Alexa+ Review: Is Amazon's AI Upgrade Worth It?

Alexa+ is Amazon's generative-AI rebuild of its voice assistant: a substantial upgrade rather than a small one, free to Prime members, and a clear answer to the question of how a voice assistant first launched in 2014 stays relevant in the era of ChatGPT and Gemini. The honest question is whether it's worth turning on (it is) and whether it's worth buying new hardware for (mostly not). This is a plain-English explainer of what Alexa+ actually adds, what's still missing, and how to decide.

Old Alexa was a command parser. You said 'Alexa, set a timer for ten minutes' and it matched the phrase against a predefined intent ('SetTimer'), pulled out the duration, and ran it. The strength of the system was speed and reliability for the small set of things it understood. The weakness was everything else: anything outside the intent library bounced back as 'sorry, I don't know that one'.

Alexa+ keeps the fast, reliable command path for simple things (timers, lights, alarms) and adds a generative layer that handles natural conversation, multi-step requests, and tasks that don't map to a single command. Three concrete shifts:

1. Conversational continuity. You can have a follow-up exchange — 'What's the weather tomorrow?', 'Should I take a coat?', 'And on Saturday?' — without restarting each time. Old Alexa lost context after one turn.
2. Agentic actions. Alexa+ can complete tasks that involve a series of steps and external services: book an Uber, order from Grubhub or Just Eat, find a plumber via Thumbtack, call a restaurant via OpenTable. The implementation is partnership-driven (each integration is built one by one) and the partner list is growing.
3. Generative routines. Instead of building routines through the app, you describe them in natural language. 'Every weekday at 7am turn on the kitchen lights, play BBC Radio 4, and tell me the day's calendar.' Alexa+ assembles the routine from your description.

Amazon has been clear that Alexa+ is built on multiple large language models running in the cloud — including Amazon's own Nova family and Anthropic's Claude. The system uses a router that picks the right model for each request, which is why responses can feel different depending on whether you're asking about a recipe, a smart-home command, or a complex agentic task. None of the LLMs run on the Echo device itself: the speaker is a microphone-and-speaker endpoint that streams audio to the cloud and plays back the response. That has implications for privacy and latency that we'll come back to.

Alexa+ is included with Amazon Prime at no extra charge. Without Prime, the price is $19.99 a month in the US. UK pricing has tracked the Prime model — included for Prime members, with non-Prime pricing announced regionally. Always check Amazon's current pricing page before subscribing because Amazon has revised the structure twice since launch.

The pricing matters because it changes the calculus: if you already pay for Prime for shipping and Prime Video, Alexa+ is effectively free. If you don't, $20 a month is roughly the price of a ChatGPT Plus subscription, and you should evaluate it against ChatGPT, Claude or Gemini rather than against the old free Alexa.

Alexa+ runs on most Echo devices made in the last five years or so, but the experience varies. The full visual interface — proactive widgets, recipe cards, agentic confirmation screens — is only available on Echo Show models (Show 8, Show 10, Show 15, Show 21). Audio-only Echos still get the conversational improvements but lose the visual layer. A short summary:

If you have an older Echo and you want Alexa+, you'll probably need to upgrade to a 4th-gen or later device. The good news: the entry-level Echo Dot at retail is the cheapest path in. The bad news: the visual experience really does add value on a Show, and if you're upgrading you should probably go to a Show rather than another speaker.

Three things stand out across reporting and user feedback aggregated across the first year of release.

Following multi-step instructions. 'Find a Thai restaurant near me that's open now and has a free table for two at 7pm' is the kind of compound query that old Alexa would refuse. Alexa+ chains the steps via OpenTable and replies with a usable answer or asks for clarification. Same for shopping: 'Reorder the laundry detergent I bought last month, but only if it's under £15' works.

Drafting and editing. Asking Alexa+ to draft a short email, summarise a recipe, or rewrite a paragraph in a different tone produces output broadly comparable to a generic ChatGPT response. It's not as deep as ChatGPT-4 or Claude on extended writing tasks, but it's right there in the kitchen when you need it, which is the point.

