Affect Performance Team

|News|Nov 11, 2026

UserTesting October 2026 Updates: AI Moderation, Card Sort, and Research Tools

In October 2026, UserTesting presented another broad update to its user experience research platform. The company focused on automating qualitative research, expanding unmoderated testing, and bringing customer insights into AI tools.

The October overview covered AI Moderation, Card Sort in New Think Out Loud, Advanced Targeting, MCP integrations, and AI Relationship Quality. UserTesting also introduced improvements for research participants, reusable study templates, and the ability to conduct certain web tests without installing a browser extension.

The features arrived at different times. AI Moderation and Card Sort in New Think Out Loud were introduced in September, while reusable templates arrived in August. In October, the company expanded international recruiting, introduced a Cursor integration, and announced a new approach to evaluating the human experience of AI.

The overall direction is to obtain more substantial user feedback without increasing the number of researcher-moderated interviews proportionally.

AI Moderation: Follow-Up Questions in Unmoderated UX Research

The AI Moderator Explores Participants' Reasons for Their Actions and Decisions

In September 2026, UserTesting introduced AI Moderation for the new Think Out Loud experience. The feature was included in the October roundup.

Traditional unmoderated studies ask participants to complete tasks and comment on their actions without a researcher present.

This method can collect feedback from many users simultaneously, but it has an important limitation: when a participant responds too briefly or their actions need clarification, the researcher cannot ask a follow-up question during the session.

AI Moderation addresses this specific issue.

After a participant gives a spoken response, the AI moderator analyzes the answer and generates further questions relevant to the research objective.

For example, an online retailer is testing a new product detail page. A participant reviews the page and says, “I'm not sure I would buy this product.”

A conventional unmoderated study might stop at that response. AI Moderation can ask what creates the uncertainty: missing product information, shipping terms, the price, reviews, or something else.

The researcher therefore receives not only the initial reaction but also explanations for the participant's behavior.

Up to Three Follow-Up Questions for Each Spoken Response

According to UserTesting documentation dated September 25, 2026, AI Moderation supports up to three follow-up questions after each task that calls for a spoken response.

A three-minute time limit applies to an individual follow-up exchange.

The moderator uses both the participant's initial answer and any additional instructions from the researcher. Questions are spoken with synthesized speech and displayed on screen.

The AI does not interrupt a participant in the middle of an answer. The next question appears only after the participant finishes and clicks Next.

For example, if a checkout participant says the process was difficult, the AI might ask:

At what step did you encounter the difficulty?
What information did you expect to see?
What would have helped you complete checkout faster?

The system generates the exact wording based on the response and the research objective.

**Important:** AI Moderation does not necessarily ask all three questions. It may stop earlier if no further clarification is needed or the time limit has been reached.

Configuring AI Moderation With Research Instructions

Researchers can guide the direction of the follow-up questions. In New Think Out Loud, they open the test settings and enable AI Moderation.

For tasks with spoken responses, an additional AI Moderation Prompt field becomes available. It accepts instructions of up to 1,000 characters.

For example: “If a participant cannot choose a product, find out what information is missing: specifications, reviews, comparisons, or purchase terms.”

This helps the AI moderator focus on a particular research goal. However, the instruction is not an exact interview script; the AI determines the wording of the follow-up questions.

Researchers are encouraged to use Preview before launching a study to check the moderator's behavior.

Results are stored with the other study materials. The recording includes participant answers and AI questions, the full conversation appears in the transcript, and the responses contribute to thematic analysis.

AI Moderation Limitations During Closed Beta

At the time of the October announcement, AI Moderation was in **closed beta** and available only to selected customers.

It supported New Think Out Loud and participants completing tests on desktop computers with the UserTesting browser extension installed.

Mobile studies and extensionless tests were not supported. The first version of the AI moderator worked in **English only**.

Follow-up questions could be asked only after spoken answers, rather than after every type of task.

