--- title: "Best voice survey tools 2026: 12 tools compared" description: "We compared 12 voice survey tools: AI interviews, recorded voice answers and research suites. See who asks the follow-up and what each one costs." url: "https://raconte.ai/en/blog/top-voice-survey-tools" --- # The best voice survey tools in 2026, sorted by how the answer gets spoken We compared 12 voice survey tools: AI interviews, recorded voice answers and research suites. See who asks the follow-up and what each one costs. October 11, 2026 Godefroy Key takeaways * A voice survey tool either holds a live conversation with each respondent or records a spoken answer to a question written in advance. Decide which one you need before you compare prices. * People say about three times more words out loud than they type in an open survey box. * A microphone button under a written question loses respondents: in a German study, 39% of people asked to speak gave no answer, against 8% of people asked to type. * Five of the twelve tools publish no price for their voice features, so ask them for the entry price in writing. * Phonic, long listed among voice survey tools, disabled its survey accounts on 1 October 2025. A voice survey tool collects answers that people speak instead of type. The twelve tools compared here do it in three ways. Six of them hold a live conversation in which an AI asks the follow-up questions: Raconte, Koji, UserCall, Glaut, Versive and Listen Labs. Four record a spoken answer to a question written in advance: VideoAsk, Speak AI, Voiceform and Yazi. Voxpopme and Qualtrics add audio and video answers to a full research suite. This page compares the twelve on how the answer gets spoken, who asks the follow-up, what each costs to start and how you connect it to your other tools, with prices read in October 2026. People say more than they type. Timo Lenzner, a survey methodologist at GESIS, the Leibniz Institute for the Social Sciences, and his co-authors Jan Karem Höhne and Konstantin Gavras asked 1,001 smartphone respondents in Germany the same open questions, by voice for some and by keyboard for others. They published the results in the [Journal of Survey Statistics and Methodology](https://academic.oup.com/jssam/article/12/5/1295/7726361) in 2024. The spoken answers averaged 44 words and the typed ones 14. In the same study, 39% of the people asked to speak gave no answer at all, against 8% of the people asked to type. That study tested a microphone button under a written question. It did not test a conversation with an AI, so it says nothing about whether a spoken follow-up keeps more people answering. Raconte publishes this article and is one of the twelve tools it compares. We put our own group first because this is our article. The three groups answer three different needs, so their order says nothing about which one is better. The numbers run from 1 to 12 so the list reads in one go, but a number only ranks a tool against the others in its group. Inside a group, the order follows one rule you can check and dispute: how far you can go on your own before talking to sales. Every entry links to the vendor’s site. “Voice survey” also names the phone survey a contact center plays after a support call, built with software such as Genesys Cloud or Verint. This page leaves those phone surveys out. It covers the tools a product, research or customer experience team sends by link. We left Phonic out. This voice and video survey tool was bought by the advertising company Infillion in February 2023, announced its closure in spring 2025 and disabled every survey account on 1 October 2025. The phonic.ai address now presents a platform for building voice agents, billed by the minute of conversation, with no survey product. ## How we compared them We read each vendor’s pricing page, product pages and help center in October 2026, and compared the twelve tools on six criteria. * We note how the answer gets spoken: in a live conversation or in a recording, in the browser, on the phone or in WhatsApp, and whether the respondent can type instead. * We note who asks the follow-up question when an answer stays vague: an AI that speaks, an AI that adds a limited number of probes, or nobody. * We list what comes back: transcript, audio, sentiment, themes across all the answers, exports. * We record the price to start. A published number counts for more than a quote. Where a vendor publishes nothing, we say so. * We check whether you can drive the tool from code: a public API, webhooks, an MCP server for AI agents. * We check the languages, French included, and where the data is hosted