Smart-home routines. 'Run the morning routine but skip the radio because I'm on a call' is a request that requires understanding a named routine, modifying it on the fly, and executing the modified version. Alexa+ handles this category of request markedly better than old Alexa, which would force you to define a new routine in the app.

The platform has limits worth flagging before you commit to it.

Latency on complex requests. Simple commands are still fast. Agentic requests that involve a partner integration (Uber, OpenTable) can take 5–10 seconds to confirm, which feels long when you're standing in the kitchen waiting. The Echo Show models help by showing a 'thinking' state on screen.

Hallucination on factual recall. Like any LLM, Alexa+ can confidently fabricate facts — particularly on local information (which restaurant is open, what time the local pool closes). It also occasionally fabricates calendar events. Cross-check anything important.

Smart-home reliability. The new conversational layer occasionally over-thinks simple commands. 'Turn off the lights' has been reported to occasionally produce a follow-up question ('which lights?') even when there's only one light group active. The classic Alexa command path is more deterministic for the basics.

UK and non-US partner integrations are uneven. The agentic action library is broader in the US than in the UK or Europe. Uber, OpenTable and Just Eat are present in most regions; other US-only integrations (Thumbtack, Vagaro) won't work outside the US.

The old privacy story for Alexa was already complicated — voice recordings sent to the cloud, with optional retention. Alexa+ doesn't change the underlying architecture (everything still runs in the cloud) but it does change what the cloud is doing with your audio: an LLM is reading your transcripts and producing a generative response that may also pull from your calendar, contacts, and shopping history.

Amazon's published policy is that Alexa+ does not use your voice or text data to train its public LLMs. Personal data is used to personalise responses (so the assistant knows your routine, your preferred news source, your saved addresses) but is held in your account, not pooled into training corpora. You can review and delete voice recordings under Alexa Privacy Settings, and you can disable specific data uses (recordings retention, ad targeting) granularly.

For users who want fully local processing of voice and AI — no cloud component — the answer is still Home Assistant with a local voice integration like Whisper plus a local LLM. That's a different shape of project and a different time investment, but it's the only path to keeping voice processing entirely on-premises today.

The right comparison is not 'Alexa+ vs old Alexa' (it's clearly better) but 'Alexa+ vs other AI assistants you might already pay for'. Quick framing:

ChatGPT, Claude, and Gemini are general-purpose assistants accessed mostly through phone and web apps. They're stronger on extended writing, document analysis, and complex reasoning. They're weaker at controlling smart-home devices and at being instantly available across the house without unlocking your phone.

Alexa+ is an ambient-first assistant. The unique value is voice access throughout the home, integration with smart-home devices, and shopping/ordering tasks via Amazon's partner network. It's not the best AI for writing a long email; it is the best for 'turn off the kitchen lights and book a restaurant for 7pm' said while you're carrying laundry.

If you want both kinds of capability you can have them — they don't conflict and there's no exclusivity. For a deeper comparison of the chat-first AIs, see our ChatGPT vs Claude vs Gemini guide.

Three honest scenarios:

You already have an Echo and Prime. Turn on Alexa+. There's nothing to lose, the headline improvements (conversation, multi-step, generative routines) are genuinely better, and you're already paying for it via Prime. Worst case it's no different to old Alexa for the basics.

You have an old Echo (1st–3rd gen). The honest answer is that the assistant has moved beyond your hardware. If you use voice assistant features daily, an Echo Show 8 or a 5th-gen Dot is a clear upgrade. If you mostly use Alexa for timers and music, you don't need to spend anything.

You don't have an Echo. Don't buy one purely for Alexa+. Try the assistant via the Alexa app first (which works on a phone with the new conversational features). If you find yourself wishing the assistant was always-listening in the kitchen, then buy a Show or Dot. If not, save the £80 to £200 for something you'll actually use.