Researchers also could not independently set the maximum number of follow-ups or enable moderation separately for each spoken question.

Businesses conducting multilingual studies or primarily testing mobile audiences need to account for these restrictions.

AI Moderation does not replace an in-depth interview with a professional researcher when the conversation must change direction freely, explore unexpected responses in depth, or address particularly complex user situations.

Card Sort: Studying Website Structure in New Think Out Loud

Testing How Users Group Products, Sections, and Information

In September 2026, UserTesting added Card Sort to the New Think Out Loud interface. This capability also appeared in the October overview.

Card Sorting is a research method for understanding how users perceive the structure of information.

Participants receive cards labeled with products, website sections, or information items and arrange them into categories.

For example, a sporting goods retailer wants to redesign its catalog navigation. The researcher creates cards labeled “Running Shoes,” “Training Apparel,” “Fitness Trackers,” “Hiking Backpacks,” and “Sports Nutrition.”

Participants group these cards according to how they think the store should be organized.

The findings help identify where users expect to find certain products and which category names create confusion.

For e-commerce, this is particularly useful when designing navigation, menus, category pages, and filters.

Closed Card Sort in the New Research Interface

At the time of the update, New Think Out Loud supported **Closed Card Sort**.

In this format, researchers define categories in advance and participants place the cards into those categories.

A retailer can use it to test whether its current menu structure is understandable. If participants repeatedly put products into different categories from those expected, this may indicate a problem with the information architecture.

Classic Think Out Loud offers additional options:

Open Card Sorting: participants create their own groups and name the categories.
Closed Card Sorting: participants sort cards into predefined categories.
Hybrid Card Sorting: researchers provide an initial structure, but participants may create additional categories.

New Think Out Loud therefore gained built-in Card Sort support but did not yet have complete feature parity with the Classic interface.

When a study requires Open or Hybrid Card Sorting, researchers should use Classic Think Out Loud.

Combining Card Sort With Video Feedback

A practical advantage of UserTesting is the ability to combine structured research with participants' qualitative comments.

While testing navigation, for example, a participant can group cards and explain why a particular category seemed appropriate.

Researchers can then examine not only the sorting results but also the reasons behind participants' choices.

This helps companies identify ambiguous section names and categories that need clarification.

However, Card Sorting does not measure a website's actual conversion rate. It informs or validates information architecture; catalog changes should still be tested in the real user interface.

Tree Testing can be used to evaluate how easily people find information within a proposed structure.

Testing Without Installing the UserTesting Browser Extension

Supported Web Studies Can Reach Participants on Restricted Devices

In September 2026, UserTesting made it possible to conduct certain recorded web studies without requiring a browser extension.

Previously, installing an extension could be an obstacle for research participants, particularly employees on corporate devices where installing software is restricted.

As a result, researchers sometimes had to exclude relevant B2B participants or use alternative feedback methods.

The updated functionality allows supported Think Out Loud and Interaction Tests to run without a separate extension.

This makes unmoderated studies more accessible to participants with limited control over browser settings, although compatibility must be checked for each study type and browser.

**This option did not extend to the initial AI Moderation beta:** AI-moderated studies still required the UserTesting extension.

Reusable Study Templates: Repeating Research Designs

Saving Existing Studies for Other Projects and Teams

In August 2026, UserTesting expanded the reuse of research designs in New Think Out Loud.

Users could save an existing study as a template and apply it in other workspaces.

For example, an e-commerce company regularly tests new product category pages. Rather than rebuilding tasks, questions, and settings each time, the team can start from an approved template and adapt it to the relevant category or audience.

This is particularly useful for international businesses and agencies whose research teams follow common standards.

Templates reduce setup time and improve the comparability of recurring studies.

However, reusing a design does not guarantee identical research conditions. Audience composition, device, language, study period, and differences between the products being tested still matter.

Advanced Targeting: Recruiting Specialized Audiences in More Countries

Expanded Geographic Coverage for Professional and Niche Research

In October 2026, UserTesting expanded the geographic availability of Advanced Targeting.

