Don't rely on a language list. Ask for results for each language measured on phone audio, then have your own staff call it. A system that greets callers in Mandarin isn't necessarily one that understands a Mandarin-speaking patient asking for Dr. Chen next Wednesday after three, and then books it correctly. These ten checks separate the two, whichever vendor you're looking at.
The 10-point checklist
| Check | What a good answer looks like | ZeroMiss |
|---|---|---|
| 1. Results for each language | A separate number for each language, not one for "all languages" | Published for Mandarin and Persian |
| 2. Measured on phone audio | Phone-quality recordings, not studio speech | 8 kHz phone-quality audio, including real Mandarin phone calls |
| 3. How it was tested | Where the recordings came from, what kind of speech, who ran the test | Public research recordings plus clinic booking lines; our own benchmark, not independent |
| 4. Switching languages mid-call | It follows the caller; English names inside a Mandarin sentence are understood | Each reply follows the caller's last turn |
| 5. A native voice | Speaks the language itself, not English translated word for word | A native voice per language |
| 6. Times, names and numbers | Times said the way a native speaker says them; phone numbers read back; practitioner names in both languages | Yes, for Mandarin and Persian |
| 7. After a misheard word | A read-back, and no booking without the patient confirming | Read-back, then the patient taps Confirm on screen |
| 8. Texts and chat in the same language | The follow-up text, and the chat if there is one, match the call | Mandarin: texts and chat in Chinese. Persian: texts in Persian, chat screens in English |
| 9. A language it doesn't support | A straight answer, not "yes" to everything | Cantonese and French: not available yet |
| 10. Your staff, your test line | A test line with your names and services, before any patient calls | Every clinic starts in test |
Five test calls to make yourself
Ask a staff member who speaks the language natively, ideally two with different accents, to call the test line and:
- Book with a named practitioner on a specific day and time, speaking naturally and quickly. Did it get the name, the day and the time right?
- Switch languages halfway. Start in Mandarin, finish in English. Did the replies follow?
- Change their mind. "Actually, Thursday is better." Did it search again?
- Give a phone number. Was it read back correctly, digit by digit, in their language?
- Ask something it shouldn't answer, like whether a symptom is serious. Did it hand over to a person, in their language?
Mandarin isn't Cantonese. If many of your patients speak Cantonese, test with a Cantonese speaker. If the system answers in Mandarin, it doesn't support Cantonese, whatever the website says.
Red flags
- "40+ languages" with no result for any one of them.
- One accuracy figure covering every language.
- Demo calls that are short, scripted, recorded in a quiet room or cut together.
- Mandarin and Cantonese listed as one "Chinese".
- A voice whose word order sounds like English.
- No clear answer to what happens when a name or time is misheard.
Frequently Asked Questions
What accuracy is good enough?
There's no official standard. More useful than one number is what the system does with a mistake: whether it reads back the time and name, and whether the patient confirms before anything is booked.
Should results be independent?
Independent results are better. Most vendors, ZeroMiss included, publish their own benchmarks, so your own staff's test calls matter most.
Does language support change privacy?
It can. Some languages use different speech providers, so ask where each language's audio is processed and whether it's kept.
Bottom Line
A language claim is only as good as its test. Ask for per-language results on phone audio, listen to full calls, and have your own staff try to trip it up before your patients do.
