G2 AI Voice Agent Platform Evaluations
Compare AI voice agent platforms on how well they resolve a caller’s request end to end, using the same simulated calls for every platform.
Agent Evaluation Score × G2 Market Presence Beta
See how each Voice AI agent platform performed in G2 evaluation, alongside the Market Presence of its G2 product.
Agent coverage · 6 evaluated agents
| Evaluated agent / slug | Market-presence product / slug | Market presence / source category | Mapping |
|---|---|---|---|
| Retell AIretell-ai | 70.4Voice AI Agents | Same product | |
| ElevenLabselevenlabsio | 96.0Voice AI Agents | Assumed match | |
| LiveKitlivekit | 5.7Voice AI Agents | Assumed match | |
| Pipecatpipecat | 10.3Voice AI Agents | Assumed match | |
| Vapi AI + Vapivapi-ai | 40.1Voice AI Agents | Assumed match | |
| Blandbland | 31.3Voice AI Agents | Same product |
Agent Evaluation Grid methodology
The Agent Evaluation Grid adapts G2’s existing Grid framework for Agent Evaluations. Unlike the standard G2 Grid, which compares products based on User Satisfaction and Market Presence, this Grid compares a platform’s Platform Score with the Market Presence of its corresponding G2 product.
In G2’s Voice AI Agent Platform evaluation, G2 worked with Bluejay, a company that specializes in testing voice AI systems, to evaluate six platforms using simulated phone calls across healthcare, legal, and customer support.
Each platform handled the same 90 tasks, repeated five times for a total of 450 scored calls per platform. The evaluation used common simulated callers, business tools, and task requirements to compare how well each platform completed the requested actions. Bluejay also ran separate tests with background noise and degraded phone connection, which contributed to the word accuracy measure.
Agent Evaluation results don’t affect a product’s G2 Score, Market Presence, or Standard Grid placement.
Platform Score
The Platform Score ranges from 0 to 100 and combines three measures:
- Task success (50 points): whether the platform completed the caller’s request, with the correct actions and the expected final state of the business’s records.
- Word accuracy (25 points): how accurately the platform transcribed the caller’s speech, including whether names, dates and numbers were captured correctly.
- Response speed (25 points): how quickly the platform replied after the caller finished speaking, measured on the recorded call.
The Platform Score measures how well each platform performs in the evaluation rather than how easy it is to build, launch, or operate an agent.
Results reflect English-language calls across three simulated industries and may not apply to other languages or use cases.
Selecting a platform in the Grid also shows its evaluation measurements.
Market Presence
Market Presence combines metrics from G2 reviews and third-party sources to indicate the market presence of a product and its vendor. Review volume is weighted for quality, recency, and source.
Each evaluated platform is mapped to the closest available G2 product. Click on a product in the Grid to see its Market Presence score and a link to its corresponding G2 product profile.
Market Presence is calculated separately from the Platform Score and is not a measure of platform performance or buyer satisfaction.
ElevenLabs Agents and LiveKit Agents: Market Presence reflects the broader ElevenLabs and LiveKit product listings on G2.
About these results
Agent Evaluations are currently in beta. Results are dated snapshots and may change as G2 refreshes evaluations and refines its methodology.