Previously, these additional recruiting capabilities were mainly focused on the United States.

Specialized participant recruiting was extended to the United Kingdom, Australia, Canada, France, Germany, and India. New Think Out Loud also gained additional availability in New Zealand and Singapore.

These changes are especially relevant to B2B research, where basic demographic filters often are not enough.

A SaaS company, for instance, may need professionals from a particular industry, seniority level, and company size rather than participants in a broad age bracket.

Advanced Targeting supports **more than 40 criteria**, organized into groups:

Personal characteristics.
Professional characteristics, including industry, seniority, and company size.
Products and services used.
Technical characteristics and experience.

This can improve the precision of research samples. However, narrower targeting can increase recruiting time and participant acquisition costs.

Recruiting Through User Interviews From Within UserTesting

On October 28, 2026, UserTesting updated its Advanced Targeting and User Interviews documentation.

For supported studies, researchers can select an automatically sourced audience and apply additional criteria.

Participants are recruited through User Interviews infrastructure, while the study and its results remain in UserTesting.

The workflow includes:

Create a study.
Open the audience settings.
Select Auto-Sourced Audience.
Enable Advanced Targeting.
Configure professional and other criteria.
Review costs and start recruiting.

Participant details and responses then appear in the UserTesting results interface.

Advanced Targeting is not a separate study type. It is an additional audience selection capability.

It is limited to eligible accounts and uses the applicable participant payment system. The account must have sufficient recruiting funds before a study can launch.

MCP Server and Cursor: Research Data Inside AI Development Tools

Using Customer Feedback Without Leaving the Development Environment

In October 2026, UserTesting announced an integration with Cursor Marketplace.

It uses the existing UserTesting MCP Server, whose functionality the company had expanded earlier, including in September.

MCP lets compatible AI tools access UserTesting data and operations through standardized interfaces.

For example, a developer working on a new signup flow in Cursor can request findings from previous tests instead of switching between the development environment and the research platform.

The integration provides access to existing studies and complete user-session transcripts.

Supported tools also provide operations for creating and launching studies and recruiting participants.

This brings real customer feedback closer to the design and development process.

Creating Studies and Working With Transcripts Through AI

A user can give natural-language instructions to a compatible MCP client.

For example: “Find studies where users struggled with registration and summarize recurring problems.”

The AI can access the relevant materials within the connected account's permissions.

Another example: “Prepare a new study of the mobile signup flow.”

If the required MCP capabilities are available, the tool can help create the corresponding test.

For product and marketing teams, this reduces the steps between identifying a problem, preparing a new study, and analyzing the findings.

MCP access does not mean an AI tool automatically receives access to all company research. Available material is determined by integration settings, authentication, and user permissions.

AI Relationship Quality: Measuring User Trust in AI Products

Five Dimensions of Human-AI Experience

In October 2026, UserTesting introduced AI Relationship Quality (ARQ), a research methodology for assessing users' experiences with conversational and agentic AI systems.

Unlike technical tests that measure answer accuracy or task completion, ARQ evaluates how users perceive the AI.

The methodology includes five dimensions:

Understanding: whether users understand what the AI does, why it makes decisions, and what its limitations are.
Trust: whether users trust the system and recognize when its results need verification.
Control: whether users can change AI actions, correct errors, or stop a task.
Outcome: whether the AI produces results that meet user needs.
Affinity: how comfortable users feel interacting with AI and whether they want to keep using it.

Results include an overall experience score from **0 to 100**, scores for the individual dimensions, and participants' comments and videos.

ARQ Is a Research Service, Not a Standalone Product Feature

At the time of the October announcement, ARQ was delivered through **UserTesting Professional Services**.

It was not a standard self-service tool that every customer could activate in the platform interface.

UserTesting specialists organized the study around the client's product, target audience, and business objectives.

