What Is CATI and Can AI Voice Interviews Replace It in 2026?

Last updated: 8 October 2026

Quick Answer: CATI, or computer-assisted telephone interviewing, is a phone survey method where a human interviewer reads an on-screen script while software handles routing. Partly: AI voice agents now run structured CATI questionnaires at scale, but a published 2,739-person deployment in Peru found its system probed less deeply than human interviewers.

CATI is being replaced not because telephones stopped working, but because of who stopped answering them. Pew Research Center's telephone response rates fell to 6% by 2018, from about 9% a few years earlier, and every unanswered call is a paid interviewer minute.

The CATI software market has split in two. One side still sells tools for human phone rooms: dialers, quota control, supervisor monitoring and, more recently, AI that audits recorded calls. The other sells AI voice agents that place the call and conduct the interview, and a team leaving a phone room buys from that side whether or not the product says CATI.

Which CATI Software Now Offers AI Voice Interviews?

AI voice interviewing now sits inside CATI software in different ways. Voiceter, Miravoice and Shrota (sold as GetAICATI) put an AI voice agent on the call itself. Voxco and IdSurvey keep human interviewers and add AI around them, and VeriCATI uses AI to audit recorded human calls. Alchemic runs AI-moderated phone interviews as managed fieldwork or on a client's own list.

CATI and AI Voice Vendors Compared

Vendors are listed alphabetically. Each description comes from the vendor's own site, checked on 8 October 2026.

Which One Fits What You Are Replacing

A team that wants to keep its interviewers and fix consistency needs a phone room tool such as Voxco or IdSurvey, perhaps with an audit layer like VeriCATI, not an AI agent. A Hindi-first tracker that must keep data in India is what Shrota is built for. Alchemic's AI phone research fits when the bottleneck is the whole fieldwork job, recruitment and languages as well as the calls.

How Does a Conversational AI Voice Survey Work?

An AI voice agent places the call, says it is automated, captures consent, then administers the questionnaire as a conversation. It handles branching, asks for clarification when an answer is thin, and transcribes and codes responses while fieldwork runs.

With no interviewer roster to schedule, calls run in parallel and the constraint moves from staffing to calling windows, consent and respondent availability.

In the United States, the FCC's Declaratory Ruling FCC 24-17, adopted 2 February 2024, confirms that AI-generated voices are "artificial" under the Telephone Consumer Protection Act, so the call needs the prior express consent of the person called. Recording consent is separate and set by state law, and the TCPA rules for AI phone research map the detail.

Gallup, reporting on 26 February 2026 on AI phone pilots across four continents and seven languages, flagged the same risk: in the United States it may not be legal to pair an autodialer with AI infrastructure or to take consent through an AI interviewer.

How Does Traditional CATI Software Run a Phone Survey?

A human interviewer works from an on-screen script while the software controls question order, skip logic, quotas and validation. CATI replaced paper telephone interviewing in the 1970s and 1980s and remains a recognized professional mode.

The defining feature is the split of labor. The computer enforces the instrument; the person conducts the conversation, handles clarification and judges when an answer is complete. CATI suites add the phone room around that split: automated, manual or hybrid dialing, call scheduling, quota dashboards, and supervisors who listen to live calls and track interviewer performance.

How CATI Differs From CAPI and CAWI

All three are computer-assisted modes, separated by how the respondent is reached. CATI is telephone with an interviewer, CAPI is in person with the interviewer carrying a device, and CAWI is web and self-administered. The guide to choosing between CATI, CAPI and CAWI covers which population each reaches.

What Is Replacing Human-Dialed CATI?

Five approaches, and they are not interchangeable. Most teams leaving human-dialed CATI split one study across two or three of them.

Why Human-Dialed CATI Lost Its Economics

Coverage collapsed while the method stayed the same. Caller ID, mobile substitution and spam filtering changed who answered, and Pew attributes part of its decline to an estimated 3.4 billion robocalls a month. A 6 percent response rate is not fatal on its own; it becomes fatal when the people who answer differ from those who do not. The mechanics are traced in why phone survey response rates collapsed.

