> voicepanel / guides / ai research platforms compared

The best AI research platforms in 2026, compared

Every research vendor now says it uses AI. The real differences are in who runs the research, whether your own systems and AI agents can reach it, whether you can see and change what the AI does, and whether it passes an enterprise security review. We compared 20 platforms that collect feedback from real people, not synthetic respondents.

Updated September 28, 2026

View as Markdown

Voicepanel wrote this guide, and Voicepanel is one of the platforms in it. We checked each claim against public sources in September 2026 and list them at the end. If you find an error, email hello@voicepanel.com and we will correct it.

The short answer

Why Voicepanel? Because your research should plug into everything you build, and never wait in a vendor's queue.

Voicepanel is the best AI research platform overall, and the most flexible one for product builders and AI agents. One study can mix AI interviews, usability tasks, and surveys. A REST API and an MCP server let your own tools and agents run studies with real people.

Best platform for each need

Best overall, for modern product and insights teams

1. Voicepanel

Enterprise survey programs in regulated industries

2. Qualtrics

Large usability programs with human moderation

3. UserTesting

UX research teams that want to hand off study setup and fieldwork

4. Outset

Consumer concept and pricing research

5. Conveo

A managed AI interview program with a large budget

6. Listen Labs

Design teams that test Figma prototypes

7. Maze

Diary studies and in-context mobile research

8. Dscout

Research operations that recruit your own customers

9. Great Question

Fast AI voice interviews that your team runs

10. Strella

In-product surveys

11. Sprig

One repository for interviews, calls, and tickets

12. Marvin

Unmoderated usability tests with unlimited seats

13. Ballpark

AI video interviews on iPhone and Android apps

14. Dialogue AI

Simple, low-cost surveys

15. SurveyMonkey

An agency that runs the study and delivers the report

16. Knit

Multi-market AI video interviews with an agency team

17. GetWhy

Consumer video feedback and showreels

18. Voxpopme

Brand teams that want fast answers from a US consumer panel

19. Suzy

Concept and pack tests for consumer brands in many countries

20. Bolt Insight

Three kinds of research companies

The platforms in this guide fall into three groups. The group tells you more than any feature list: it decides who does the work, what it costs as you grow, and whether your own systems can use it.

Services businesses

Most new AI research companies grow like agencies. Their staff set up and run your studies, and they price like consultants. The work stays inside the vendor, where your team and your AI agents cannot reach it.

Incumbents that add AI

Qualtrics, UserTesting, and SurveyMonkey built their products before large language models. Private equity firms own all three, and all three add AI on top of the old product.

Infrastructure

Voicepanel is infrastructure. The app, the REST API, and the MCP server share one public study schema, and you can edit every AI rule. Your team or your AI agents run the research, with our help when you want it.

How we compared

We compared all 20 platforms on 10 questions. We used each vendor's own documentation, pricing pages, job postings, and independent profiles, and we list the sources at the end.

Self-serve

Can your team run the research itself, or does the vendor run it for you?

API access

Can your code create and launch a study, or only read results?

MCP server

Can an AI agent such as Claude or ChatGPT build, launch, and analyze a study?

AI control

Can you read and edit the rules that drive the AI?

Methods

Can one study mix AI interviews, voice, video, desktop and phone screen recording, usability tasks, and survey questions?

Recruitment

Where do participants come from: a panel, your own customers, or your website?

Governance

Can a large team control who launches studies, what they cost, and how good they are?

Enterprise readiness

Does it pass a large company's security review: SOC 2 or ISO 27001, HIPAA, SSO, and user provisioning?

Support

How do you reach the vendor, and who answers?

Pricing

Can you buy one project as well as an annual contract, and do you pay for seats?

Comparison table

Full

Partial

Limited or vendor-run

Missing or unknown

Platform	Self-serve	API access	MCP server	AI control	Methods	Recruitment	Governance	Enterprise readiness	Support	Pricing
1. Voicepanel
2. Qualtrics
3. UserTesting
4. Outset
5. Conveo
6. Listen Labs
7. Maze
8. Dscout
9. Great Question
10. Strella
11. Sprig
12. Marvin
13. Ballpark
14. Dialogue AI
15. SurveyMonkey
16. Knit
17. GetWhy
18. Voxpopme
19. Suzy
20. Bolt Insight

Hover or tab to a square to read the details, or switch to the detailed view. Facts checked in September 2026.

