Google Assistant
Last reviewed
Sources
28 citations
Review status
Source-backed
Revision
v2 ยท 2,561 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
28 citations
Review status
Source-backed
Revision
v2 ยท 2,561 words
Add missing citations, update stale details, or suggest a clearer explanation.
Google Assistant is a voice assistant developed by Google that responds to voice and text commands on phones, smart speakers, smart displays, televisions, cars, and wearables. Unveiled at the Google I/O developer conference on May 18, 2016, it shipped later that year inside the Google Allo messaging app and the Google Home smart speaker, and grew into one of the most widely distributed consumer AI products of its era: Google said in January 2019 that Assistant would be available on nearly one billion devices by the end of that month, and in January 2020 that it had more than 500 million monthly active users [1][2][3][4]. Beginning in 2024 and 2025, Google replaced it with Gemini, its large language model based assistant, on phones, and with Gemini for Home on smart-home devices, making the transition one of the largest assistant migrations in consumer software history.
Built as the conversational successor to Google Now, Assistant was Google's answer to Amazon's Alexa and Apple's Siri during the voice assistant boom of the late 2010s. It parsed spoken queries with a pipeline of speech recognition and natural language processing against a largely fixed set of intents, the dominant design for conversational AI before large language models reset expectations. After ChatGPT shifted attention to generative chatbots, Google moved to retire it. The company pruned features through 2023 and 2024, announced on March 14, 2025 that the "classic" Assistant would leave most mobile devices in favor of Gemini, and began replacing it on speakers and displays with Gemini for Home in late 2025 [5][6]. As of June 2026 the mobile transition, originally targeted for completion in 2025, remains in progress after Google pushed the timeline into 2026 [7].
Sundar Pichai introduced Google Assistant during his I/O 2016 keynote, describing it as "a conversational assistant" capable of "an ongoing two-way dialogue" and framing it as the evolution of Google search into "an ambient experience that extends across devices" [1][2]. Where Google Now returned one-shot answers and predictive cards, Assistant was built to understand follow-up questions in context and to complete multi-step tasks such as booking movie tickets [1].
The product debuted in September 2016 as a text chat agent inside Allo, Google's short-lived messaging app [8]. In October 2016, Google announced the Pixel smartphone line with Assistant built in as a temporary exclusive, alongside the $129 Google Home speaker, which shipped on November 4, 2016 as a direct competitor to the Amazon Echo [9][10].
Distribution then widened quickly. Assistant rolled out to existing Android phones running Marshmallow and Nougat beginning in February 2017, gained a software development kit for third-party hardware in April 2017, arrived on iOS as a standalone app in May 2017, and extended to Android 5.0 devices and tablets in December 2017 [8]. The first third-party smart displays were announced at CES in January 2018 and shipped that July, followed by Google's own Home Hub (later Nest Hub) in October 2018 [8].
At CES in January 2019, Google said Assistant would be available on nearly one billion devices by the end of the month, up from 500 million the previous May, with the total dominated by Android phones; the number of connected smart home devices had grown 600 percent year over year [3]. A year later, at CES in January 2020, Google disclosed for the first time that Assistant had more than 500 million monthly active users across more than 90 countries and over 30 languages [4][8].
Assistant was invoked with the "Hey Google" or "OK Google" hotwords, or by gestures and dedicated buttons on supported hardware. It answered general knowledge questions using Google search, managed timers, alarms, reminders, and calendars, controlled media playback, placed calls and sent messages, ran multi-step Routines, recognized individual users by voice for personalized results (Voice Match), and controlled a broad smart home ecosystem of lights, thermostats, locks, and cameras. Interpreter Mode, which provided live spoken translation between languages, reached the smartphone apps on December 12, 2019 [8]. Third-party developers could build conversational apps and games for speakers and displays through the Actions on Google platform [13].
The product's most ambitious extension was Google Duplex, demonstrated at I/O in May 2018: an Assistant feature that placed real phone calls with a synthesized voice, complete with human-sounding filler words, to book restaurant tables and salon appointments [11]. Pichai presented Duplex as technology that "brings together all our investments over the years in natural language understanding, deep learning, text to speech" [25]. The demo was so lifelike that some outlets, including Axios, questioned whether it was real, and critics pressed Google on disclosure; the company committed that Duplex would identify itself at the start of each call [11]. By late 2018, Duplex was completing real reservation calls in limited release, opening with a statement that Google was calling and that the call would be recorded [12].
Google Assistant was a pre-LLM conversational system: it turned speech into structured intents using deep neural networks rather than a single generative model. Google described the flow as two stages. First, speech recognition would "receive an input, like the audio of someone speaking, and process that information across a stack of layers to turn it into text." Then natural language processing would take over, where "the text is processed by another stack of layers to parse it into pieces of information that help the Assistant understand what you need" [26]. A query such as "where's the closest dog park?" was parsed "with another neural network that tries to identify the semantics, i.e. the meaning, of your question," sorting candidate answers by how confident the system was and how relevant each response seemed [26].
To make multi-turn dialogue feel natural, Google layered on contextual rephrasing, combining linguistic rules, large amounts of historical query logs, and machine-learning models based on Transformer architectures so the Assistant could resolve a follow-up like "When?" into "When did William Shakespeare write Romeo and Juliet?" [8]. Despite these advances, the core design still matched utterances against a largely fixed library of intents and actions, the standard pattern for the voice-assistant generation. That structural limit is what Google ultimately judged too costly to retrofit, opting to replace Assistant with a large language model rather than extend it [5].
