Qualtrics AI: Text iQ and AI Topic Models for Research Data

Qualtrics AI, for research purposes, means the machine-learning text analysis built into Qualtrics: Text iQ on the survey platform, which tags topics and scores sentiment in open-ended survey responses, and AI-powered topic models in Text Analytics, which build a topic hierarchy from text with Qualtrics AI. Both work on text you have already collected, so they help you code and summarise free-text answers; they do not replace the human review that makes coded qualitative data defensible. This guide separates the two tools, gives the menu paths and permissions each one needs, explains how to read the output, and lists open-source alternatives.

What Qualtrics AI is used for in research

Researchers who run surveys with open-ended questions end up with free text that has to be coded before it can be counted. Qualtrics offers two AI-assisted routes for that job, and they sit in different parts of the product. Treat them as separate tools with separate access rules.

Text iQ on the survey platform

Text iQ lives in the Data & Analysis tab of a survey project. According to the Qualtrics support documentation, its functions are topic tagging (with parent and child topics), sentiment analysis (a score shown per response), widgets, a Search and Query Builder, Recommended Topics, Translate Comments and Text iQ-powered survey flows, plus spell check and lemmatization. Several of these are limited to the Advanced tier, which the table in the next section sets out.

AI-powered topic models in Text Analytics

Text Analytics is a separate area reached from the global navigation. Its AI-powered topic models build a topic hierarchy from your text together with context you supply: use case, persona, industry and company. The hierarchy can have one to five levels; Qualtrics suggests 3 to 5 levels for an in-depth analysis. The model can draw on survey projects, imported data, online reputation management, email, chat and voice sources, but not on EX projects such as Engagement, and one model can cover up to 10 projects.

What the documentation does not describe

The official support pages for these two tools describe analysis of text that respondents have already written. They do not describe an AI function that writes or reviews survey questions in the survey builder, so this guide does not cover one. If you need help with question design, see the CASRAI guides on Likert scales in Qualtrics and how to use Qualtrics.

Access and requirements

Access depends on three things: your licence tier, your user permissions and the project type. Qualtrics support states that Text iQ Basic allows up to 20,000 responses per survey and has no summary mode, widgets, sentiment analysis or Query Builder. Text iQ Advanced has unlimited responses and all features. The Qualtrics documentation says to reach out to your Account Executive to obtain Advanced. Because licensing differs by institution, ask your Qualtrics administrator which tier your account holds before planning an analysis around sentiment scores or Recommended Topics. This guide gives no pricing.

• Text iQ: you need the Edit Survey Responses permission on the project.
• AI topic models: you need the Automated Text Analytics user permission; without it, the Text Analytics section does not appear in the navigation.
• Language: sentiment analysis is optimised for a stated list of languages (see the output section below).

The data table below compares the Basic and Advanced Text iQ tiers feature by feature, using the Qualtrics support page for Text iQ functionality.

Step-by-step: run Text iQ on a survey project

The steps below follow the Qualtrics support documentation for Text iQ. Menu labels can change between product releases, so match them against your own account.

1. Open your survey project and go to the Data & Analysis tab. Confirm you hold the Edit Survey Responses permission.
2. Open Text iQ. On your first visit, tick the open-ended fields you want analysed using the checkboxes.
3. Click Upload data. Dashboards and data are unavailable until the upload finishes, so wait for it to complete before judging the results.
4. Use the Field dropdown to switch between the open-ended fields you selected. Use Edit topics to build or change topics and Return to summary to go back to the overview (summary mode is Advanced only).
5. Use Select fields to display and Sort by to control which columns show and how responses are ordered while you review.
6. When you finish editing topics, choose Review and publish. The options are Publish, Revert Changes and Cancel. Publish the topics once you are satisfied so the tags become part of your data.

Step-by-step: generate an AI topic model in Text Analytics

The path below follows the Qualtrics support page on AI-powered topic models.

1. In the global navigation choose Text Analytics, then Generate a topic model.
2. Name the model and choose the source Engage, then choose the projects and fields to include and click Next.
3. Choose Let Qualtrics AI build it and click Save.
4. Enter your organization name, industry, persona role and use case, and choose how many levels the hierarchy should have. Click Save. Qualtrics states that these parameters cannot be edited after generation, so plan them before you click Generate.
5. Click Generate, then Review the proposed hierarchy. Deselect any topic that does not fit your codebook.
6. Click Accept topics. The results appear in the Data & Analysis tab.

If you already have a codebook, the same support page describes an import route: choose Upload your topic hierarchy, then Choose file, Save and Upload, and finally Accept topics. For research this is often the better starting point, because the topics come from your own framework rather than from the model.

Recommended topics for uncategorized sentences

The model can recommend topics for sentences that have no topic yet. The Recommend new topics button appears at least 4 days after the last set of recommendations.

Enrichments you can add

The documented enrichments are actionability, emotion, emotional intensity, sentiment and effort. A data source called Granular Text Analytics provides sentence-level data.

Output and how to read it

Topic tags

Text iQ topics are organised as parents and children, so a response can carry a broad tag and a narrower one. Treat each tag as a candidate code, not a finding. A topic list produced by a model reflects patterns in your text; whether those patterns match your research question is a judgement only you can make.

Sentiment scores

Sentiment analysis assigns each response one of these labels: Very Negative, Negative, Neutral, Positive, Very positive or Mixed. The numeric score runs from -2 to +2, with Neutral at 0. Scores appear as embedded data in Data & Analysis and in Reports, so you can export them with your other variables. The documentation describes the model as transformer-based and cross-lingual, trained on a large and diverse body of real experience data; it uses the question text when that is available, and new model versions apply only to new comments.