The platforms

1. Voicepanel

Research infrastructure

Best for: Product builders, research teams, and AI agents that need the most compatible and flexible research platform.

Pricing: An Enterprise contract, which includes the REST API, the MCP server, and custom AI rules, or pay-as-you-go one-off projects on the Agent plan. No seat licenses, and recruiting credits pay panel fees and incentives.

Why Voicepanel? Because your research should plug into everything you build, and never wait in a vendor's queue.

Voicepanel is research infrastructure for people and agents. The app, the REST API, and the MCP server share one public study schema, so research fits into the tools and agents that your team already uses. With Grade, Voicepanel also takes research into the evals that product teams use to decide what ships.

Strengths

Your code and your AI agents run real research. The REST API runs the whole loop, and the MCP server gives Claude and ChatGPT 41 tools.
Full control of the AI, with rules that you read and edit for 13 AI stages.
Every method in one study: AI interviews, usability tasks on websites, prototypes, and mobile apps, and surveys.
Recruitment from 10+ panels, your own customers, or your website, in 37 languages, with no seat licenses.

Limits

No real-time visual intelligence: the AI moderator does not react to the participant's screen, as it does in Listen Labs, Outset, and Conveo (beta).
No dedicated researcher for every project.
Prices are quoted, not published.

Customers: Consumer brands and AI product teams

Nestlé
Poshmark
NETGEAR
Gamma
Ancestry

2. Qualtrics

Enterprise survey depth

Best for: Large enterprises that run complex survey programs and need government-grade compliance.

Pricing: Free plan with 500 responses. A self-serve research license costs $420 a month, billed annually. Enterprise contracts are quoted.

Why Qualtrics? Because it has the deepest survey toolset and the broadest compliance set in this guide.

Qualtrics is the enterprise survey leader. It has the deepest survey methods and the broadest compliance set in this comparison. Its AI push centres on Edge Audiences, a synthetic panel.

Strengths

The deepest survey toolset, including conjoint and MaxDiff, in 78 languages.
FedRAMP High, HITRUST, ISO 27001, and ISO 42001.
Project approvals and ExpertReview, an automated check before launch.

Limits

Surveys come first. AI adds text follow-ups, but there are no voice or video interviews.
The API costs extra on top of the license, and the status of its MCP server is unclear.
Plan a real budget: buyers report a median contract of about $30,000 a year.

Customers: Hospitality, healthcare, and government

Hilton
Shake Shack
Stanford Health Care
U.S. Department of State
Elevance Health

3. UserTesting

Usability testing at scale

Best for: Enterprises that run large usability programs with recorded and human-moderated sessions.

Pricing: Quoted annual subscriptions. No self-serve paid plan.

Why UserTesting? Because it covers the most usability methods, with a large panel of its own.

UserTesting has the widest set of usability methods and a large panel. It merged with UserZoom in 2023 and bought User Interviews in 2026. AI moderation is in early access.

Strengths

Native iOS and Android tests, prototypes, card sort, tree test, and live moderated sessions.
A 7M+ participant panel, plus the User Interviews network of professionals.
An MCP server that can create and launch tests and follows approval workflows (beta).

Limits

AI-moderated interviews are still in limited early access.
Your code can read results, but it cannot build or launch a test.
Only 4 participant languages, and buyers report a median contract of about $41,000 a year.

Customers: Financial services, telecom, and insurance

Wells Fargo
AWS
BT Group
Vitality

4. Outset

Vendor-run UX research

Best for: UX research teams that want to hand study setup, fieldwork, and reporting to the vendor's research staff.

Pricing: Quoted, based on "your research team, your research needs, and your support needs".

Why Outset? Because UX research teams can hand study setup and fieldwork to its research staff.

Outset runs AI-moderated interviews with strong agent access. It is a good fit when you want the vendor to do the research work: its pricing page sells guide programming, data collection, panel sourcing, and project management by the Outset team.

Strengths

An MCP server that creates studies, sets up recruitment, launches, and analyzes.
Visual Intelligence: the AI moderator watches a screen share during a usability task, then asks about the clicks, hesitation, and friction it saw.
25+ panel integrations and 40+ languages.
UX Evals for AI products, SOC 2 Type II, HIPAA, and ISO 42001.

Limits

AI agents can drive it through MCP, but there is no REST API for your own code.
Demo only, with no public prices, and billing is by questions answered.
It also sells Digital Twins, which are synthetic stand-ins for real respondents.