when the vendor says. Inside each group, a free plan that renews every month and includes spoken answers comes first, then a free trial or a one-off free allowance, then the tools with no free start for voice. Between two tools at the same level, the lower published price comes first and a published price comes before a quote. That order suits a small team that wants to hear from its customers this month. A team that ranks on panel size, video, the number of languages or analysis across hundreds of answers would put Listen Labs, Voxpopme or Qualtrics higher. Tool How the answer is spoken Follow-up questions Free start Paid plans from Languages Developer access [Raconte](https://raconte.ai) Live conversation in the browser AI, out loud, during the whole conversation 60 interview minutes every month $0.12 per minute of speech English, French REST API, MCP, webhooks, CLI, SDK [Koji](https://www.koji.so/) Live conversation, by voice or chat AI, out loud 10 credits, once €29 a month 30+, no list published REST API, webhooks, MCP [UserCall](https://www.usercall.co/) Live conversation, by voice or text AI, out loud First interview $89 a month 25+, no list published Webhooks, MCP [Glaut](https://www.glaut.com/) Voice interview, typing allowed AI, in real time Free trial, terms not stated Not published 50+, no list published Embeds in survey tools [Versive](https://www.getversive.com/) Live conversation, by voice, text or video AI, out loud Free plan without AI interviews $99 a month, billed yearly 20, French included MCP on every plan, API on Enterprise [Listen Labs](https://listenlabs.ai/) Interview moderated by an AI AI Demo only Quote 120+, no list published Not documented [VideoAsk](https://www.videoask.com/) Recorded audio or video answer None, the next step is scripted 20 minutes a month, per Capterra $24 a month, billed yearly, per Capterra Not stated API, webhooks, Zapier [Speak AI](https://speakai.co/audio-video-surveys/) Recorded audio or video answer None 7-day trial $25 per user a month, or credits at $1 50 to 100+, depending on the page API, MCP, CLI, webhooks [Voiceform](https://www.voiceform.com/) Recorded voice or video answer in a survey AI probes, 40 a month on Lite Free trial Not published 50+, no list published API, SDK, webhooks, Zapier [Yazi](https://www.askyazi.com/) WhatsApp voice note AI, on each answer None $600 a month or per project 100+, no list published Custom, on Enterprise [Voxpopme](https://www.voxpopme.com/) Recorded video, plus AI interviews by voice AI moderator ”Try it free”, terms not stated Quote Not stated Not documented [Qualtrics](https://www.qualtrics.com/) Recorded audio in a survey question None on the audio answer None for the audio question Quote Not stated REST API, MCP ## An AI holds the conversation Raconte, Koji, UserCall, Glaut, Versive and Listen Labs replace the survey with a conversation. An AI asks an opening question, listens to the spoken answer and chooses the next question from what it just heard. ### 1\. Raconte ![Screenshot of the Raconte homepage](/.netlify/images?url=_astro%2Fhomepage-en.evuUlE9t.png&w=1920&h=1200&dpl=6ad1cc428d394a5c89e0155b) [Raconte](/en) runs voice interviews with an AI. A team writes the interview as a prompt, then shares it as a public link or a QR code, or sends email invitations from a pasted list or a CSV file. The respondent opens the link in a browser, with no account and nothing to install, and talks whenever it suits them. The AI asks its questions out loud, one at a time, and asks a follow-up when an answer stays unclear. After the call, the team reads the transcript and a summary, replays the audio answer by answer and sees the sentiment of each message. The first 60 interview minutes each month are free and need no card. After that, Raconte charges $0.12 per minute of active speech, silences excluded, with top-ups from $12 and no subscription, which puts a ten-minute interview at about $1.20. The team sets a maximum length per interview. Raconte also sells an [Enterprise plan](/en/pricing), with volume discounts, white-label, SSO and an SLA. A team can drive Raconte entirely from code. Its OpenAPI spec is public at [raconte.ai/openapi.json](https://raconte.ai/openapi.json), `@raconte/cli` and `@raconte/node-sdk` are on npm, the [MCP server](/en/docs/mcp) exposes eight tools to AI agents, and [webhooks](/en/docs/webhooks) fire when an interview starts and when it ends. All of it is open on the free allowance. Respondents answer by voice only, with no typed fallback, no video and no phone line. In exchange, the respondent has no form to fill in, since the AI asks