Correcting a score

You can override the model. Enter edit mode, click the edit comment icon, and choose Comment Overall Sentiment, then pick the score and Save. Choose Topic Sentiment to correct the score for a single topic. Qualtrics notes that your edits do not immediately change sentiment in your own account; they are analysed periodically across users to improve the model.

Language coverage

Sentiment analysis is optimised for English, French, Spanish, German, Korean, Italian, Dutch, Japanese, Portuguese, Traditional and Simplified Chinese, Polish, Swedish, Thai, Indonesian and Russian. For other languages the documentation says Qualtrics cannot guarantee the same level of quality. Multi-language support is available to all clients with Sentiment Analysis, which itself is limited to Advanced Text clients.

Using the output in a research design

The following points are editorial guidance, not Qualtrics statements.

• Validate by human coding. Draw a sample of responses, code them yourself or with a second coder, and compare against the model’s tags and sentiment scores before using the model for the full dataset. The product itself supports this: you can edit sentiment and deselect topics before accepting them.
• Report the method. State that tags or scores came from Qualtrics Text iQ or an AI topic model, the date you ran it, the tier and the number of levels or settings you chose. Because new model versions apply only to new comments, responses analysed at different times may have been scored by different versions.
• Check language and domain fit. If your sample includes languages outside the optimised list, or specialised vocabulary, review those responses by hand.
• Handle participant data properly. Open-text answers can contain identifying details. Follow your ethics approval and data management plan when you decide what is analysed and exported; see the CASRAI entry on qualitative data collection techniques for background on the kind of data involved.

Troubleshooting

Symptom	Likely cause according to the documentation	What to do
No Text iQ or Text Analytics access	Missing permission: Edit Survey Responses for Text iQ, Automated Text Analytics for topic models	Ask your Qualtrics administrator to grant the permission.
No Text Analytics section in the navigation	The Automated Text Analytics user permission is not assigned	Request the permission.
Dashboards or data not showing after you select fields	The upload has not finished	Wait for Upload data to complete.
Sentiment, widgets, Query Builder or summary mode missing	These are not part of Text iQ Basic	Confirm your tier with your administrator; Advanced is obtained through your Account Executive.
Basic account stops at a response limit	Basic allows up to 20,000 responses per survey	Split the analysis or move to Advanced.
Topic model cannot be built from a project	EX projects such as Engagement are not supported; one model covers up to 10 projects	Use a supported project type and stay within 10 projects.
Generation parameters look wrong	They cannot be edited after generation	Generate a new model with corrected parameters.
Recommend new topics button is missing	It appears at least 4 days after the last set	Return after the waiting period.
Weak sentiment results in a smaller language	The language is not on the optimised list	Review those responses manually.

Free alternatives for coding open-text responses

If your institution does not license the Advanced tier, free and open-source software can support the human-coding step. These tools code text by hand rather than score sentiment automatically, so they answer a different need from Text iQ. For the wider survey tool landscape see the CASRAI comparison of survey tools for academic research.

Tool	Licence and cost	What it does	Where it runs
Taguette	Free and open source	Import materials, highlight passages and attach tags; export the codebook, highlights and tagged quotes	Locally on macOS, Windows or Linux, on a server, or hosted at app.taguette.org
QualCoder	LGPLv3	Assign codes to text, images and audio or video; hierarchical categories; coding frequency and coder comparison reports; optional AI features set up by wizard	Windows installer, macOS bundles for Apple Silicon and Intel, Linux from source

Frequently asked questions

Is Qualtrics AI the same as Text iQ?

No. Text iQ is the text analysis tool inside the survey platform’s Data & Analysis tab. AI-powered topic models are a separate feature in Text Analytics that uses Qualtrics AI to build a topic hierarchy. They have different permissions and different project-type rules.

Does Qualtrics AI work in every language?

Sentiment analysis is optimised for 16 listed languages, including English, Spanish, Japanese and Simplified Chinese. For others, Qualtrics says it cannot guarantee the same quality.

Can I correct the AI’s sentiment scores?

Yes. Use edit mode and the edit comment icon to set Comment Overall Sentiment or Topic Sentiment. Your edit does not immediately change your account’s model; Qualtrics analyses edits across users periodically.

Can I edit the AI topic model settings after generating it?

No. Qualtrics states the generation parameters cannot be edited afterwards, though you can review and deselect topics before you accept them.

Do I need a special licence?

Some features, including sentiment analysis, widgets, Query Builder and Recommended Topics, are Advanced only. Ask your Qualtrics administrator or Account Executive which tier you hold.

Can I pull the results out programmatically?

Sentiment scores are stored as embedded data, which means they travel with the response data. For export automation see the Qualtrics API guide.

Sources

• Qualtrics Support, Text iQ Functionality: qualtrics.com/support/…/text-iq-functionality (checked 2026-11-14).
• Qualtrics Support, Sentiment Analysis (Text iQ): qualtrics.com/support/…/sentiment-analysis (checked 2026-11-14).
• Qualtrics Support, AI-Powered Topic Models: qualtrics.com/support/…/ai-powered-topic-models (checked 2026-11-14).
• Taguette: taguette.org (checked 2026-11-14).
• QualCoder repository: github.com/ccbogel/QualCoder (checked 2026-11-14).