Customers: Software and internet UX teams, plus consumer goods

Microsoft
Nestlé
Indeed
HubSpot
Glassdoor

5. Conveo

Consumer concept research

Best for: Consumer brands that run concept, pricing, and diary research.

Pricing: Quoted, paid in credits. Outsourced research work goes to a marketplace of third-party agencies.

Why Conveo? Because it has deep concept and pricing methods, with strong spend controls.

Conveo has deep consumer research methods and the strongest spend controls of the AI-native group.

Strengths

Monadic concept tests, MaxDiff, Van Westendorp, Gabor-Granger, and diary studies.
Video analysis that detects emotions, brands, and products on camera, and a screen-watching moderator in private beta.
Spend approvals, launch approvals, and spend caps for partners.
An automated quality check on each topic guide.

Limits

Your code can read finished studies and an AI agent can build one, but neither can launch it.
Interviews are asynchronous only, with no live conversation.
Demo only, with no public prices.

Customers: Consumer goods (CPG), many in Europe

Canva
JDE Peet's
Georgia-Pacific
Edgard & Cooper

6. Listen Labs

Managed AI interviews

Best for: Enterprises that want a managed AI interview program and have a large budget.

Pricing: Quoted subscription plus credits. Contrary Research describes it as a managed engagement.

Why Listen Labs? Because it runs a managed AI interview program with strong agent access.

Listen Labs has one of the strongest AI interview products and a strong agent surface. It also runs as a services business: it staffs strategists and engagement managers, and it prices research like a consulting firm.

Strengths

An MCP server that creates, edits, and launches studies and starts recruitment.
Visual Insights: the AI interviewer watches a live screen share and asks follow-ups in real time when behavior contradicts what a participant said.
MaxDiff, emotion analysis, and 100+ languages.
ISO 27001, ISO 27701, and ISO 42001.

Limits

Budget for it: buyers report about $20,000 a year, plus $300 to $400 per interview.
It hires like a services firm. In September 2026, about 10 of 36 open roles were services roles, and 3 were engineering.
The REST API works only when you bring your own participants.

Customers: Tech and AI companies, plus consumer brands

Microsoft
Anthropic
Sweetgreen
Chubbies
Cognition

7. Maze

Prototype tests for designers

Best for: Product design teams that test Figma prototypes and live sites.

Pricing: Free plan, a Starter plan paid by card, and a quoted Enterprise plan. Panel credits cost $5 each.

Why Maze? Because design teams can test Figma prototypes and live sites in one tool.

Maze mixes prototype tests, surveys, card sorts, and AI-moderated interviews in one tool, with clear panel pricing. Its developer access is limited.

Strengths

AI-moderated voice interviews in 20 languages, and an AI voice block inside unmoderated studies.
Figma prototypes, live websites, native mobile tests, card sort, and tree test.
Transparent panel credits that include the participant incentive.

Limits

An AI agent can read your Maze data, but it cannot create or launch a study.
The AI moderator, mobile tests, and moderated interviews need the Enterprise plan.

Customers: Banks and software product teams

Itaú Unibanco
Braze
Uizard
HqO

8. Dscout

In-context mobile research

Best for: Enterprise UX teams that run many diary, mobile, and usability studies, with a budget of $20,000 or more a year.

Pricing: Quoted annual plans with flexible seats and activity credits. Buyers report a range of about $24,000 to $110,000 a year.

Why Dscout? Because it puts a vetted panel, diary studies, mobile usability tests, and AI interviews in one platform.

Dscout is a mature research platform from Chicago, founded in 2011, and customers run their own studies on it. In 2026 it added AI-moderated voice interviews in open beta and an MCP server that can build, recruit for, and launch studies. It sells custom annual subscriptions only.

Strengths

An official MCP server with 60+ tools that drafts a study, sets targeting and incentives, recruits from the Dscout panel, and launches.
Whole-screen phone recording through native iOS and Android apps, for up to 40 minutes and 15 tasks.
100K+ active, verified participants, 3M+ through partner panels, and private panels of your own customers.

Limits

The AI moderator is desktop-only, in open beta, and cannot see the screen during a task.
No public prices. Buyers report a median contract of about $49,000 a year, plus incentives.
To build a study from code, use MCP: the REST API only manages panel members and users.

