Phone Survey Companies in the US and AI Phone Interviews 2026

Last updated: 23 September 2026

Quick Answer: Phone survey companies and AI phone interview alternatives include Alchemic, Cora Intelligence, Drive Research, I/H/R, Ironwood, Koji, Miravoice, Precision Opinion and SSRS. Five dial with human interviewers; four place AI calls. Pick a CATI house for cold cell phone samples and probability estimates, and AI calling for consented lists, many languages and fast turnaround.

Almost every American adult can be reached by phone, and most ignore calls from numbers they do not know. In late 2024 only 0.5% of US adults lived in a household with no phone at all, according to the CDC's National Health Interview Survey wireless substitution estimates. Yet in a 2020 survey of 10,211 adults, Pew Research Center found only 19% generally answer a cellphone call from an unknown number.

Coverage is not the phone's problem; cooperation is, and US law decides who may place the call. An AI voice counts as an artificial voice under the Telephone Consumer Protection Act, so an AI call to a cell phone needs prior consent. Cold random-digit dialing of cell numbers stays a human-dialed job, while a consented customer list is where AI calling earns its place. Choose by the sample you hold, then by people or software.

Key Takeaways

• Two kinds of company answer this query: CATI houses with trained interviewers, and AI phone services that read a fixed script or probe.
• The sample decides more than the technology. Cold cell samples need human dialing in the US; consented lists open the AI route.
• Coverage is near universal, pickup is not. Only 0.5% of US adults live in a phoneless household; about one in five routinely answers unknown numbers.
• Capacity is published unevenly, from 52 to 600 seats, or not at all.
• Languages separate the field. CATI houses that publish languages center on English and Spanish; AI services that publish a count list 14 to 57+.

How This Guide Evaluates Phone Survey Companies

Each company was read on its own website on 23 September 2026; none was briefed or paid. The list started from the companies ChatGPT named for this query on a US connection that day, plus firms ranking in US search.

Five criteria shaped each entry:

1. Who places the call: a human interviewer, an AI reading the questionnaire verbatim, or an AI that probes.
2. Published scale: seats, stations or concurrent calls.
3. Published languages.
4. Sample fit: random-digit dialing, a probability panel, or your own list.
5. Service model and compliance posture: full service, data collection only or self-serve, plus any published TCPA practice or security certification.

Three groups were left out:

1. QuestionPunk, which the same ChatGPT answer credited, runs voice interviews in the browser, not over a call.
2. VeriCATI checks CATI recordings against data files rather than running studies.
3. Voxco, Askia and NIPO sell CATI software; NIPO says its product runs on 10,000 seats worldwide.

The Quirk's and GreenBook directories list more telephone interviewing firms.

What a Phone Survey Company Actually Provides

A full-service firm designs the questionnaire, sources the sample, dials, records calls, codes open ends and reports. A data collection house stops at a clean data file. An AI phone service replaces the dialing and interviewing.

Phone Survey Companies at a Glance

Sorted alphabetically by company name, from each company's own pages on 23 September 2026. The numbered list below puts Alchemic, the publisher, first and the rest alphabetically.

Seeing an AI Phone Study Before You Buy

The AI phone research service page walks through a full AI phone study, from consent to live themes, including where it does not fit.

Which Companies Run Telephone Interviewing in the US

Five human-dialed houses and four AI phone services, each with one limit.

1. Alchemic: Best for AI Phone Interviews That Probe, in Many Languages

Alchemic runs end-to-end consumer research at scale, and its phone service places outbound AI calls to any working number in supported regions. The AI introduces the study, captures consent on the first turn, probes, and keeps pace when a respondent switches languages mid-call. Calls run concurrently, are recorded and searchable, and code into themes as they complete. The service publishes 57+ languages including Hindi, Tamil and Telugu, recruits through managed fieldwork or bring your own across 14 markets including the USA and the UK, and can mix WhatsApp and web interviews into the same study. A 200-interview qualitative study runs from brief to live dashboard in 3 days.

• Strengths: Qualitative and quantitative in one call, a research team alongside the software, and context that carries across waves.
• Limitations: Its own page lists extremely rare languages, samples under 20 and interviews needing a live clinician as poor fits. In the US an AI call to a cell phone needs prior consent, so it suits lists and panels, not cold dialing.
• Best for: Churn, satisfaction and category interviews with customers you can already reach.

2. Cora Intelligence: Best for Running a Fixed Questionnaire With AI Callers

Cora Intelligence builds voice AI agents for business calls; its research-firm offer applies them to fieldwork. A firm uploads its questionnaire, even as a PDF; Cora's agents call the sample in parallel, run every interview exactly as written in multiple languages, and return structured responses with a transcript of every call. It charges per completed interview and says it can scale up or down overnight.

• Strengths: Identical wording on every call, with no interviewer drift.
• Limitations: Research is one use case beside payment-recovery calls, and exact wording means it is built not to probe.
• Best for: Research firms moving an existing CATI questionnaire to AI calling.

