS in iOS 18 in late 2024 was a watershed moment. According to Infobip, global RCS traffic spiked 500% following the iOS rollout, with regional penetration in the United States leaping from roughly 4-5% to 70% [17]. The implications are clear: every AI feature Google builds into its Messages app now reaches a far larger and more diverse audience than ever before.

Source: GDELT Project

Data as of Mar 9, 2026CSV

The Competition: An AI Arms Race in Messaging

Google is not the only company racing to embed AI into messaging. The competitive landscape has shifted dramatically in the past 18 months.

Samsung introduced AI-powered suggested replies with its Galaxy S25 series, integrating the feature at the keyboard level so it works across multiple messaging apps, including WhatsApp [8]. The Samsung approach notably made draft-before-send the default behavior — the very design choice Google is now adopting.

Apple Intelligence, Apple's suite of on-device AI tools, includes notification summaries and writing assistance features that operate across iMessage and third-party apps. Apple has marketed its approach as privacy-first, with notification summaries generated by on-device models that do not transmit message content to Apple's servers [18].

Meta has been aggressively integrating its own AI assistant into WhatsApp and Messenger, offering conversational AI, image generation, and message summarization. However, Meta's approach has drawn scrutiny because its language models do not run locally, meaning messages must be transmitted to Meta's servers for processing [18].

The result is a three-way design philosophy divergence: Apple emphasizes on-device privacy, Google offers a hybrid of cloud-based Gemini and on-device Gemini Nano processing, and Meta relies primarily on server-side AI. Each approach carries different implications for user privacy, feature capability, and the speed of innovation.

The Privacy Question: Who Reads Your Messages?

The integration of AI into messaging apps has raised significant privacy concerns that the Tap to Draft feature — while focused on user experience — does not directly address.

The Electronic Frontier Foundation published a detailed analysis in October 2025 warning that AI integration into messaging platforms creates new vectors for data exposure. When users employ AI features like Gemini to draft or summarize messages, the content may be transmitted to external servers for processing. "By default, everything you do in Gemini is stored in 'Gemini Apps Activity' where messages are stored forever and used to train Google's products," the EFF noted [18].

The concern extends beyond voluntary AI usage. Google's Utilities app can "read, summarize, and reply to notifications" including those from end-to-end encrypted messaging apps — yet Google's documentation provides limited clarity on what data is collected or retained from these notifications [18].

The ACLU has gone further, arguing that "secure messaging and AI don't mix" as a fundamental principle, since server-side AI processing inherently undermines the privacy guarantees of encrypted communication [19].

Google's Smart Reply feature specifically has operated in a somewhat ambiguous privacy space. The AI suggestions are generated based on analysis of incoming messages, but Google has maintained that this processing occurs on-device for Smart Reply (as opposed to the cloud-based Gemini features). The Tap to Draft change does not alter this architecture — it merely changes what happens after the suggestion is generated.

Still, the feature's emphasis on user control is philosophically aligned with growing public demand for AI transparency. A Stanford study published in October 2025 found significant privacy risks in AI chatbot conversations, adding academic weight to consumer concerns about how AI systems handle personal communications [20].

What Comes Next

The Tap to Draft toggle is currently available only in the Google Messages beta channel, with no confirmed date for a stable rollout. Given Google's typical release cadence, a wider deployment could come within weeks — though the company has been known to hold features in beta for extended periods.

The feature's significance extends beyond its immediate functionality. It represents a broader industry acknowledgment that AI automation in personal communication needs guardrails. Speed and convenience remain important, but users increasingly demand the right to review, edit, and approve what AI systems say on their behalf.

As RCS adoption accelerates, as Gemini becomes more deeply integrated into Android, and as the competitive pressure from Apple and Samsung intensifies, the way Google handles the human-AI boundary in messaging will have implications for billions of users worldwide.

For now, the ability to pause before sending a Smart Reply may seem like a modest improvement. But in a world where AI is increasingly putting words in our mouths, the right to a second look is anything but trivial.

This article is based on reporting from multiple technology publications and publicly available data. Google did not respond to requests for comment on the timeline for a stable-channel rollout of the Tap to Draft feature.