Customers: Consumer tech, telecom, and retail

T-Mobile
Best Buy
Lenovo
Bose
Vodafone

9. Great Question

Research ops for enterprises

Best for: Research operations teams that let many people run studies with their own customers, under admin approvals and budgets.

Pricing: Self-serve costs $129 per seat a month, for up to 5 seats. Enterprise contracts are quoted, and buyers report a median of about $23,000 a year.

Why Great Question? Because large teams can recruit their own customers, run studies, and control spend in one platform.

Great Question is self-serve software that combines a CRM of your own participants, study methods, and an AI research repository. Its strengths are recruitment from your own customer data and controls for large teams. Its in-house AI moderator launched in August 2026 and is on the Enterprise plan only.

Strengths

Admins can require approval before a study gets funds or goes live, and each team can have its own incentive wallet.
An MCP server with 95+ tools that builds studies and screeners and invites your own customers.
Recruit your own customers from Salesforce, Snowflake, or a CSV file at no fee per participant, or use the User Interviews or Respondent panel.

Limits

Participants on a phone cannot share their screen, so you get audio and camera only.
The AI moderator, MCP, SSO, and a dedicated CSM need Enterprise, with at least 5 seats.
The HIPAA agreement does not cover recordings or transcripts.

Customers: B2B software and fintech

ServiceNow
Canva
Brex
Miro

10. Strella

Fast AI voice interviews

Best for: In-house research teams that want fast AI voice interviews.

Pricing: Quoted, per seat, with prepaid credits for recruitment.

Why Strella? Because in-house teams get a deep AI voice moderator that they run themselves.

Strella is software that customers run themselves, with a deep AI voice moderator and fast releases. Access goes through a demo.

Strengths

Voice interviews up to 90 minutes, with mobile screen sharing and live prototype tasks.
Moderator guidance for the whole study and probing instructions for each question.
46 languages, and a connector for Claude and ChatGPT.

Limits

Built for researchers, not for code: there is no public REST API.
The Claude and ChatGPT connector is for current customers only, with no public tool list.
No public prices, and recruitment goes through Respondent and User Interviews.

Customers: Consumer brands and tech; case studies are mid-market brands

Amazon
Chobani
Duolingo
Daily Harvest

11. Sprig

In-product surveys

Best for: Product teams that run in-product surveys on web and mobile.

Pricing: Quoted Enterprise contracts based on response volume.

Why Sprig? Because product teams can survey users inside their web and mobile apps.

Sprig moved from in-product surveys to an enterprise survey platform run by AI agents. It has good developer access for reading data, and clear controls on what agents can do.

Strengths

In-product survey SDKs for web, iOS, Android, React Native, and Flutter.
A public REST API for exports, and an MCP server that creates draft studies.
SOC 2 Type II, SAML SSO, and admin controls on agent access.

Limits

No live AI-moderated voice or video interviews.
An AI agent can draft a study, but it cannot launch one, and the REST API cannot create one.
Enterprise-only pricing since it dropped its public plans.

Customers: Consumer and B2B software

Notion
Thumbtack

12. Marvin

Research repository and analysis

Best for: Research teams that want one repository to analyze interviews, calls, surveys, and tickets.

Pricing: Free plan with 5 uploads a month. Starter, Pro, and Enterprise are quoted per seat.

Why Marvin? Because it brings interviews, calls, surveys, and tickets into one repository with cited AI answers.

Marvin (HeyMarvin) is a research repository with strong AI analysis. It added an AI interviewer in 2025 and a website intercept in beta. It is software that your team runs, with no managed research services, but it does not recruit participants.

Strengths

30+ integrations that bring calls, support tickets, and surveys into one repository, with cited AI answers.
Strong PII redaction, including face blur and voice changes, plus SOC 2 Type II, ISO 27001, ISO 42001, and HIPAA.
A free self-serve plan, and discussion-guide controls for probing depth and branching.

Limits

You bring the participants. Marvin says recruitment "is not currently available inside the platform".
No native survey builder yet. Mixed-method studies are on the roadmap.
Strong for analysis, weak for fieldwork: the API only imports data, and MCP cannot launch a study.

Customers: Software and tech research teams

Microsoft
VistaPrint
Criteo
Razorpay

13. Ballpark

Unmoderated usability tests

Best for: Product and design teams that want unmoderated usability tests with unlimited seats.

Pricing: Quoted Enterprise plans based on study volume, with a 14-day trial. A managed tier delivers research for you.

Why Ballpark? Because it gives unlimited seats and responses for unmoderated usability tests.

Ballpark is a London usability testing platform from the team behind Marvel. Your team builds and runs the studies, and it now adds AI voice interviews and an MCP server. Many of its AI features are in beta, and you buy it through sales.

Strengths

Surveys, card sorts, Figma prototypes, live websites, native mobile apps, and AI interview steps in one study.
Unlimited users and responses on Enterprise. Your own participants are free, and panel credits cost $1 to $4 per minute.
An MCP server that creates and edits studies, with publishing and recruitment in beta.

Limits

Much of the new AI work is still in beta or behind a feature flag.
The GraphQL API does not document how to launch a study or recruit.
No public prices, HIPAA is still in progress, and there is no ISO 27001 certificate.

Customers: UK and European telecom, finance, and tech

Vodafone
Coventry Building Society
Soldo
Trade Me

14. Dialogue AI

Mobile AI video interviews

Best for: Consumer product and insights teams that want fast AI-moderated video interviews, including mobile usability sessions, with recruitment included.

Pricing: Quoted annual contracts in three tiers, from a startup tier to a custom enterprise tier. There is no public price list, and the site shows only a demo form.

Why Dialogue AI? Because its AI moderator runs voice and video interviews in native iPhone and Android apps that record the screen.

Dialogue AI is a Los Angeles startup, founded in 2025 by former Nextdoor staff, with a $6M seed round led by Lightspeed. Its AI moderator plans studies, recruits, runs voice and video interviews, and writes the synthesis. Its own research team also runs full studies. It publishes little about its API, AI control, governance, or prices.

Strengths

Native iOS and Android participant apps record the screen, camera, and microphone while the AI moderator runs the interview.
A claimed global panel of 50M+ people, a Prolific partnership, and invites to your own customers.
Wayfair reports insights in 36 hours from 20 participants, and Kalshi ran 5+ studies in one week with one researcher.

Limits

Hard to judge before a call: no public prices, and the founders describe three annual tiers.
The site does not say how you edit the moderator's rules, or describe roles, approvals, or SSO.
The developer docs need a login, and there is no MCP server.

Customers: US consumer tech and fintech

Wayfair
Block
Nextdoor
Kalshi

15. SurveyMonkey

Simple, low-cost surveys

Best for: Teams that need simple, low-cost surveys without a sales call.

Pricing: Free plan. Paid plans from $99 a month billed monthly, or less billed annually. Team plans from $30 per user per month.

Why SurveyMonkey? Because you can start a survey today, for free, without a sales call.

SurveyMonkey is the easiest survey tool to start with, and it publishes its prices. It stays a survey tool: it has no voice or video answers and no usability testing.

Strengths

Public prices, a free plan, and full self-serve sign-up.
The Audience panel reaches 335M+ people in 130+ countries.
Claude and ChatGPT connectors that create and send surveys on all plans.

Limits

It stays a survey tool: no AI interviews, voice or video answers, or real-time probing.
No usability, prototype, or screen-recording tests.
Need HIPAA? It is an Enterprise add-on, and the AI features turn off.

Customers: Mid-market companies, sports, travel, and events

Golden State Warriors
Greyhound
Noom
Hornblower

16. Knit

AI-native research agency

Best for: Brand and insights teams that want an agency to deliver a finished report.

Pricing: Quoted annual subscription with a respondent bank, or quoted per project. No public prices.

Why Knit? Because a dedicated researcher runs each study with AI and delivers a finished report.

Knit calls itself "The AI-Native Research Agency". A dedicated Knit researcher runs each project with AI: Knit programs the survey, fields it, and delivers a branded report. Clients can review the work in the Knit platform.

Strengths

Quantitative surveys and AI-moderated video questions with the same respondents in one study.
65M+ respondents through 15 named panel suppliers, in 50+ languages.
SOC 2 Type 2, HIPAA, GDPR, and COPPA, with a senior researcher on every account.

Limits

It works like an agency, so there is no self-serve plan, API, or MCP server.
Expect to wait: Knit's own comparison puts time to insights at 1 to 2 weeks.
No usability, prototype, or screen-recording tests.

Customers: Consumer brands and trade associations

Mars Wrigley
NASCAR
JBL
Timex

17. GetWhy

Enterprise AI video interviews

Best for: Enterprise consumer insights teams that test brands, concepts, and ads in many markets and want agency support.

Pricing: No price list, and sales go through a demo. GetWhy says a study typically costs $2,000 to $4,000, on annual contracts invoiced in advance.

Why GetWhy? Because it runs AI-moderated video interviews in 100+ languages and can add a senior research team to do the work.

GetWhy started in Copenhagen in 2011 as UserTribe and took its current name in 2024. It sells AI-moderated video interviews with managed recruitment, and it markets a team of 70+ senior researchers who can run the work. You can run studies yourself, but only after a demo.

Strengths

The AI moderator runs video, voice, or text interviews in 100+ languages.
A network of panel vendors that GetWhy says reaches 300M+ people in 100+ markets, plus your own panel.
An audit by Nucleus Research reports that eBay got insights in 3 working days instead of 15.

Limits

Contracts run for a year, renew automatically, and are invoiced up front.
No documented usability tasks, survey blocks, or phone screen share.
It runs research for you, so there is no public API or MCP server to plug into.

Customers: Global consumer brands (CPG), many in Europe

eBay
Heineken
L'Oréal
Arla Foods
Maersk

18. Voxpopme

Video feedback and showreels

Best for: Consumer brands that want video feedback, showreels, and a partner to deliver projects.

Pricing: Quoted annual subscription. Project delivery, advisors, and full-service research cost extra.

Why Voxpopme? Because it turns consumer video answers into showreels, with a services team to help.

Voxpopme started in 2013 as a video survey tool and now sells itself as a customer intelligence platform. It adds an English-only AI moderator, the Compass research agent, and an early-access MCP server. It is sold on annual contracts, and a large services team can run projects for you.

Strengths

Video questions, AI and live interviews, screen tasks, and closed questions in one project.
A widget that embeds video questions in Qualtrics, Decipher, and Confirmit surveys.
ISO 27001, ISO 42001, and ISO 9001, plus more than 20 automated checks on participant quality.

Limits

The AI moderator speaks English only and shows images only.
The Voxpopme team can review and approve your project before it launches.
Developer access is gated: the API docs need a key, and MCP needs an onboarding call.

Customers: Consumer goods (CPG), energy, and travel

Mars
Shell
Philips
Alaska Airlines
Liberty Mutual

19. Suzy

Fast consumer panel research

Best for: Consumer brand and marketing teams that want fast answers from a US consumer panel.

Pricing: Quoted annual contracts paid in credits. The public pricing page was removed in 2026.

Why Suzy? Because brand teams get fast answers from its own US consumer panel, with quant and qual in one place.

Suzy is a consumer research platform built on its own US panel. In April 2026 it relaunched as a "decision engine" for marketing, with AI agents that search past studies and draft deliverables. It offers a self-serve interface, but its AI-moderated interviews were sold as fully managed by Suzy staff.

Strengths

Surveys, MaxDiff, TURF, AI voice interviews, live interviews, and in-home usage tests in one platform.
An owned US panel of 1M+ members, partner panels in 130+ markets, and bot detection.
SOC 2 Type II, ISO 27001, ISO 27701, and ISO 42001.

Limits

The highest reported price in this guide: a median contract of about $99,000 a year.
AI interviews run in 8 markets, and you cannot see or edit the AI prompts.
A closed platform, with no public API or official MCP server.

Customers: US retail and consumer goods (CPG)

Walmart
PepsiCo
Mondelēz
Estée Lauder
DoorDash

20. Bolt Insight

Global CPG qual at scale

Best for: Large consumer goods insights teams that test concepts, packs, and ads in many countries at once.

Pricing: Quoted credit plans with no public price list. The price is the same in every market.

Why Bolt Insight? Because it pairs AI-moderated interviews in 50+ languages with an in-house research team.

Bolt Insight is a London AI research platform, founded in 2019, with a large in-house research team. The team co-designs studies, checks quality, and writes reports, and customers can also run projects themselves with credits. It still sells itself as full-service.

Strengths

The vendor says it has interviewed 5M+ consumers in 90 markets, for customers such as Unilever, Danone, and Reckitt.
One UX study can combine website or Figma tasks with screen and webcam recording, and the AI checks each recording for task completion.
Video emotion analysis flags gaps between what participants say and what they do.

Limits

Sessions stop at 30 minutes and 20 questions, with at least 10 participants a project.
Participants chat in a mobile browser, with no native app or phone screen share.
Nothing documented for developers: no public API docs or MCP server.

Customers: Global consumer goods (CPG)

Unilever
Danone
Reckitt
Haleon
Vodafone

Why teams choose Voicepanel

Services businesses grow by adding people to each project. Voicepanel grows by giving your people and your AI agents direct access to research with real people. These are the differences that follow from that choice.

Trusted by enterprise teams

Nestlé, Poshmark, NETGEAR, Gamma, and Ancestry run their research on Voicepanel. Ancestry reports research 10 times faster than with its survey vendors, at one third of the budget, across 100+ completed projects.

Read the Ancestry case study →

Your AI agents can run the research

Connect Claude, ChatGPT, or your own code, and it can build a study, recruit real people, and analyze the answers. The REST API runs the whole loop, and the MCP server has 41 tools. Both use one public study schema. We found no other platform in this guide that publishes its full schema. The REST API and the MCP server come with an Enterprise contract.

Read the study reference →

You can see and change the AI

Most platforms let you give the AI moderator a few instructions. With Voicepanel, you read and rewrite the actual rules for 13 AI stages, from study design and screeners to probing and analysis. For most stages, you can test a change on sample data before you save it. Custom AI rules come with an Enterprise contract.

Any method, any participant

Mix AI conversation, voice answers, rating questions, and usability tasks in one study. Test websites, prototypes, mobile apps, or physical products, and record the screen or an iPhone screen. Then recruit from 10+ panel partners, your own customers, or your website visitors, in 37 languages.

Grade: research where product decisions happen

Teams that build AI products decide what ships with evals, not slide decks. Grade takes research into that loop. It is an expert judge, calibrated against real feedback on your AI product, that scores every new output. Grade is in early access.

Learn about Grade →

Ready for large teams, with answers in Slack

Set roles, require approval before a study goes live, and cap the cost of each study. On Enterprise, an AI reviewer checks every plan against your own research rules, and you share a Slack channel with the engineers who build Voicepanel. It is SOC 2 Type II, HIPAA, and GDPR compliant, and it does not train on your data.

Where other platforms lead

No platform wins every row. If one of these points decides your purchase, look closely at the platform that leads it.

Area	Leaders	Detail
Visual intelligence	Listen LabsOutsetConveo	The AI moderator reads what is on the participant's screen. Listen Labs asks follow-ups in real time when behavior contradicts what a participant said. Outset asks about clicks, hesitation, and friction after each task. Conveo's version is in private beta.
Emotion detection	OutsetListen LabsConveoDscoutBolt Insight	The AI reads facial expressions or voice cues and tags emotions in the recording. Conveo also detects brands and products on camera. The Dscout version is in beta.
Fully managed research	KnitOutsetListen LabsGetWhyBolt Insight	A dedicated researcher or project team runs each study for you. Knit assigns a researcher to every project at no extra cost. GetWhy markets 70+ senior researchers.
Languages	Listen LabsGetWhyQualtricsSurveyMonkey	Listen Labs and GetWhy support 100+ languages, Qualtrics 78, and SurveyMonkey 56.
Compliance breadth	QualtricsSuzyListen LabsOutsetMarvinVoxpopme	Qualtrics holds FedRAMP High and HITRUST. All six hold ISO 42001 for AI management.
Survey depth	QualtricsConveo	Qualtrics has conjoint and MaxDiff. Conveo has Van Westendorp and Gabor-Granger pricing questions.
Usability breadth	UserTestingMaze	Card sort, tree test, and human-moderated sessions.
Spend approvals	ConveoGreat Question	An owner or admin approves spend before a study launches. Great Question also gives each team its own incentive wallet.
Panel size	SurveyMonkeyQualtrics	SurveyMonkey reaches 335M+ people and Qualtrics 200M+.
Agents that launch panel studies	Listen LabsOutsetDscout	Their MCP servers can launch a study and start panel recruitment from a chat.
Public prices	SurveyMonkeyQualtrics	Both publish prices and offer a free plan.
Mobile screen share on iOS and Android	UserTestingMazeBallparkOutsetDscoutDialogue AI	Participants record the phone screen through the vendor's app. On Ballpark, Outset, and Dialogue AI, the AI moderator works with the mobile session.
Diary studies	Dscout	Participants record moments over days or weeks from their phones, with a vetted panel of 100K+ people.
Research repository	Marvin	30+ integrations bring calls, tickets, and surveys into one repository with cited AI answers.