3. Drive Research: Best for a Full-Service Phone Study With a Report at the End

Drive Research sells phone surveys as a managed project: a proposal, a dedicated project team, screener and questionnaire design, and recruiting with online pre-screening, phone re-screening and reminders. After trained interviewers finish, the team codes open ends, builds tables and charts, writes the report and runs a debrief.

• Strengths: Analysis included, for B2B and B2C.
• Limitations: Its page says fieldwork begins "in the next few weeks" after sign-off, and it publishes no seat count or language list.
• Best for: A team without in-house analysts that wants a recommendation, not a data file.

4. I/H/R Research Group: Best for Cell Phone Samples Dialed by Hand

Established in 1976, I/H/R runs a 52-station Las Vegas call center, which it says is expanding to 90, with live digital monitoring and recording on every station. It uses no predictive, power or auto dialers, which keeps the whole center TCPA-compliant for wireless numbers, and fields pure probability random-digit-dial samples to AAPOR and government-style standards. A bilingual team interviews in English and Spanish, and it has also fielded Mandarin, Japanese, Vietnamese, Thai, Portuguese and French.

• Strengths: A compliance posture built for cell-only adults, with sampling from sister company Scientific Telephone Samples.
• Limitations: Fifty-two stations caps nightly output below the largest houses, and hand dialing costs interviewer time.
• Best for: General-population and public-opinion studies that must include cell-only adults with no consent on file.

5. Ironwood Insights Group: Best for High-Volume Quotas Across Several Call Centers

Ironwood operates wholly owned call centers it describes as 100% TCPA and HIPAA compliant. The largest is a 90-seat center in Rexburg; Houston moved into a new 50-seat center in July 2026 with expanded Spanish capacity; and a 35-seat center in San Jose, Costa Rica, works in English and Spanish. It pairs CATI with online panels and text-to-web.

• Strengths: Several centers that can split a large quota.
• Limitations: Its pages give different totals, 250 seats on one and 300+ stations on another, so confirm current capacity.
• Best for: Healthcare, customer satisfaction and current-events tracking that needs many completes a night.

6. Koji: Best for Adding Phone to a Self-Serve AI Interview Tool

Koji, built in Amsterdam, runs AI-moderated interviews by voice, chat or phone, probing for the reason behind each answer in the participant's own language. It lists 30+ languages and a voice library matched to audience and region, pitches one platform in place of recruiters, moderators and transcription vendors, offers EU or US data residency, and runs studies from Claude, ChatGPT or any MCP client.

• Strengths: Fast self-serve setup and one study across markets.
• Limitations: Its home page shows no phone-specific language list or US calling-consent process, so ask for both.
• Best for: Product and UX teams already running AI interviews who want some respondents on a call.

7. Miravoice: Best for AI Calling on Long Structured Surveys

Miravoice, from VKL Research, runs phone surveys with AI interviewers that speak 14 languages, call thousands of people at once and answer callbacks day or night. It publishes its field research: work with NORC at the University of Chicago, a human-to-AI handoff on the CDC's BRFSS in Florida and Ohio alongside ICF, and a pilot on the SSRS Opinion Panel published with SSRS researchers. In its second wave the AI ran a 123-question, 30-minute omnibus instrument, and 86% of finishers rated the experience neutral or positive.

• Strengths: Methodological transparency and public health experience.
• Limitations: Its own paper with Trent D. Buskirk, comparing AI interviewers with IVR, finds them for now a better fit for quantitative than qualitative work, and in that second wave 29 of 70 people who picked up hung up during the introduction.
• Best for: Survey organizations automating long, exact-wording quantitative instruments.

8. Precision Opinion: Best for Very Large CATI Projects Staffed In Person

Precision Opinion has sold data collection to research firms for nearly 20 years. In 2022 it built a 42,000-square-foot, 600-seat CATI center, and it says it never moved to remote interviewing, never offshores and never outsources. It adds SMS and mail-to-web, was SOC 2 Type II certified in 2020, and has been sole subcontractor for nearly 20 years on what it calls the industry's largest federal data collection project.

• Strengths: Scale, in-person supervision and data security proven on federal work.
• Limitations: It sells data collection to research firms, so a brand team needs its own or a partner's design and analysis.
• Best for: Social science and healthcare firms co-fielding very large telephone samples.

9. SSRS: Best for Probability Samples That Must Stand Up in Public

SSRS runs the SSRS Opinion Panel, a probability-based panel of US adults recruited through address-based sampling and random-digit dialing of prepaid cell numbers. Panelists answer on the web or by phone, so adults without internet stay in the frame, and recruitment runs in English and Spanish. Its omnibus fields twice a month with 1,000 adults. SSRS also runs a TCPA-compliant call center, holds ISO 27001 certification, and works with Pew Research Center on the American Trends Panel.

• Strengths: A true selection probability for every respondent, so weights can fully correct for the sample design.
• Limitations: That rigor is built for publishable estimates; a quick directional read of your own customers does not need it.
• Best for: Health, policy, media and public-opinion research that will be published or cited.