For instance, an e-commerce company may launch an AI shopping assistant. Technical testing might show that it finds products and generates recommendations correctly.

ARQ can additionally explore whether customers understand why products were recommended, trust the system, and feel in control of their choices.

Findings can inform changes to the AI interface, action confirmation rules, and interaction design.

ARQ does not replace technical testing for accuracy, safety, or task performance. Its purpose is to supplement those evaluations with evidence about the human experience.

Practical Value for E-commerce

For online retailers, the main benefits relate to research on product selection and checkout.

AI Moderation helps investigate purchase hesitation, reasons for abandoning a purchase, and difficulties during checkout.

For example, if a participant says they could not select the correct size, the AI moderator can ask what information was missing and what would have helped them decide.

Card Sort is useful for checking catalog structure, categories, and navigation.

Reusable Templates support recurring tests across product categories, while Advanced Targeting helps recruit shoppers with relevant characteristics in different countries.

A retailer introducing its own AI shopping advisor can use ARQ to evaluate buyers' trust in its recommendations.

These studies complement quantitative GA4 analytics, heatmaps, and A/B testing.

However, findings from small qualitative studies should not automatically be generalized to the store's full audience. They should inform hypotheses that are subsequently tested.

Practical Value for Customer Services

For service businesses, UserTesting can help investigate application forms, appointment scheduling, plan selection, and customer support processes.

AI Moderation explores the reasons users struggle with digital services.

Advanced Targeting recruits participants with relevant professional or consumer experience, while reusable templates allow teams to repeatedly test changes in customer journeys.

When introducing AI support, ARQ can complement customer satisfaction and issue resolution metrics.

An AI agent may resolve requests successfully while users still do not understand which actions are being taken on their behalf. ARQ can investigate trust and control separately.

Practical Value for B2B SaaS

For B2B SaaS, the most relevant capabilities are Advanced Targeting, AI Moderation, and MCP integrations.

Advanced recruiting helps find participants with specific roles, responsibilities, and experience using particular software categories.

A company might recruit marketing leaders who regularly use analytics and advertising platforms.

AI Moderation provides more detailed feedback about complex interfaces, while MCP makes research findings available directly in development and product workflows.

This is especially useful when evaluating multi-step tasks in B2B software.

However, the initial AI Moderation release supported only English and certain testing configurations, a limitation to consider in international research.

Availability and Limitations of the October Updates

As of early November 2026, UserTesting's features had different availability statuses.

AI Moderation was in closed beta and supported English-language New Think Out Loud studies on desktop computers with the browser extension.
Card Sort in New Think Out Loud supported Closed Card Sorting. Open and Hybrid methods required Classic Think Out Loud.
Extensionless Testing was available for supported web studies, but not the initial AI Moderation beta.
Reusable Study Templates were introduced in August 2026 and included in the October product overview.
Advanced Targeting reached additional markets, with availability depending on the account, study type, and recruiting terms.
UserTesting MCP Server evolved throughout the year, with the Cursor Marketplace integration announced in October.
AI Relationship Quality was delivered through Professional Services rather than as a universal self-service module.

Teams therefore need to confirm that each specific feature is available in their UserTesting account and supports the intended research method.

UserTesting: Platform Capabilities and Target Audience

UserTesting is a user experience research platform for moderated and unmoderated studies of websites, mobile apps, digital products, and prototypes.

It combines participant recruiting, session recording, information architecture research, feedback analysis, and collaboration tools for product teams.

The October 2026 overview illustrates three directions of development.

First, AI Moderation aims to collect richer qualitative feedback.

Second, Card Sort, Advanced Targeting, and the updated Think Out Loud expand available research methods and participant access.

Third, MCP brings research findings into AI-enabled development workflows, while ARQ evaluates the quality of AI product experiences.

The updates are relevant to UX researchers, product teams, e-commerce companies, customer service providers, and B2B SaaS businesses.

Their practical value lies in faster study preparation, richer customer insights, and more effective use of research findings to improve digital products.