each question out loud. The transcript holds the whole conversation, follow-ups included. Interviews run in English and French only, with twenty voices in those two languages. Anyone can listen to those voices before sending an interview. Raconte has no closed questions, no respondent panel, no file export and no report across interviews. Every question is open, so the AI can follow up on each of them. Raconte interviews the people a team already knows and sends each result to the team’s other tools through the [API](/en/docs/api) and the webhooks. To compare interviews, a team gives the transcripts to an AI assistant through the MCP server. The data is stored in the European Union, at the French host Scaleway. ### 2\. Koji [Koji](https://www.koji.so/) is built by Koji Labs, a company based in Amsterdam. Its AI moderator interviews each participant by voice or by chat, in the browser, and answers as soon as the participant stops speaking. Koji also mentions interviews by phone on its homepage. Each interview gets a quality score, a summary and a sentiment reading. Koji then writes a report across all the interviews of a study and exports the data as CSV or JSON. It handles structured question types alongside the open conversation and recruits participants through the Cint panel network. Koji advertises more than 30 languages without publishing the list. Koji bills in credits, priced in euros. The free plan gives 10 credits once. Insights costs €29 a month for 29 credits and Interviews €79 a month for 79 credits, with Enterprise from €500 a month. A voice interview uses 3 credits, a text interview 1 and a report refresh 5, so the €29 plan covers nine voice interviews a month. Extra credits cost €1 each. Koji does not charge for an interview that scores under 3 out of 5 on quality. Koji documents a REST API, webhooks, embeddable widgets and an MCP server. Koji includes API keys from the Interviews plan, at €79 a month. ### 3\. UserCall ![Screenshot of the UserCall homepage](/.netlify/images?url=_astro%2Fusercall.CzdMtXfo.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [UserCall](https://www.usercall.co/) runs AI interviews by voice, by text or both. Its moderator follows the discussion guide and decides after each answer whether to probe, clarify or move on. Participants join by link, and outbound phone calls are an add-on to the Pro plan. UserCall’s analysis groups the answers into themes, links each finding to the quotes behind it and flags a finding as strong, weak, overstated or contradicted. Interviews can start from a product event through integrations with PostHog, Mixpanel, Amplitude, Segment and Stripe. UserCall also gives access to a panel it puts at more than four million participants. The first interview is free. After that, Lite costs $89 a month for 100 credits, which UserCall says covers ten to twenty interviews of up to twelve minutes. Core costs $199 a month for 30 to 60 interviews and Pro $399. UserCall advertises more than 25 languages without publishing the list, and it has webhooks and an MCP server. Electrolux, ClassDojo and Penn State appear among its customers. UserCall is one of several [user interview platforms](/en/blog/top-user-interview-platforms) with an AI moderator. ### 4\. Glaut ![Screenshot of the Glaut homepage](/.netlify/images?url=_astro%2Fglaut.BK-3ch1i.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Glaut](https://www.glaut.com/) puts AI-moderated voice interviews and classic survey questions in one study. The respondent chooses between speaking and typing, and the AI asks its follow-up questions in real time. Glaut sells mostly to research agencies and institutes: Ipsos, Kantar, BVA Doxa and Mondadori appear among its customers. A Glaut interview can be embedded in the survey software a research team already uses, so a panel survey can open a conversation in the middle of a questionnaire. Glaut then codes the answers and writes the report. It advertises more than 50 languages without publishing the list. Glaut reports an average completion rate of 86% and open answers 2.3 times longer than in a survey. Both figures are its own and come without the sample behind them. Glaut publishes no price. Its homepage has a “Try Glaut for free” button and does not state the terms of the trial. Glaut’s site has no public API documentation, so ask Glaut if you need to start interviews from your own code. ### 5\. Versive ![Screenshot of the Versive homepage](/.netlify/images?url=_astro%2Fversive.PNSW1RnU.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Versive](https://www.getversive.com/) came out of Y Combinator and combines surveys, AI interviews and prototype tests. Its AI moderator runs interviews by voice, text or video, and the voice mode is a real-time spoken conversation in the browser. Versive lists 20 languages, French included, and offers hosting in the United States or in the European Union, with a SOC 2 Type 2 report. Beyond interviews, Versive tests websites and Figma prototypes and can run a study on synthetic personas before real participants. It recruits through an integration with Respondent. The free plan covers 100 survey responses a month and includes no AI interviews. Pro costs $99 a month billed yearly, with 240 AI interviews or AI tests a year and two seats. An interview on the real-time voice engine counts double, which makes 120 voice interviews a year. Enterprise is on quote. The MCP server is available on every plan, while the public API and SSO come with Enterprise. ### 6\. Listen Labs ![Screenshot of the Listen Labs homepage](/.netlify/images?url=_astro%2Flistenlabs.DKEwGieW.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Listen Labs](https://listenlabs.ai/) announced in January 2026 a $69M Series B led by Ribbit Capital, which brought its total funding to $100M. Its AI moderator interviews participants and asks follow-up questions, then the platform delivers reports, highlight reels and slides, and keeps past studies searchable. Listen Labs recruits from a network it puts at more than 50 million people and advertises more than 120 languages without publishing the list. Anthropic, Microsoft and Sweetgreen appear among its customers. Listen Labs says its customers’ answers run three times longer than average, a figure it publishes without a method. The site shows no price and has no self-serve signup: the first step is a demo. The site has no public API documentation. For a large organisation that wants the participants, the interviews and the deliverables in one contract, Listen Labs is the most complete vendor of the twelve. Outset and Strella, two other [user interview platforms](/en/blog/top-user-interview-platforms), sell the same kind of contract. ## The respondent records a voice answer With VideoAsk, Speak AI, Voiceform and Yazi, the team writes the questions in advance. The respondent presses a button and records an answer, in a form, a video thread or a WhatsApp chat. Voiceform and Yazi add AI follow-up questions on top. ### 7\. VideoAsk ![Screenshot of the VideoAsk homepage](/.netlify/images?url=_astro%2Fvideoask.BmNMCbvu.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [VideoAsk](https://www.videoask.com/) belongs to Typeform. A team records a short video question, and the respondent answers by video, by audio or by text. The team decides the next step in advance, with branches that depend on the answer. No AI asks a follow-up question. The format suits a message that works better with a face: a founder asking customers for a testimonial, a recruiter screening candidates, a coach collecting questions before a session. The respondent can pick audio when they prefer to stay off camera. VideoAsk has an API, webhooks and a Zapier connector, and says more than 10,000 companies use it. VideoAsk bills by the minute of video and audio processed, yours and your respondents’. The software directory Capterra lists a free plan with 20 minutes a month, a Grow plan at $24 a month billed yearly for 100 minutes and a Brand plan at $40 a month for 200 minutes. VideoAsk’s own pricing page tells teams that process more than 200 minutes a month to start from Brand and add minutes. ### 8\. Speak AI ![Screenshot of the Speak AI audio and video surveys page](/.netlify/images?url=_astro%2Fspeakai.-yXKNyfI.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Speak AI](https://speakai.co/audio-video-surveys/) started as a transcription and analysis tool for recordings and added audio and video surveys to it. Participants record their answer in the browser, with no app and no account. Speak AI transcribes each recording when it arrives, then extracts keywords, topics and sentiment. A team can also question the whole library of answers through an AI chat. Speak AI analyses any recording a team uploads, such as sales calls, focus groups and past interviews, next to the survey answers. No AI asks the respondent a follow-up question. Speak AI’s trial lasts 7 days. After that, pay-as-you-go credits cost $1 each, with standard transcription at 2 credits per hour of audio. Pro costs $25 per user a month, or $20 billed yearly, and includes the branded recorder and the surveys. Speak AI has an API, an MCP server, a command line tool, webhooks and a Zapier connector. Its pages give two different language counts, 50 and more than 100, so ask which one applies to your survey. ### 9\. Voiceform ![Screenshot of the Voiceform homepage](/.netlify/images?url=_astro%2Fvoiceform.D5A_9GG-.