Your voicebot talks to your customers all day. Know what it says.
Upload the knowledge base behind your company’s voicebot, and Voicemark phones it with dozens of generated calls, the way your customers do. You get a score you can report upward, the exact prompt or knowledge-base change for every wrong answer, and the recording behind every verdict. Rerun after every change to keep it that way.
Free trialFull plan, 4 weeks free, no credit card required.Sipfront runs Voice AI tests for teams like
Replace mystery calls with a repeatable check, in 5 minutes of your time.
Score, fix and proof for every run.
One score your management understands
Task completion, correct answers, turn-taking, how the bot sounds, reliability and compliance, scored per run and shown side by side. Report a trend instead of a feeling, and see straight away when a change made things worse.
A fix you can paste or forward
Each finding names where it belongs: system prompt, knowledge base or the platform’s transfer settings. Paste it yourself, or forward it to your integrator or vendor with the recording that proves it.
Does it sound like your brand?
Naturalness, synthetic share, words per minute, pauses, pitch variation and loudness, each compared with the caller on the same recording, plus reply delay. A flag says too fast or monotone in words, so the bot sounds like the company you want to be.
Evidence when someone asks
Audio, timed transcript and SIP trace for every call. AI disclosure and recording notice marked with the second they were spoken. When data protection, the works council or a complaining customer asks, the answer is one click away.
Score your company's voicebot this afternoon.
Upload the knowledge base, start the run and get on with your day while it calls. Full plan, 4 weeks free, no credit card required.
Keep the voicebot right after go-live, not just on launch day.
Prices change, opening hours change, campaigns start, the vendor updates the model. Each change is a chance for the bot to say something wrong to a customer.
Rerun after every change
New content, new prompt, new platform release: start the same test set again and compare the score with the last run. By hand, or from your own tooling over the API.
Right answers in your brand's voice
Scenarios come from your own knowledge base, so a wrong price, a missed handover to an agent or an off-brand tone shows up before your customers hear it.
A number to report
One score per run, broken down by dimension. Show the improvement after your changes, or the regression after the vendor’s, with the report attached.
Nine dimensions, scored from phone calls to your bot.
The bot needs a phone number, a SIP URI or a WebRTC endpoint, nothing else. No access to the bot platform, no integration project. fonio.ai, sipgate sona, CloudTalk, Parloa, NICE Cognigy and Zoom Virtual Agent are reached the same way a caller reaches them.
Calls over PSTN, SIP and WebRTC
Speech-to-speech callers in more than 50 languages, on the line your callers use. Jitter, loss and MOS from the same calls, so a carrier fault is not reported as a slow bot.
Content correctness
Scenarios generated from the knowledge base and judged per call against its facts. A calibration round aligns the judge with the bot’s wording before scoring.
Security probes
Ten caller archetypes mapped to the OWASP Top 10 for LLM Applications (2025), each reported as held, failed, or held only sometimes.
Robustness
Interruptions, silence mid-turn, language switches, circular callers and street noise. Barge-in, dead air and loops are counted on the recording.
Compliance checks
AI disclosure and recording notice with their timing, no leakage of verification data, language adherence and a turn budget. Transcript attached to each check.
API, webhooks and MCP
Start runs and read results from your own tooling, or from Claude, Cursor and other assistants over MCP. Results also arrive by email.
Voicemark checks what the vendor's dashboard can't.
The platform dashboard counts calls and containment. It doesn’t tell you whether the answers were right or how the bot sounded to your customer.
| Platform dashboard | Mystery calls by colleagues | Voicemark | |
|---|---|---|---|
| Calls the bot over the line your callers use | - | ✓ | ✓ |
| Checks answers against your knowledge base | - | Spot checks | ✓ |
| Repeatable after every change, with a score | - | - | ✓ |
| How the voice sounds, barge-in, reply delay | Reply delay only | By ear | ✓ |
| AI disclosure and recording notice, timed | - | - | ✓ |
| Cost | Included, but blind to content | Hours of staff time per round | €99 a month, flat |
€99 a month to test 3 voicebots simultaneously.
One flat monthly price with the test allowance included, so you know the cost before the first call. No quote to request, cancel monthly.
3 voicebots simultaneously
- Test allowance included in the monthly price
- 30 tests in 15 to 20 minutes, results by email and in the app
- Test callers in more than 50 languages
- Security probes, barge-in tests and the measurement of how the bot sounds
- A fix to paste for every finding, REST API and MCP server included
Talk to us when you outgrow it
If you run more than 3 voicebots, want the bot watched around the clock, or need to test the contact centre platform and phone lines behind it, ask us about combining Voicemark with Sipfront’s monitoring and testing products. Built and hosted in the EU by Sipfront in Vienna.
Article 50 of the EU AI Act has applied since 2 August 2026.
It is the transparency rule that covers every voicebot, whatever its risk class: callers must be told that they are talking to an AI system. Here is what that means for the company whose number the customer dialled.
What Article 50 requires
A caller must be informed that they are interacting with an AI system, early enough and clearly enough to notice, unless it is obvious from the context. The recording notice under GDPR and national telecom law comes on top, in every language the bot answers in.
Why it concerns you
The duty to build the disclosure in sits with the provider of the system, and Article 99(4)(g) sets fines of up to €15 million or 3% of total worldwide annual turnover. But when a caller complains, it’s your brand in the headline, and the recording notice under GDPR is yours as the controller. A rewritten greeting can drop either one.
Where Voicemark helps
Every run calls the bot like a customer and marks the AI disclosure and the recording notice to the second, with the recording, transcript and timestamps kept as a record. Whether that satisfies your obligations is a question for your legal advice, and this summary does not replace it. How Sipfront tests voice AI in production.
Questions voicebot owners ask before the first run.
Do I need technical skills or test cases?
No. Upload the knowledge base and prompt the bot works from, and enter the number customers call. Voicemark generates the scenarios, calibrates them against the live bot and shows them to you before the scored run. You can edit, add or remove any of them in plain language.
Do we need access to the bot platform or our vendor?
No. Voicemark calls the bot like a customer, so it works with any platform without an integration. If your integrator or vendor maintains the bot, forward them the findings with the recordings.
Will the test calls show up in our statistics and bill?
The test calls reach the bot like customer calls, so they count against the platform’s minute allowance and appear in its call statistics. A run of 30 tests at two to three minutes each uses one to one and a half hours of bot minutes. Use a separate test number if your platform offers one.
How often should we run it?
After every change to the prompt, the knowledge base or the platform, and at a fixed rhythm in between, for example weekly, to catch changes on the vendor’s side. Runs can be started by hand or over the API.
How do I know the judge is right?
A calibration round of three runs, done in under an hour, aligns the scenarios and the judging criteria with how your bot actually phrases its answers. Every verdict links to the transcript and the audio, so you can check it yourself in seconds.
Can I try it before I pay?
Yes. The free trial runs for four weeks on the full plan and needs no credit card. If you want to see a report before you register, ask us for a sample.
Hear what your customers hear, before they call.
Four weeks free, no credit card required. Your bot’s first score is in your inbox this afternoon.