CATI, IVR and AI Phone Interviews for the Buyer

How each mode works is covered in AI phone interviews vs IVR vs voice bots and what CATI is and what is replacing it. Three differences change a quote:

• Billing unit. CATI fieldwork is built on interviewer hours; Cora, for one, charges per completed interview, shifting low-incidence risk to the supplier.
• Turnaround. Human fieldwork is paced by seats and shifts; Drive quotes weeks, while Miravoice says interviews complete in days.
• Depth. IVR records keypad answers, AI readers keep exact wording, and AI moderators such as Koji and Alchemic follow up on what was said; how much latitude a system takes varies by tool.

For choosing among AI platforms, the guide to AI phone interview platforms sets out the checks.

What a Phone Study Costs and What Drives the Quote

Only Cora publishes a starting price per complete, and every quote turns on the same inputs; pricing models across research types are compared in what consumer research costs.

• Length and incidence. A 25-minute call where one person in ten qualifies costs far more than a 10-minute call to a list where everyone qualifies.
• Sample type. Hand-dialing cell numbers takes more interviewer time than dialing a consented list.
• Attempts per number. The CDC's 2024 BRFSS data quality report allows up to 15 attempts before a case closes, for a median state response rate of 43.9%, ranging from 30.7% to 64.8%. Commercial cold calling runs far lower, as why telephone response rates collapsed traces.
• Languages. Each needs bilingual interviewers or a tested AI voice.
• Deliverable. A data file costs less than a written report.

With concurrent AI calls, cost per complete stops tracking headcount.

Who a Phone Sample Reaches That Online Panels Miss

The NHIS estimates for July to December 2024 found 78.7% of adults, about 205 million people, in wireless-only households, 19.8% with both a landline and a cell phone, and 0.9% landline-only. Adults with their own cell phone and no landline at home made up 88.1% of renters and 83.0% of adults below the poverty line, but only 57.9% of adults 65 and over, which is where landlines still matter. The NHIS is the right source because it rests on household interviews, not on anyone answering a cold call.

Pickup is uneven. Pew found adults 18 to 29, Hispanic and Black adults, and lower-income adults more likely than older, White or higher-income peers to answer unknown numbers, so quotas and weights should track who picks up, not just who was dialed.

A voice call needs no smartphone, app, data plan or reading, which keeps feature-phone and low-literacy respondents in; reaching respondents without smartphones covers those modes. For people who ignore calls but read messages, Alchemic runs WhatsApp-native interviews with no link and no app inside the same study.

When a CATI House or Another Approach Is the Better Choice

An AI phone interview that probes is the wrong buy in six situations:

1. Cold random-digit dialing of cell phones. With no consent on file, a human-dialed house such as I/H/R or SSRS is the lawful route; see TCPA rules for AI phone research.
2. Estimates that will be published. A probability panel such as SSRS's gives every respondent a known selection probability.
3. Very long standardized instruments. Precision Opinion's seats, or Miravoice and Cora for exact-wording AI calls, fit better than an open interview.
4. Visual stimuli. Packs, ads and prototypes need a screen.
5. A handful of senior executives. Fewer than 20 hard-to-schedule interviews reward a skilled human interviewer.
6. Sensitive clinical topics that need a live clinician on the line.

Gallup's February 2026 note on its AI phone pilots, more than half a million call attempts in seven languages, reached the same point: in the US, AI calling may need a preconsented sample or a human who starts the call and hands off.

How to Shortlist a Phone Survey Company

Ask every shortlisted company, human or AI, for six things:

1. Unedited recordings from a comparable study, in each language you need.
2. The consent basis for every cell number it will dial, in writing.
3. Capacity: seats or concurrent lines, and attempts per number.
4. The deliverable: data file, coded open ends, or a report.
5. Security evidence, such as a SOC 2 report or ISO 27001 certificate.
6. A paid pilot of 30 to 50 completes, read by break-off point.

Where Phone Research Breaks Down, Human or AI

• Answering is the bottleneck. With about one adult in five routinely picking up unknown numbers, and over 2.6 million Do Not Call complaints filed with the FTC in fiscal 2025, legitimacy signals do as much as any script.
• Nothing can be shown. A call carries no pack or prototype, and voice-only interviews carry no facial signal.
• Length punishes every caller. Hanging up is one gesture, so long batteries lose respondents.
• People and systems both drift. Human interviewers vary in tone and probing; a rigid AI reads the guide regardless of the answer, and latitude varies by tool.
• In the US, consent narrows AI sampling to people who opted in to something.

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.

Sources and Methodology

• CDC NCHS, NHIS wireless substitution estimates, July to December 2024: phone status by age, income and tenure.
• CDC, BRFSS 2024 Summary Data Quality Report: state response rates and maximum call attempts.
• Pew Research Center, 2020: who answers calls from unknown numbers.
• FTC, National Do Not Call Registry Data Book, fiscal 2025: complaint volume behind unanswered calls.

Sources are linked inline; a research seller's note and Miravoice's pilot paper are named, unlinked. Company facts come from each company's own site on 23 September 2026; unpublished facts are marked as such.