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Voiceform](https://www.voiceform.com/) is a survey builder where an open question can be answered by voice or by video, next to the usual closed questions. Its AI probing generates a follow-up question when an answer deserves one. When Phonic closed, it recommended Voiceform to its customers. Voiceform transcribes the answers, groups them into themes and writes summaries and reports. It advertises more than 50 languages with AI translation, without publishing the list, and it lists a phone survey (IVR) integration. Stanford, RingCentral, Shopify and Ipsos appear among its customers. Voiceform publishes a customer case where spoken open answers averaged 82 words against 40 typed, a figure that comes from Voiceform. The Lite plan includes 200 answers and 40 AI follow-ups a month, with recordings limited to two and a half minutes per question, and starts with a free trial. Voiceform shows no price for Lite on its pricing page, and larger plans are on quote. For developers, Voiceform documents an API, an SDK, webhooks and connectors for Zapier, Salesforce, HubSpot and Prolific. ### 10\. Yazi ![Screenshot of the Yazi homepage](/.netlify/images?url=_astro%2Faskyazi.BcoW_qi2.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Yazi](https://www.askyazi.com/) runs its research inside WhatsApp. The participant receives the questions in a chat and answers in text or with a voice note, which Yazi transcribes and groups into themes. An AI interviewer asks follow-up questions on each answer. Yazi does not say whether those follow-ups arrive as text or as audio. The respondent needs WhatsApp and nothing else, which matters in the countries where people use it far more than email. Yazi also runs diary studies over several days and sells its own audience panels. Virgin Active, Old Mutual, Capitec and Ipsos appear among its customers. Yazi hosts data in South Africa and in the European Union and advertises translation across more than 100 languages without publishing the list. The Basic plan starts at $600, per month or per project, for up to 100 AI interviews or 250 survey respondents. Enterprise is on quote and adds white-label and custom integrations. Panel respondents cost $6 to $9 per response. ## Research suites with audio and video answers Voxpopme and Qualtrics sell a full research platform to large companies. Spoken answers are one feature among many. ### 11\. Voxpopme ![Screenshot of the Voxpopme homepage](/.netlify/images?url=_astro%2Fvoxpopme.nUTLvzVJ.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Voxpopme](https://www.voxpopme.com/) built its name on video surveys, where consumers film a short answer on their phone. The platform now also covers live interviews, an AI moderator that runs voice interviews and asks follow-up questions, a research agent named Compass and a repository for past studies. Respondents come from Voxpopme’s own panel, from a customer list or from a custom audience, and studies run in 27 countries. McDonald’s, Microsoft, PepsiCo, Mars and Airbnb appear among its customers. The video is the deliverable as much as the transcript, since insight teams cut clips of customers to show in a meeting. Voxpopme publishes no price. Its site has a “Try it free” button and offers pilot projects, whose terms you discuss with sales. The site has no public API documentation. ### 12\. Qualtrics ![Screenshot of the Qualtrics homepage](/.netlify/images?url=_astro%2Fqualtrics.Be2HOh8H.png&w=1920&h=1080&dpl=6ad1cc428d394a5c89e0155b) [Qualtrics](https://www.qualtrics.com/) is a survey platform that many large research departments already run. Its Video Response question type has an option to record audio only, with a default limit of three minutes per answer that can be raised to ten. Qualtrics then analyses the topics and the sentiment of the recordings next to the rest of the survey data. A company that already runs its surveys on Qualtrics can add a spoken question to a questionnaire and keep its logic, panels, quotas and dashboards. No AI follows up on a recorded answer. The Video Response question needs a specific permission on the licence, so check with your Qualtrics administrator before you plan a study around it. Qualtrics publishes no price. It has a REST API and an MCP server. Qualtrics is above all a [customer feedback tool](/en/blog/top-customer-feedback-tools) in