This tracker follows Amazon patents that let Alexa read what's on your screen, catch a skill collecting sensitive data, sense when you're about to speak, and explain its own answers. Together they show Amazon building an assistant that watches, judges, and acts instead of just answering.
14 filings · tracking since May 2026 · latest Nov 2026 · updates weekly
based on all tracked filings in this watchlist · refreshes every week
Amazon is filing patents around teaching its voice assistant to make smarter choices on its own, from picking the right tool for a job to adjusting how it talks to you mid-conversation.
The filings cluster most heavily around two areas: helping the assistant decide what to do and how to do it without you spelling everything out, and making sure it handles sensitive information and explains itself when asked.
a dated entry each week this watchlist moves · older entries stay archived
This week's filing focuses on helping Alexa choose the right image to go with a spoken answer. Amazon is exploring ways to make voice replies more visual.
This week's filing focuses on helping Alexa decide which of its own abilities to use when you ask it something. The idea is to make the assistant better at picking the right action for each request.
Both new filings show Alexa moving from answering questions to taking action: one would let it browse websites on your behalf, the other would track which topics you feel most strongly about during calls.
This week's filing shows Amazon exploring ways for Alexa to choose how it sounds and behaves before it responds to you. The focus is on giving the AI a way to pick its own tone or style based on the situation.
Amazon’s tracked filings by month · newest first
15 or more filings that month · 5 to 14 · under 5
the problems Amazon keeps filing on · each with its three newest filings · new filings join every week
A voice assistant works better when it understands context beyond the words spoken. These filings cover systems that track what is on your screen, sense when you are about to speak, and adjust tone mid-conversation.
Acting on a request means understanding it fully, not just the words used. These filings cover systems that convert requests into working software commands, help an AI explain its own choices, and guide an AI through complex real-world situations like difficult roads.
Voice assistants and robots need to understand the space around them to be useful. These filings cover systems that map objects in a room to find where a sound is coming from, and swap out skill sets on a robot based on what a person asks it to do.
The choice of personality is made automatically from what you asked and the situation around the request.
If you sound frustrated, replies get shorter and calmer on their own, with no settings to change.
Phone numbers, health details or card numbers are caught before the skill ever receives them.
It tells a page command like go back apart from a question the voice assistant should answer out loud.
every tracked filing, month by month · counts are USPTO pre-grant publications, one per publication number (an application, not a granted patent)
US 2026/0351250 A1
Scoring answer elements by relevance to the user's question helps Alexa decide which fact deserves visual prominence when multiple candidates compete for screen space.
US 2026/0331107 A1
Reducing noise in skill selection: a filtering stage narrows the full app catalog to relevant candidates before the main model picks which one to use, cutting both latency and mismatches.
US 2026/0312784 A1
Distinguishing voice commands meant for a website from questions for the assistant itself lets Alexa handle both without requiring explicit switching signals from the user.
US 2026/0311692 A1
Detecting emotional peaks in a caller's voice and linking them to specific topics lets Alexa surface what genuinely matters to users during conversations, sharpening its ability to act on priorities rather than just parse words.
US 2026/0289186 A1
Automated personality selection based on question type and context lets Alexa vary its tone per response rather than maintaining a fixed default, extending the system's ability to adapt behavior to different user needs and situations.
US 2026/0188312 A1
Getting Alexa's intent into actual system calls requires mapping fuzzy user requests to specific APIs and services. This filing describes the translation layer that converts "check my flight" into the right backend command.
US 2026/0175868 A1
A robotaxi stuck on an unusual road situation could query an AI chatbot instead of waiting for human intervention, compressing the decision cycle from minutes to seconds.
US 2026/0175417 A1
To physically act on decisions, Alexa's robot body needs to switch between specialized skills without losing basic competence, this filing shows how general movement and task-specific control can run simultaneously.
US 2026/0172743 A1
Predicting speech intent from acoustic cues alone lets Alexa skip the wake word entirely, moving the assistant from passive listening to reading conversational context without explicit commands.
US 2026/0155148 A1
Once Alexa starts acting on its own, you need to know why it acted. This patent sketches how the system could show its reasoning in plain language, turning the black box into something you can actually audit.
US 2026/0155136 A1
Detecting user frustration from conversation signals lets Alexa shift its response style without explicit commands, moving the assistant closer to reading emotional cues rather than just parsing words.
US 2026/0148739 A1
An Alexa that grasps screen context could execute commands like "add that to my cart" without requiring you to name the product, letting the assistant move from parsing speech to understanding intent.
US 2026/0141903 A1
Preventing Alexa skills from intercepting unintended data requires a filter that knows what each skill is supposed to hear. This patent describes a system that monitors data flowing to skills and blocks sensitive information they weren't authorized to collect.
US 2026/0141917 A1
Matching audio arrival angles against spatial maps of furniture and walls lets the system ignore bounced sound and lock onto the actual speaker's location, sharpening how Alexa pinpoints where commands originate.
This tracker follows Amazon patents that push Alexa past simple Q&A: tying commands to what's on your screen, finding the true source of a sound, sensing when you're about to speak, auditing skills for privacy leaks, and explaining its own decisions. It's a research signal, not a list of confirmed products.
One filing describes tying voice commands to on-screen content, but it is a patent application, not a shipped feature. It shows Amazon exploring how phrases like 'add that' could resolve against what's displayed, not a confirmed product timeline.
One patent in this batch is built to catch Alexa skills that collect sensitive data like phone numbers without clear need. It suggests Amazon is building internal auditing tools alongside new sensing features, likely because more cameras and context awareness raise more privacy questions.
Expect more filings on context awareness, like combining screen data with audio and speech timing, plus continued work on explainability and robot skill switching. The tracker updates as new Amazon filings publish, so check back rather than expecting a fixed list.
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →