An AI video research platform is software that conducts moderated video customer interviews with customers using an AI moderator, with optional screen sharing for live URLs, Figma prototypes, and design mockups. The category settled on eight serious 2026 contenders, each making a different architectural bet on how to combine AI moderation with video and screen-share evidence.
If you’re shortlisting AI video research platforms in 2026, you have eight serious contenders to weigh: User Intuition, Conveo, Outset, Listen Labs, Voxpopme, Maze, Strella, and HeyMarvin. Each makes a different architectural bet — adaptive laddering depth, multimodal signal extraction, researcher-configured probing, full research lifecycle, video-first incumbency, prototype-first infrastructure, lean AI moderation, or AI-native customer insights breadth. This guide evaluates all eight on the criteria that actually decide the purchase, so you can match the platform to the research deliverable instead of to brand recognition.
What are the best AI video research platforms in 2026?
The category settled on roughly eight serious platforms by 2026, each with a different theory of how to combine AI moderation with video and screen-share evidence:
- User Intuition — Adaptive 5-7 layer laddering on every video + screen-share interview, $150/study, 4M+ panel
- Conveo — Multimodal voice + video + tone + facial extraction, live Figma prototype embeds, 8+ partner panels by its count
- Outset — Self-paced, live AI-moderated video/voice/text interviews with real-time follow-ups across 40+ languages
- Listen Labs — Full research lifecycle with fraud detection and multimodal emotional analysis
- Voxpopme — Qualitative video research incumbent with diaries, interviews, and showreels
- Maze — Live website and prototype testing infrastructure with AI moderator added
- Strella — AI-moderated interviews with adaptive probing for lean teams
- HeyMarvin — AI-native customer insights platform supporting 40+ languages
Pick by research deliverable, not by feature checklist. The buyer guide below evaluates each on identical criteria.
How we evaluated platforms
Six criteria decide most AI video research purchases. We applied them consistently across all eight platforms in this guide.
Most buyers ask the wrong opening question. They start with “which platform has the most features” and end up with a platform that does many things shallowly. The better starting question is “what does my research deliverable need to be true,” then work backward to the architectural fit. Concept testing with screen sharing rewards deep laddering plus synchronized video and mouse capture. Multimodal signal research rewards platforms that extract voice, tone, and facial reaction together. Live website testing at high volume rewards click-pattern density over interview depth. Multilingual qualitative at scale rewards owned panels with fraud screening built in. Pricing transparency matters more than headline price — a flat $30-per-interview rate with no monthly fee is structurally cheaper than a $45,000 annual enterprise-tier floor for teams running fewer than ten studies. The platforms differ enough that the wrong fit costs more than the price tag suggests.
The six criteria we used:
- Screen-sharing depth — Does the platform support live URLs, Figma prototypes, hosted mockups, and any web-accessible asset? Or only specific formats?
- Video capture fidelity — Face video and click activity synchronized to the transcript by default, extending to mouse movement and scroll behavior with a lightweight snippet? Or video clip aggregation only?
- AI moderation depth — Adaptive 5-7 layer laddering on every interview? Or shallower probing with AI added on top of a survey-style core?
- Panel access — Owned vetted panel with fraud screening built in, partner-network access, or bring-your-own recruitment expected?
- Pricing transparency — Per-study self-serve with public pricing, per-seat monthly licensing, or enterprise quote-based?
- Async vs synchronous flow — Concurrent hundreds-per-week throughput, sequential scheduling, or hybrid?
Every section below follows the same template: strengths, weaknesses, ideal customer, pricing summary, screen-share specifics, video capability specifics. Each platform gets fair treatment — this is a buyer’s guide, not a sales pitch.
User Intuition
Strengths: Adaptive 5-7 layer laddering on every interview — the deepest documented AI moderation methodology in the category. Synchronized capture of face video and click activity by default, tied directly to the verbatim transcript with a replayable clip per session, extending to mouse position and scroll behavior once the one-minute User Intuition snippet is added to the page. Three-level fraud defense: a vetted panel tracked across studies, per-interview screening of each transcript for AI use (on video, confirming a real, identifiable person is on screen and engaged), and laddered depth that is hard to fake. Owned 4M+ pre-vetted panel across 80+ languages with bring-your-own customer support in the same study. Pay-only-for-high-quality conversations commitment. Studies start at $150 with public pricing. 4.9/5 on G2 and 5/5 on Capterra. Customer Intelligence Hub indexes every interview into queryable knowledge that compounds across studies.
Voice interviews cost $30 and video interviews cost $60 on Starter, with recruitment additional. Choose the modality based on the evidence needed and inspect the pricing page before estimating the project.
Ideal customer: Product, design, research, and CX teams running concept testing, prototype testing, win-loss, churn, and broad qual-at-quant-scale work. Especially fit for teams that need 100-300 panel interviews in 24 hours and want a knowledge layer that compounds across studies rather than per-study PDF exports.
Pricing: Starter has no monthly fee and includes 3 free interviews on signup, no card required. User Intuition voice interviews cost $30 each with no subscription required. Recruitment is priced separately. Studies from $150. Public pricing page; no annual commitment required. See /platform/in-depth-interviews/ for the screen-share modality detail.
Screen-share specifics: Live URL, Figma prototype URL, hosted design mockup, JPEG/PNG concept board, marketing landing page, app prototype — any web-accessible asset, no developer integration required to start. The participant interacts inside the interview while the AI probes pause points, click logic, and what the page is for in their words. On a live URL, Figma Make prototype, or any asset the team controls, add the one-minute User Intuition snippet and that same probing extends to scroll behavior and element-level clicks (classic Figma prototypes run on Figma’s own player and stay at the click-and-page-visibility tier).
Video capability specifics: Face video, click and page-visibility activity, and verbatim transcript captured together by default, synchronized to a single replayable clip per interview — extending to mouse position and scroll behavior once the one-minute snippet is added. Asynchronous concurrent flow — hundreds of sessions run in parallel 24/7 across timezones. The methodology overview lives at /platform/in-depth-interviews/.
Conveo
Strengths: Distinctive multimodal video signal extraction architecture — voice, video, tone, facial expressions, and emotional nuance all extracted as theme synthesis sources. Live Figma prototypes embed as interview stimuli for Figma-heavy product teams. Conveo says it works with 8+ panel providers: Respondent and User Interviews are API-integrated, and the rest (such as Norstat, Bilendi, Sago, Rakuten, DataDiggers, and NewtonX) quote per project. Conveo emphasizes research rigor; its head of research is a former ESOMAR president, and it supports MaxDiff and pricing research inside AI interviews. Its $50M Series A in September 2026, bringing total funding to $55.8M, is earmarked for U.S. expansion, per Conveo. AI-moderated follow-up across 50+ languages, by Conveo’s count.
Weaknesses: In Conveo the researcher sets follow-up depth per question (automatic probing of up to 5 follow-ups, or a fixed range), while User Intuition runs 5-7 level laddering by default in every conversation; the architectural bet is signal breadth (extract more from one session) rather than methodological depth (probe deeper across more sessions). Pricing requires sales conversation — both Pay & Go and Enterprise tiers go through a demo rather than self-serve signup. The Enterprise plan starts at $45,000/year, prepaid (per Conveo’s G2 listing, vendor-supplied, March 2026), with recruitment billed separately.
Ideal customer: UX teams with heavy Figma workflows that want live prototypes embedded as interview stimuli. Insights teams whose research deliverable depends on facial expression and tonal signal extraction more than verbal motivational depth. Enterprise consumer-insights and brand teams that want MaxDiff or pricing research inside AI interviews. Teams with panel needs in geographies served by Conveo’s partner network.
Pricing: Two editions on Conveo’s G2 listing (vendor-supplied, March 2026) — Pay & Go credits for agencies and smaller teams, plus an Enterprise plan prepaid annually from $45,000/year; credits are charged per interview minute, with recruitment billed separately. Conveo publishes no pricing for its interview platform on conveo.ai, and pricing may have changed since its September 2026 Series A, so get a current quote. No self-serve sign-up, and its G2 listing shows no free trial. See /compare/conveo-vs-user-intuition/ for the architectural side-by-side.
Screen-share specifics: Live Figma prototype embeds are the differentiator — the participant clicks through a Figma prototype embedded as an interview stimulus, and the AI can probe on what it sees on a shared screen. Screen sharing (a single window or the entire screen) covers other web-accessible assets.
Video capability specifics: Voice or video interviews that participants take on their own time (Conveo calls them asynchronous), with the AI asking follow-ups after each answer and the researcher setting the camera to required, optional, or off; multimodal signal extraction is layered on top — facial reaction, tonal shift, voice, and verbal response synthesized together. The architectural bet is wider signal capture per session rather than deeper laddering across sessions.
Outset
Strengths: Multimodal video, voice, and text moderation in any combination, across 40+ languages — useful for global insights teams running mixed-modality work. Its $30M Series B in December 2025, led by Radical Ventures and bringing total funding to $51M, signals capital to extend the platform. AI-moderated follow-up adapts across modalities within the same study. Strong fit for teams that want one platform spanning text-based diary studies through voice and video moderated interviews.
Weaknesses: Outset’s AI moderator asks real-time follow-ups inside a researcher-written guide, and the researcher sets probing depth per question (up to 10 follow-ups in Abyss mode), while User Intuition runs 5-7 level laddering by default in every conversation. Outset recruits through 25+ third-party panel integrations, such as Prolific and User Interviews, rather than a first-party panel like User Intuition’s 4M+ panel. Subscription pricing is sales-led without public per-study transparency.
Ideal customer: Large-scale insights teams running global research where modality flexibility (text + voice + video in one platform) matters more than deepest laddering on a single modality.
Pricing: Subscription, sales-led — verify current pricing on outset.ai. No public per-study price. See /compare/outset-vs-user-intuition/ for the side-by-side detail.
Screen-share specifics: Picture-in-picture screen-share usability studies on Figma files, clickable prototypes, and live websites and mobile apps, with the moderator probing on clicks, hovers, and navigation paths it observes. Less prototype-specific tooling than Conveo or Maze; closer to Listen Labs in the multimodal-research category.
Video capability specifics: Researcher-configured AI probing across 40+ languages, with Visual Intelligence reading screens, facial expressions and tone of voice, and physical behavior to shape follow-up questions. Strong on multilingual reach. Buyers who need synchronized mouse + face + transcript clips should confirm that output in an Outset demo; User Intuition’s snippet-enabled tier documents it.
Listen Labs
Strengths: Full-funnel research lifecycle in one platform — recruitment, moderation, fraud detection, multimodal emotional analysis, and synthesis. Covers video, voice, and text in one workflow. Multimodal emotional analysis is a documented architectural feature rather than a roadmap item. Strong fit for insights teams managing mixed-methods research where consolidating tools matters.
Weaknesses: Listen Labs publishes no per-study or per-seat price — organizations over 100 employees buy through a demo and pilot, and smaller teams can start through a self-serve sign-up — so budget scoping requires a quote. The full-lifecycle bet puts the value in integration breadth; on moderation, the researcher sets follow-up depth per question (up to 2-3 follow-ups in the editor) and can add probing instructions, while User Intuition runs 5-7 level laddering by default in every conversation.
Ideal customer: Insights teams with mixed-methods research and an explicit goal of consolidating onto one vendor. Teams whose research needs span recruitment fraud detection through emotional signal analysis through multimodal synthesis, where switching costs across point tools outweigh per-tool depth.
Pricing: Not published. Listen Labs describes a subscription with credits spent per participant, costing more for harder-to-reach audiences; get a current quote from listenlabs.com. See /compare/listen-labs-vs-user-intuition/ for the architectural comparison.
Screen-share specifics: Optional screen recording with text, voice, or video interviews. Since June 2026 the interviewer can watch a participant’s shared desktop screen, ask follow-ups about what they do, and redirect off-task sessions; iOS mobile screen recording rolled out from January 2026.
Video capability specifics: Multimodal emotional analysis layered on top of video capture — emotional state inference is a documented capability. Less specific public detail on synchronized mouse + scroll behavior data relative to User Intuition or Maze.
Voxpopme
Strengths: The qualitative video research incumbent. Long-running platform with established Fortune 500 customer logos including McDonald’s and Microsoft. Video diary studies and editable showreels are mature features. AI Moderator added on top of the video-survey core covers async voice and video moderation. ChatGPT-powered theme aggregation works well for fast video clip synthesis. Strong fit for teams whose research deliverable is video evidence for stakeholders.
Weaknesses: The architectural origin is video survey aggregation, not AI-native moderation — the AI Moderator feature was added on top rather than designed as the primary research instrument. Per-user licensing at $199-$499/month scales cost with team size before any research runs. Ladering depth on the AI Moderator product is shallower than the native AI-first cohort.
Ideal customer: Qualitative research teams scaling video studies with stakeholder showreels as a primary deliverable. Organizations with multi-user research teams (5+ concurrent users) where per-seat licensing economics work. Teams whose research is more about visual evidence aggregation than motivational depth.
Pricing: Per-user licensing at $199-$499/user/month per public references. Five-person team approximately $12,000-$30,000 annually before any research runs. Verify current pricing on voxpopme.com. See /compare/voxpopme-vs-user-intuition/ for the format-difference detail.
Screen-share specifics: Standard screen-share within the moderated flow. Less prototype-specific tooling than Conveo or Maze.
Video capability specifics: Asynchronous video survey responses with editable showreels and ChatGPT analysis is the core competency. Replayable showreel output is the strength; synchronized mouse + face + transcript clip per interview is shallower than User Intuition’s snippet-enabled tier.
Maze
Strengths: Deep prototype testing infrastructure built before AI moderation became table stakes. Live website testing, click-pattern density data, heatmaps, and Figma integration all mature features. AI moderator added on top covers post-test follow-up. Free tier exists for small teams. Strong fit for product and UX teams running unmoderated tests at high volume with optional moderated follow-up.
Weaknesses: Maze’s architectural bet is unmoderated testing infrastructure with AI added on top — the AI moderator is a feature, not the primary research instrument. Bring-your-own recruitment is generally expected; no large owned panel like User Intuition’s. Methodological depth on moderated sessions is shallower than the native AI-first cohort.
Ideal customer: UX and product teams running unmoderated prototype tests at high volume with AI-moderated follow-up as a complement. Teams with strong existing recruitment pipelines (CRM customer lists, internal user panels). Organizations using Figma as the primary design source of truth.
Pricing: Subscription with a free tier scaling to enterprise. Public pricing tiers on maze.co. Verify current numbers before commitment. See /compare/maze-vs-user-intuition/ for the comparison detail.
Screen-share specifics: Strongest in the category for click-pattern density on Figma prototypes and live URLs — heatmaps, click counts, and unmoderated behavior data are core competencies.
Video capability specifics: Video capture is supported but the bet is behavior-data fidelity (clicks, paths, time-on-task) more than synchronized face + transcript moderated video. Different deliverable than User Intuition’s synchronized clip output.
Strella
Strengths: AI-moderated interviews with adaptive probing for lean research teams, run as real-time voice-to-voice conversations with optional webcam and desktop screen recording. AI-guided project setup, real-time synthesis with highlight reels, and a research repository its AI Chat can query across studies. Strong fit for small product, marketing, or research teams that want moderated interview depth and a research-partner relationship; Strella Advisory adds a full-service option.
Weaknesses: Strella recruits through integrated partner panels (it names Respondent and User Interviews), which it markets as access to up to 8M participants, plus bring-your-own participants; User Intuition recruits from its own 4M+ panel. Pricing is sales-led: no public sign-up or price list. Strella’s moderator asks dynamic follow-ups, and researchers set a project-wide default follow-up depth, a Strict or Adaptive wording mode, and moderator guidance, while User Intuition runs 5-7 level laddering by default in every conversation.
Ideal customer: Lean research teams (1-3 researchers) and small CX or product teams that want voice-first moderated interviews on a quote-based annual plan, recruiting from partner panels or their own customers. Teams that want a research partner can add Strella Advisory.
Pricing: Quote-based annual plans that Strella describes as usage-based, with no per-seat charges; strella.io publishes no prices, so get a current quote. See /compare/strella-vs-user-intuition/ for the architectural comparison.
Screen-share specifics: Desktop screen recording, switched on by the researcher, with usability heatmaps, task analytics, automatic pass/fail task grading (June 2026), and card sort studies (July 2026).
Video capability specifics: Voice-to-voice AI-moderated interviews with optional webcam recording set by the researcher. Synthesis output covers themes, verbatim quotes, highlight reels, and automatic coding of open-ended answers (since August 2026). Buyers who need synchronized mouse + scroll + transcript clips should confirm that output in a Strella demo; User Intuition’s snippet-enabled tier documents it.
HeyMarvin
Strengths: AI-native customer insights breadth — covers research repository, transcript analysis, AI-assisted synthesis, and AI moderation across 40+ languages. Strong fit for insights teams with broad research scope where consolidating onto one repository platform matters. AI Moderator and AI Insights features cover the moderation-plus-synthesis layer.
Weaknesses: HeyMarvin’s center of gravity is repository and synthesis more than category-leading moderated interview depth. Laddering depth on AI Moderator appears competitive with the AI-native cohort but not category-leading. Pricing is subscription-based with limited public per-study transparency.
Ideal customer: Insights teams managing broad research scope (repository + analysis + synthesis + moderation) where consolidating onto one platform outweighs per-tool depth. Teams that already have a research repository need and view AI moderation as one feature among many.
Pricing: Subscription, sales-led — verify current pricing on heymarvin.com.
Screen-share specifics: Standard screen-share within the moderated flow. Less prototype-specific tooling than Maze or Conveo.
Video capability specifics: AI Moderator covers video and audio. Synchronized mouse + scroll + transcript clip detail varies. Repository and synthesis are stronger than per-interview synchronized capture relative to User Intuition.
Decision matrix: which to choose for which job
The same platform rarely wins every job. Match the use case to the architectural fit.
| Use case | Primary recommendation | Strong alternative |
|---|---|---|
| Concept testing with screen sharing | User Intuition | Conveo (Figma-heavy workflows) |
| Prototype testing (Figma) | Conveo (Figma prototype embeds) or Maze (click density) | User Intuition (laddering depth) |
| Live website testing | User Intuition (depth) or Maze (behavior data) | Listen Labs (multimodal emotional) |
| Win-loss research | User Intuition | Strella (lean teams) |
| Churn motivation research | User Intuition | Listen Labs (multimodal emotional) |
| Broad qual at quant scale | User Intuition | Conveo (multimodal extraction) |
| Async video diaries with showreels | Voxpopme | Outset (multilingual reach) |
| Multilingual research at depth | User Intuition (80+ languages, owned panel) | Conveo (50+, per Conveo), Outset, HeyMarvin (40+) |
| Multi-modal mixed methods | Outset, Listen Labs, HeyMarvin | User Intuition (depth on each modality) |
| Lean-team moderated interviews | Strella, User Intuition (no monthly fee) | HeyMarvin |
| Unmoderated prototype tests | Maze | User Intuition (moderated complement) |
The recommendation column reflects the platform whose architectural bet aligns most closely with the use case. The strong alternative column captures the second-best fit when team-specific constraints (existing tools, panel relationships, budget structure) favor a different platform.