Synthetic respondents versus a calibrated judge

Several vendors now sell AI models that answer in place of people. Qualtrics sells Edge Audiences, Outset sells Digital Twins, Knit sells Digital Personas in beta, and Conveo offers study personas that it labels as simulated.

Digital twins and synthetic personas both simulate the respondent. A digital twin predicts what a person would say from past data, so it repeats the past and cannot react to something new. A synthetic persona asks a language model to act as a type of customer, so it gives the model's average view of that group, not the views of your users. Neither has a ground truth, so nothing tells you when a simulated answer is wrong.

Grade does not simulate respondents. It is an expert judge that scores the outputs of your AI product against a rubric built from real user feedback. Grade is provably calibrated: we measure its scores against the judgments of real users, so you know how far to trust each score.

We manage the calibration for you. When your product or your users change, we run new research with real people and recalibrate the judge. Real people stay the source of truth, and the judge applies their standard to thousands of scenarios.

Learn about Grade →

Frequently asked questions

What is the best AI research platform in 2026?

Voicepanel is the best overall choice, and the most compatible and flexible platform for product builders and AI agents. It combines AI-moderated interviews, usability tasks, and surveys in one study, with a public schema, editable AI rules, a REST API, and an MCP server. For survey-only programs in regulated industries, Qualtrics is strong. For UX research teams that want the vendor to run studies, Outset offers managed services.

Which enterprises use Voicepanel?

Enterprise teams at Nestlé, Poshmark, NETGEAR, Gamma, and Ancestry run research on Voicepanel. Ancestry reports research 10 times faster than with its survey vendors, at one third of the budget, across 100+ completed projects.

Which AI research platforms have an MCP server?

Voicepanel, Listen Labs, Outset, and Dscout have official MCP servers that can launch studies and recruit. UserTesting (beta), Ballpark (publish in beta), Voxpopme (early access), Strella (customers only), Great Question (no launch), Maze (read-only), Conveo (no launch), Marvin (no launch), and Sprig (drafts only) have limited MCP servers. SurveyMonkey has Claude and ChatGPT connectors. The status of the Qualtrics MCP server is unclear. We found no MCP server for Suzy, Knit, GetWhy, Bolt Insight, or Dialogue AI.

Which AI research platforms have a public REST API?

Voicepanel, Qualtrics, SurveyMonkey, UserTesting, Listen Labs, Conveo, and Sprig have public REST APIs. The UserTesting, Conveo, and Sprig APIs only read data. The Listen Labs API can launch studies only with your own participants. Voicepanel's API can also recruit a panel and publish the study. Ballpark has a GraphQL API, Voxpopme's API docs need a key, and Marvin's API only imports data. Great Question has a beta API that creates studies but cannot launch them, and the Dscout API only manages private panels. Maze, Outset, Strella, Suzy, Knit, GetWhy, Bolt Insight, and Dialogue AI have no public API.

Can I customize the prompts of an AI interviewer?

Most AI interview platforms let you add instructions for the moderator, such as probing depth or topics to avoid. Voicepanel goes further: on an Enterprise contract, you can read and replace the rules for 13 AI stages, from study creation and screening to probing, conversation, and analysis.

Are AI-native research companies software or agencies?

It varies. Knit calls itself "The AI-Native Research Agency", and a Knit researcher runs every project. Outset sells guide programming, data collection, and project management on its pricing page. Listen Labs staffs strategists and prices like a consulting firm, according to Contrary Research. GetWhy and Bolt Insight sell software with large in-house research teams, and the Dialogue AI research team also runs full studies. Conveo, Strella, and Sprig are software, but all three sell through demos. Voicepanel is infrastructure that your team and your agents run, with support from the team that builds it.

Do synthetic respondents replace real people?

No. Digital twins and synthetic personas simulate respondents from past data or a language model's general knowledge, and nothing checks their answers against your users. They can help you draft a study, but they cannot tell you something new about real customers. Grade takes a different approach: it is an expert judge for AI products, provably calibrated against real user feedback, and Voicepanel manages the calibration for you.

What support does Voicepanel offer?

Every plan includes email support. Enterprise customers also get a shared Slack channel with the team that builds Voicepanel, hands-on onboarding, and priority input on the roadmap.

How much does Voicepanel cost?

Voicepanel is flexible. Buy one project at a time on the pay-as-you-go Agent plan, or sign an Enterprise contract for full platform access, including the REST API, the MCP server, and custom AI rules. There are no seat licenses, and recruiting credits pay panel fees and incentives directly.