What Is the Smartest Home Assistant? Alexa, Siri, Google Compared

There is no single smartest home assistant, and none has an IQ. The only side-by-side test we found, Loup Ventures' 2019 test of 800 questions, had Google Assistant answering 93% correctly, Siri 83% and Alexa 80%, and nearly every question was understood. Assistants are changing fast: in August 2025 Google announced Gemini for Home as the replacement for Google Assistant, so a 2019 ranking does not describe today's products.

Below: what that test measured and the three kinds of "smart" inside a voice assistant. Then whether Alexa is smarter than Siri, why AI benchmarks cannot rank a speaker, and how to judge a "smartest assistant" claim.

What is the smartest home assistant?

The smartest home assistant depends on the job, and no source we found settles it for all jobs. The same goes for the smartest voice assistant, since the phrase hides three separate abilities. A voice assistant does three things in a row: it hears you, works out what you mean, and does something about it. "Smart" can mean any of the three, and a product can be strong at one and weak at another.

• Step 1, hearing: turning speech into words. In 2019 all three assistants did this almost perfectly
• Step 2, understanding and answering: working out what you mean and finding the right answer. This is where the 2019 scores differed
• Step 3, acting: setting the timer, dimming the lights, sending the message. Which assistant ranked first depended on the type of request

Is Alexa smarter than Siri? What the only side-by-side test found

Is Alexa smarter than Siri? Not in the one test we found: in 2019 Siri scored slightly higher than Alexa on answering, and Alexa was behind on commands. Loup Ventures asked each assistant the same 800 questions about finding nearby places, ordering goods, navigating and more, and CNBC reported the results on 16 August 2019. The firm called it an IQ test, but it reports a percentage correct, not a score on the IQ scale (mean 100, SD 15).

• Google Assistant: 93% answered correctly, and it understood all 800 questions
• Siri: 83% answered correctly, and it misunderstood 2 questions
• Alexa: 80% answered correctly, and it misunderstood 1 question

All three had improved since July 2018, when the scores were 86% for Google Assistant, 79% for Siri and 61% for Alexa. That is a rise of 7, 4 and 19 percentage points, our arithmetic on those figures. The test covered smartphone assistants, not speakers, so it compares the phones' built-in assistants and the Alexa app.

Why did the smartest assistant change from one category to the next?

Because each category tests a different step. Loup Ventures said Google outperformed on information questions and Siri handled commands best. In the commands category both Siri and Google Assistant "far outperformed" Alexa. The authors linked that to the way they are built into the phones' operating systems, while Alexa runs as a third-party app.

In the commerce category, which covers ordering goods, Google Assistant answered 92% of questions correctly against 71% for Alexa and 68% for Siri. The researchers added that they did not think many consumers were buying by voice. One overall percentage hides all of this, which is why a "smartest assistant" headline needs the category and the year.

Which AI assistant is smartest today?

Which AI assistant is smartest? We found no current independent test that answers it. The big change since 2019 is that assistants are moving onto large language models. Google said on 20 August 2025 that Gemini for Home "is built on the same powerful AI models" as the Gemini app. It would replace Google Assistant on existing speakers and displays, with early access starting in October 2025. Google also called it more capable, and no independent test backs that up. We did not check how far the rollout has gone.

We also could not verify from a primary source which model runs Alexa or Siri today, so we do not rank them.

Do not read a leaderboard as a speaker ranking. The Artificial Analysis Intelligence Index scores AI models on ten evaluations. As of 14 October 2026 its top model scored about 58 out of 100, and the index had been rebuilt five times in two months. Our intelligence index explainer and the review of the most intelligent AI model show how those scores move. Why benchmarks disagree explains why two labs report different numbers for one model. None of them tests a smart speaker as a product.

Are smart gadgets actually smart?

When people search for the smartest gadgets, the word "smart" on a product usually tells you the device connects to a network and takes commands, not how well it reasons. A smart plug that obeys "off" is following a fixed instruction, which says little about reasoning. In the 2019 test the assistants differed on answering and on commands, not on hearing the question.

The gap with people is still wide where the task is new. On ARC-AGI-3, an interactive benchmark published in March 2026, the ARC Prize Foundation reports that humans score 100% and frontier AI 0.51%. Read our write-up, Can AI pass an IQ test?, for what that does and does not show. For the wider debate, see AI versus humans and the ChatGPT IQ score page.

How can you judge a "smartest assistant" claim?

• Look for the year: a 2019 result says little about assistants rebuilt since 2025
• Look for the category: information, commands and shopping gave different winners
• Look for what was scored: percentage correct on 800 questions is not an IQ
• Ask whether it tested the product you own: phone assistant, speaker or app can differ
• Distrust model leaderboards as product rankings: they score models, not speakers

Our wider coverage sits in the AI and intelligence news section. Human reasoning is measured against a norm group instead: for that, take the IQ test and read the result with the IQ bell curve.

Common questions