Common buyer questions: what’s in this category and what isn’t?
A handful of questions come up in nearly every shortlist conversation. The honest answers below.
What about HireVue and Mercor — should they be on this list? No. Both are video interview platforms for hiring and candidate screening. They evaluate job applicants. Customer research platforms test products, prototypes, websites, and concepts with real customers or panel members. Different buyer, different methodology, different category. Don’t shortlist HireVue or Mercor for product, design, or market research work.
What about Otter or Fireflies — are they competitors? No. Otter and Fireflies are meeting transcription and note-taking tools. They transcribe and summarize meetings; they don’t moderate research interviews, recruit participants, or synthesize cross-study insights. They sit alongside research platforms (transcribing your team’s internal calls), not in competition with them.
Do I actually need video specifically? Sometimes. Video matters when the research deliverable depends on facial reaction, on-screen behavior, or visual evidence (concept testing with mockups, prototype walkthroughs, design validation). Video adds noise when the research question is fundamentally about motivation and audio-only laddering reaches the same depth at half the cost. User Intuition’s voice Starter pricing ($30/interview equivalent) is structurally cheaper than its video-included sessions ($60/interview), so default to audio when video isn’t load-bearing.
Should I run a paid pilot before committing? Yes. Most platforms offer a paid first study (User Intuition’s Starter plan has no monthly fee and includes 3 free interviews on signup, no card; others typically require a sales conversation). Run identical research questions across two platforms; compare the synthesis output and the depth of evidence per insight. The architectural difference between platforms is invisible in feature comparisons and obvious in the side-by-side output.
What about platforms not on this list? A handful of adjacent platforms (Genway, dscout, Discuss, Great Question) appear on some shortlists. Genway and Strella sit close to each other architecturally; dscout is closer to Voxpopme on the video-diary axis; Discuss and Great Question are repository-plus-services platforms with AI features added rather than AI-native moderation. The eight platforms above represent the serious AI-native and AI-mature contenders by 2026.
What’s coming next in this category
Three architectural shifts will reshape AI video research in 2026-2027.
Figma’s own AI test generator. Figma’s Make + AI moves prototype generation upstream of testing. The next-gen workflow looks like: AI generates the prototype, AI tests it with customers, AI summarizes the findings — all inside the design tool. Platforms that bring live Figma prototypes into the interview itself (Conveo embeds them as interview stimuli today) will see the prototype-testing share consolidate. Platforms whose strength is independent of design tool (User Intuition’s adaptive laddering across modalities) keep the share of motivational research that doesn’t start in Figma.
AI-prototyping tools as a new test surface. v0, Lovable, Bolt, and similar AI prototyping tools generate working web apps in minutes. Live URL testing volume per team is going up, not down — the bottleneck moves from build to validate. Platforms with deep live-URL screen-share capability (User Intuition, Maze, Conveo) capture more of this volume than platforms whose strength is video clip aggregation.
Async video diaries replacing scheduled UX sessions. The Zoom-plus-recruiter scheduled session is structurally expensive and capacity-capped. Async video + screen-share with AI moderation collapses the scheduling tax. Voxpopme’s async-first architecture had this thesis early; the AI-native cohort (User Intuition, Conveo, Outset) extends it with depth on every session rather than aggregation across many shallow ones. Both bets compete for the same UX research budget that used to fund scheduled Zoom.
For teams investing in video research today, the durable bet is platform architecture over feature checklist. A platform built AI-first, with deep moderation methodology, owned panel infrastructure, and pricing that scales with research cadence rather than team size, will absorb each of these shifts more gracefully than a platform retrofitting AI features onto a survey or repository core. See /posts/video-customer-interviews-complete-guide/ for the full methodology overview, /posts/video-customer-interviews-cost/ for the pricing math across the category, and /platform/in-depth-interviews/ for screen-share modality detail. The decision is research-object fit. Match the platform to the deliverable, run the paid pilot, and let the synthesis output decide.