which the audio answer is one question type. ## How to choose between them Start with what you want back from each respondent. * For the story behind a score or a cancellation, take a tool that holds a conversation. You can start alone with three of them: Raconte on 60 free minutes a month, Koji from €29 a month and UserCall from $89 a month. Glaut fits a research agency that mixes closed questions and interviews in one study. Listen Labs fits a large company that wants the participants in the same contract. * For one spoken answer inside a questionnaire, take a tool that records. Voiceform adds a limited number of AI probes, Speak AI adds the analysis of other recordings, and Qualtrics keeps the question inside a survey program that already exists. * For a face or a testimonial, take video. VideoAsk and Voxpopme are built for it, while a voice-only tool gives you no clip to show. Then check how your respondents will answer. Many people skip a question they are asked to answer out loud: 39% in the German study by Lenzner, against 8% for a typed answer. Glaut lets the respondent type instead. Koji, UserCall and Versive have a text mode. Yazi goes through WhatsApp, where a voice note is a habit. Raconte takes spoken answers only, so send your interview to people who have a few quiet minutes. Check the languages too. Raconte interviews in English and French only, with twenty voices to choose from in those two languages. Versive lists French among its 20 languages. Koji, UserCall, Glaut, Voiceform, Yazi and Listen Labs advertise between 25 and 120 languages without a public list, so ask for yours in writing. If you need one analysis across two hundred answers, Koji, UserCall, Glaut, Speak AI, Voiceform, Yazi, Listen Labs, Voxpopme and Qualtrics do it for you. Raconte gives you each interview on its own, so you compare them through its API or its MCP server. The tools bill in different units. Raconte bills by the minute of speech, Koji and UserCall by credit, Versive by interview per year, VideoAsk by processed minute and Speak AI by seat or by credit. Count your real volume before you compare. Fifty ten-minute interviews a month come to about $53 at Raconte once the free hour is used, to €150 at Koji with the Interviews plan and 71 extra credits at €1 each, and to $199 at UserCall, whose Core plan covers 30 to 60 interviews. A technical team can also build its own voice survey on a voice agent platform billed by the minute. It then writes the interview logic, the consent screen, the storage of transcripts and the analysis on its own. ## Frequently asked questions ### What is a voice survey? A voice survey is a survey people answer by speaking instead of typing. The respondent opens a link, allows the microphone and talks. The tool transcribes the answer and usually analyses it. Some voice survey tools record one answer per written question. Others hold a conversation in which an AI asks the follow-up questions. The same words also name the automated phone survey a contact center plays after a support call. ### What is the best voice survey tool? The best voice survey tool depends on what you want back from each respondent. For a conversation with follow-up questions, you can start Raconte, Koji and UserCall without a sales call. Raconte takes spoken answers only, in English or French, while Koji and UserCall also have a text mode and more languages. For a recorded answer inside a classic survey, Voiceform and Qualtrics keep the closed questions next to it. VideoAsk and Voxpopme are built for video answers. For respondents on WhatsApp, Yazi collects voice notes there. ### Do people give better answers by voice than by typing? People give longer answers by voice. In a 2024 study of 1,001 smartphone respondents in Germany, published in the Journal of Survey Statistics and Methodology, spoken answers averaged 44 words and typed answers 14. Fewer people answered, though: 39% of the people asked to speak skipped the question, against 8% of the people asked to type. Melanie Revilla and Mick Couper, survey researchers at Universitat Pompeu Fabra and the University of Michigan, reported in [Survey Research Methods](https://ojs.ub.uni-konstanz.de/srm/article/download/8456/7886?inline=1) in 2026 that about one respondent in five chose voice when both options were offered. These studies tested a one-way recording. They did not test a conversation with an AI. ### How much does a voice survey tool cost? Seven of the twelve tools publish a price for their voice features. Raconte gives 60 free minutes a month, then charges $0.12 per minute of speech. Koji starts at €29 a month, UserCall at $89 a month and Versive at $99 a month billed yearly. Speak AI costs $25 per user a month. Yazi starts at $600 a month or per project. VideoAsk’s plans start at $24 a month billed yearly according to Capterra. Voiceform, Glaut, Listen Labs, Voxpopme and Qualtrics give their price on request. ### Can respondents answer a voice survey by phone or on WhatsApp? A few tools go beyond the browser. Yazi runs entirely in WhatsApp, where respondents answer with voice notes. UserCall sells outbound phone calls as an add-on to its Pro plan, Koji mentions phone interviews on its homepage, and Voiceform lists a phone survey (IVR) integration. Raconte, Versive and Speak AI work in the browser only, from a link. ### Which voice survey tools work in French? Raconte interviews in French and English, and Versive lists French among its 20 languages. Koji, UserCall, Glaut, Voiceform, Yazi and Listen Labs advertise between 25 and 120 languages without publishing the list, so ask them to confirm French in writing. Check the language of the spoken questions as well as the transcription: a tool can transcribe French answers and still ask its questions in English. ### What happened to Phonic? Phonic was a voice and video survey tool founded in 2020. The advertising company Infillion bought it in February 2023. In spring 2025 Phonic announced it was being discontinued: surveys stopped collecting answers on 30 June 2025, exports closed on 30 September 2025, and accounts were disabled on 1 October 2025. The phonic.ai address now shows a platform for building voice agents, with no survey product. Phonic recommended Voiceform to its customers when it closed. ### Can an AI agent run a voice survey through an API or an MCP server? Yes, with the tools that expose one. Raconte, Koji, UserCall, Versive, Speak AI and Qualtrics publish an MCP server, so an agent can create a study, send it and read the answers. Raconte publishes its OpenAPI spec, a command line tool and a Node SDK, all open on the free allowance. Koji gives API keys from its €79 plan and Versive keeps its public API for Enterprise. Listen Labs, Glaut and Voxpopme have no public API documentation on their sites. ### Do respondents need to install an app? No, in most cases. Raconte, Koji, Versive and Speak AI open in the browser from a link, and the respondent only allows the microphone. Speak AI and Raconte state that the respondent needs no account. Yazi needs WhatsApp on the respondent’s phone. ## Where to start A voice survey tool holds a conversation or records an answer. Pick between the two first, then compare prices on your real volume and ask for your language in writing. If you want to hear the story behind a score from customers you already have, [Raconte](/en) runs that conversation by voice, in English or French. Write the interview as a prompt, share the link and read the transcripts as the answers come in. The first 60 minutes each month are free and need no card. After that, a minute of speech costs [$0.12, with no subscription](/en/pricing). To measure a score before you ask why, compare the [customer feedback tools](/en/blog/top-customer-feedback-tools). Ready to launch your first interview? Describe your interview, share the link: an AI agent runs it by voice with each participant. You get back the transcript and analysis. [Try for free](/login) ## Continue reading [![Best customer feedback tools 2026: 11 tools compared](/.netlify/images?url=_astro%2Fthumbnail.BEBgL6FX.jpg&w=1376&h=768&dpl=6ad1cc428d394a5c89e0155b) October 4, 2026 ## Best customer feedback tools 2026: 11 tools compared We compared 11 customer feedback tools across interviews, surveys, in-app prompts and VoC platforms: who asks the follow-up, what it costs, how to plug it in. ](/en/blog/top-customer-feedback-tools)[![Best user interview platforms 2026: 13 tools compared](/.netlify/images?url=_astro%2Fthumbnail.CKayV9Va.jpg&w=1376&h=768&dpl=6ad1cc428d394a5c89e0155b) August 1, 2026 ## Best user interview platforms 2026: 13 tools compared Most user interview platforms never ask a question. We compared 13 tools on who runs the interview, what it costs and whether you can call it from code. ](/en/blog/top-user-interview-platforms)[![Typeform alternative: when the answer is a story](/.netlify/images?url=_astro%2Fthumbnail.CNe6xQoR.jpg&w=1200&h=630&dpl=6ad1cc428d394a5c89e0155b) September 28, 2026 ## Typeform alternative: when the answer is a story Tally, Google Forms or a voice interview: which Typeform alternative fits, depending on whether you need a short answer or the story behind it. ](/en/blog/alternative-typeform)