---
type: "post"
slug: "how-to-evaluate-live-and-voice-agents-in-adk"
title: "How to evaluate live and voice agents in ADK"
description: "Google brought live evaluation into ADK: a simulated user speaks its turns as audio so spoken replies can be scored inside the same eval loop already used for text agents."
retrieval_nugget: "Google brought live evaluation into ADK: a simulated user speaks its turns as audio so spoken replies can be scored inside the same eval loop already used for text agents."
published_at: "2026-09-01"
updated_at: "2026-09-12"
record_date: "2026-08-25"
date_kind: "discovered_at"
topics: ["agent-security","agents","ai","evaluation","voice-agents"]
entities: ["Google"]
source_url: "https://developers.googleblog.com/how-to-evaluate-live-voice-agents-in-adk"
source_title: "How to Evaluate Live & Voice Agents in ADK- Google Developers Blog"
source_domain: "developers.googleblog.com"
source_terms: ["evaluate","live","voice","model","import"]
summary_word_count: 199
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:how-to-evaluate-live-and-voice-agents-in-adk"
status: "published"
source_urls: ["https://developers.googleblog.com/how-to-evaluate-live-voice-agents-in-adk"]
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:586e19c5450f2f94bdd26a633e32cd5793780038e3bab5d82aee94634eddee24","reviewed_at":"2026-09-12T10:00:00Z","source_evidence_count":1,"verified_claim_count":2,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":2,"implication_count":1,"watch_condition_count":1,"related_record_count":3}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Google added native live evaluation to the Agent Development Kit on 24 August.","observed_facts":[{"text":"ADK can now drive a live voice agent with a simulated user that speaks its turns as audio and score the spoken replies.","source_urls":["https://developers.googleblog.com/how-to-evaluate-live-voice-agents-in-adk"]},{"text":"The post names tools failing to fire, context slipping between turns and ignored interjections as live-agent failure modes.","source_urls":["https://developers.googleblog.com/how-to-evaluate-live-voice-agents-in-adk"]}],"mechanism":"A simulated user drives a live voice agent by speaking its turns as audio, the spoken replies are scored, and the whole thing runs inside the same evaluation loop teams already use for text agents rather than in a separate voice harness.","why_now":"behaviour that sounded perfect yesterday can change quietly on the next prompt tweak or model iteration: tools stop firing, context slips between turns, interjections go ignored","implications":["None of those are audible in a single-turn demo, which is why a demo is not evidence."],"evidence_boundary":"This is a how-to post with a worked example: it documents the loop without publishing scoring rubrics or reliability data for the simulated user itself.","watch_conditions":["The condition to watch is whether simulated-user audio becomes reproducible across runs, because an eval whose synthetic caller varies between executions moves the noise rather than removing it."],"related_records":[{"url":"https://newruntime.com/posts/langchain-agent-environments-spec-to-task-pipeline","relation":"Prior coverage of turning agent evals into a production pipeline."},{"url":"https://newruntime.com/posts/evals-move-from-leaderboards-to-operating-diagnostics","relation":"Prior coverage of evaluation as an operating diagnostic."},{"url":"https://newruntime.com/topics/agents","relation":"Topic hub for agent evaluation practice."}]}
routes: {"html":"https://newruntime.com/posts/how-to-evaluate-live-and-voice-agents-in-adk/","markdown":"https://newruntime.com/posts/how-to-evaluate-live-and-voice-agents-in-adk.md","json":"https://newruntime.com/posts/how-to-evaluate-live-and-voice-agents-in-adk.json"}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
---

# How to evaluate live and voice agents in ADK

## Retrieval answer

Google brought live evaluation into ADK: a simulated user speaks its turns as audio so spoken replies can be scored inside the same eval loop already used for text agents.

Google added native live evaluation to the Agent Development Kit on 24 August. A simulated user drives a live voice agent by speaking its turns as audio, the spoken replies are scored, and the whole thing runs inside the same evaluation loop teams already use for text agents rather than in a separate voice harness.

The failure modes named in the post are what makes this more than a convenience. A live agent has to take the right actions across a spoken conversation turn after turn, where timing and recovery matter as much as content, and behaviour that sounded perfect yesterday can change quietly on the next prompt tweak or model iteration: tools stop firing, context slips between turns, interjections go ignored. None of those are audible in a single-turn demo, which is why a demo is not evidence.

Turning the conversation into a repeatable case is the same production move as [LangChain turning agent evals into a spec-to-task pipeline](https://newruntime.com/posts/langchain-agent-environments-spec-to-task-pipeline/), and it is the direction described in [evals moving from leaderboards to operating diagnostics](https://newruntime.com/posts/evals-move-from-leaderboards-to-operating-diagnostics/). The [agents hub](https://newruntime.com/topics/agents/) collects both.

This is a how-to post with a worked example: it documents the loop without publishing scoring rubrics or reliability data for the simulated user itself. The condition to watch is whether simulated-user audio becomes reproducible across runs, because an eval whose synthetic caller varies between executions moves the noise rather than removing it.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Agent Security](https://newruntime.com/topics/agent-security.md): Explore the agent-security topic hub.
- topic hub: [Agents](https://newruntime.com/topics/agents.md): Explore the agents topic hub.
- topic hub: [Ai](https://newruntime.com/topics/ai.md): Explore the ai topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)