When Traditional CATI Is Still the Right Answer

Human-dialed CATI is still the better choice in three cases, the same three Alchemic's own phone research page lists:

1. An extremely rare language outside the set an AI system supports.
2. A very small specialist sample, under 20, that needs a bespoke one-off moderator.
3. A topic where a respondent may need a person on the line, such as a sensitive medical interview that requires a live clinician.

Who Can an AI Voice Survey Reach That a Link Cannot?

Anyone with a working number, including people on feature phones, shared devices and unreliable connections. That is a wider frame than any browser method, and in several large markets it is the difference between a national sample and an urban one.

The ITU's Facts and Figures 2025 counts 2.2 billion people still offline, most of them in low and middle income countries, while mobile broadband coverage is nearly universal.

Which Languages Can a Phone Study Run In?

A phone study is only as wide as the languages it can be conducted in, and the respondents hardest to reach by link are the least likely to be comfortable in English. Alchemic publishes 60+ languages including Spanish, Hindi, Tamil and Telugu, with code-mixing handled mid-call, and runs managed fieldwork or bring-your-own-list studies across 14 markets that include the USA and the UK.

Where a study needs pictures as well as reach, mix modes: phone for the segments a link cannot reach, interviews inside WhatsApp or the browser for the rest.

What Does the Evidence on AI Voice Interviewing Show?

Early, encouraging and thin. Two published tests show AI voice agents can administer real questionnaires at scale; probing depth and response rates are the open questions.

• Peru and US, 2025. An LLM-based telephone survey system ran a 75-person US pilot and a 2,739-person deployment in Peru. It handled open and closed questions, clarifications and branching. Data quality approached human-led standards on structured items, while its probing for qualitative depth was more limited than human interviewers'.
• SSRS Opinion Panel, 2025. An AI interviewer on a US probability panel delivered a 30-minute, 123-question survey with exact wording and randomized answer order, and 86 percent of respondents rated the experience neutral or positive.

Mode effects do not disappear because the interviewer is synthetic. Pew's randomized mode-of-interview experiment, 3,003 panelists split between phone and web, found phone respondents more upbeat about their lives: 62 percent said they were very satisfied with their family life by phone, against 44 percent on the web. A synthetic interviewer is still a voice on the line, so test sensitive items against a self-administered baseline before trusting them.

Where CATI and AI Voice Interviews Both Fall Short

Neither replaces the other.

• No visual stimulus, in either mode. Concept boards, packaging and interfaces cannot be tested by voice.
• A call carries voice, never a face. Facial signal is absent in both modes. How well a system reads tone and hesitation varies by tool, and a pilot recording is the only way to check.
• Distress detection is unvalidated. No AI system has been validated as a distress detector on a call, so none should be relied on to notice it. Route foreseeable-distress topics to a person.
• Probing depth varies. In the Peru test, the system's probing was shallower than human interviewers'; how much latitude a given system has to leave the script varies by tool.
• Compliance is local and unforgiving. Calling windows, do-not-call registries and consent rules differ by country, and a mistake is a legal problem rather than a data-quality one.

Once it has the client's brief, Alchemic designs and tailors the discussion guide to handle these risks before fielding, rather than leaving that work to the buyer.

How Do You Choose Between CATI and an AI Voice Survey?

Start from who you need to reach, not from what the method costs, because coverage failures are invisible in a completion report. Three questions settle most cases:

1. Can your target segment be reached by a link at all? If a meaningful share cannot, a browser-only design has decided your findings before fielding starts.
2. Does the instrument need visual stimulus? If yes, voice is out for that section, whoever administers it.
3. Could a respondent become distressed? If yes, route that portion to a human, regardless of budget.

Everything after those three is economics: volume, languages and how much of the fieldwork you want to run yourself. The buyer's guide to AI phone interview platforms compares the vendors on those terms.