The launch of ChatGPT in November 2022 shifted industry attention from command-driven voice assistants to generative chatbots, and Google's investment visibly moved with it. The first major retrenchment came on June 13, 2023, when Google sunset Conversational Actions, ending third-party voice apps and games on Assistant devices; integrations that depended on the platform, such as MyQ garage door openers and Alarm.com devices, stopped working with Assistant that day [13][14].
In January 2024, Google laid off hundreds of employees across the Assistant, hardware, and central engineering organizations as it consolidated Pixel, Nest, and Fitbit into a single hardware team; the Alphabet Workers Union put the combined figure at more than 1,000 [15]. Days later, on January 11, 2024, Google announced it was removing 17 "underutilized" Assistant features, including voice control of Google Play Books audiobooks, cookbook and recipe tools, stopwatches on speakers and displays, and the ability to send emails, make payments, or post to social media by voice [16][17]. Users began seeing deprecation notices on January 26, 2024, and most of the features were removed on February 26, 2024; at the same time, the microphone icon in the Google app was repurposed to return only search results rather than perform Assistant actions [16][17]. The cuts were widely read as Google triaging Assistant while it concentrated resources on its Gemini models [15][17].
Google signaled Assistant's successor in October 2023, when it previewed "Assistant with Bard," a hybrid of the assistant and its Bard chatbot, at the Pixel 8 launch event; the product never shipped broadly under that name [8]. On February 8, 2024, Bard was rebranded Gemini and a Gemini app launched on Android in the United States, letting users opt in to replace Assistant as their phone's default helper while keeping the "Hey Google" invocation; on iOS, Gemini was folded into the Google app [18][19]. Early versions lacked full parity with Assistant, and Google gradually restored utility features such as timers, media controls, and smart home commands [19].
The formal sunset plan arrived on March 14, 2025, when Google announced it was "upgrading more users on mobile devices from Google Assistant to Gemini" and that later in 2025 "the classic Google Assistant will no longer be accessible on most mobile devices or available for new downloads on mobile app stores" [5][20]. The same post said Gemini would expand to tablets, cars, headphones, and watches, that a "new experience, powered by Gemini" would come to home devices such as speakers, displays, and TVs, and that devices unable to meet minimum requirements would keep the classic Assistant [5].
The home half of that plan became Gemini for Home, announced on October 1, 2025. Google said the new assistant would support every Google Home and Nest speaker and display it had released, back to the original 2016 Google Home, and teased a new Gemini-first Google Home Speaker for spring 2026 [6][21]. Gemini for Home is built on the same large language models as the mobile Gemini app, and Google framed the shift as moving past rigid syntax: "we're replacing rigid commands so you can use more nuanced or complex requests," enabling reasoning-based instructions such as "turn off the lights everywhere except my bedroom" [27]. The change reaches an unusually large installed base: Google said it has "over 800 million devices" in the Works with Google Home ecosystem, counting its own hardware and third-party products [28]. Early access began rolling out in the United States in late October 2025, initially for Google Home Premium subscribers, and opened to broad opt-in by December 10, 2025, by which point Google said "millions of households" had access [22]. Once a compatible device is upgraded, it uses Gemini for Home permanently in place of Google Assistant [23].
The phone timeline slipped. On December 19, 2025, Google said it was adjusting its schedule "to deliver a seamless transition" and that the work of upgrading Assistant users to Gemini on mobile devices would continue into 2026; when the transition completes, Assistant will no longer be available on Android phones and tablets and the standalone iOS app will be discontinued [7][24]. As of June 2026, Google had not announced a single firm shutdown date for the remaining surfaces, and classic Assistant continues to run on hardware that cannot meet Gemini's minimum requirements [5][7].
| Date | Transition milestone |
|---|---|
| October 2023 | "Assistant with Bard" previewed at the Pixel 8 event; never widely released [8] |
| February 8, 2024 | Gemini app launches on Android in the US as an opt-in Assistant replacement [19] |
| March 14, 2025 | Google announces classic Assistant will leave most mobile devices "later this year" [5] |
| October 1, 2025 | Gemini for Home announced for all Home and Nest speakers and displays [6] |
| Late October 2025 | Gemini for Home early access begins in the US [22] |
| December 19, 2025 | Mobile transition delayed into 2026 [7] |
| June 2026 | Transition ongoing; no final shutdown date announced [7] |
Google Assistant anchored Google's ambient computing strategy for nearly a decade, helped create the smart display category, and built the device ecosystem that Gemini for Home now inherits. Its scale means its retirement is one of the largest assistant migrations in consumer software history, affecting hundreds of millions of users and a decade of Home and Nest hardware [3][21]. Duplex, its most celebrated feature, was a landmark of pre-LLM conversational AI and an early flashpoint in debates over AI disclosure, anticipating norms later applied to voice agents broadly [11][12].
The product's arc also illustrates the platform shift that large language models forced on the industry: Assistant parsed commands against largely fixed intents, and Google ultimately judged replacement, rather than retrofit, to be the path forward, a conclusion mirrored by Amazon's LLM-based Alexa+ overhaul. Elements of Assistant persist through the transition: the "Hey Google" hotword carried over to Gemini, Gemini for Home runs on speakers dating to 2016, and the classic Assistant remains the voice layer on legacy and low-spec devices that cannot run its successor [5][19][21].