If you run surveys for a living, the hard part of open-ended questions is not reading them. It is turning a column of text into codes that sit next to each respondent in your data file, so they can be cross-tabulated against age, region or brand, and coded the same way again next wave. Most "AI text analysis" tools answer a different question: what are people talking about? This guide compares the tools that market research agencies and in-house insights teams actually consider for the first job, and says plainly where each one wins.

Disclosure and method. We make Survey Coder, one of the tools below. Every statement about another product comes from that vendor's own website or help centre, linked in its section. Where we could not confirm something in a public source, the table says "not documented" rather than "no": the feature may exist, we just could not read it. Last verified: 9 October 2026.

The short answer: who should pick what

  • Your study lives in Qualtrics and your licence includes Text iQ: use Text iQ. Topics and sentiment come out in the response export, which Qualtrics downloads as an SPSS .sav file.
  • Your study lives in QuestionPro: try its TextAI dashboards first. It can follow a codebook you upload and reuse one from another dashboard.
  • You need to manage a team of human coders, or buy coding as a service: Ascribe is built for that, and it exports a .sav.
  • You mix surveys with reviews and tickets and want dashboards: Caplena.
  • You want coding, tables and reports in one analysis platform: Displayr.
  • Your work is academic or mixed-methods (interviews plus surveys): NVivo or MAXQDA.
  • You field on several platforms, run trackers, and need the codes back in your .sav and then in a banner table: that is the job Survey Coder was built for.
  • You have a one-off of a few hundred answers: ChatGPT or a spreadsheet will do.
  • IBM SPSS Text Analytics for Surveys is often recommended for this question, but IBM support staff say it is no longer sold. See its section below.

Comparison table

Seven criteria that decide whether open-ended coding fits a research workflow. "Codes to SPSS" means the codes arrive as variables per respondent in a .sav, not just a chart or a summary.

Tool AI coding of open ends Same codebook across waves Human review of codes Codes to SPSS .sav Multilingual Tabulation / banners Public pricing
Survey Coder Yes: proposes a codebook or applies yours; several codes per answer Yes: the next wave inherits it Your team reviews only the answers the AI is unsure about Yes: a column per code, per respondent, in your .sav or Excel Yes Yes: tab book with banner and significance, from an uploaded .sav Yes: credit packs per coded response
Qualtrics Text iQ Topics by query and by hand; recommended topics in Advanced Topics export and import between surveys Add or remove topics per response Yes: topics are in the response export, available as .sav Yes: 14 languages listed Qualtrics dashboards and widgets One plan with Text iQ is published; others on request
QuestionPro TextAI Themes, sub-themes and sentiment; can follow your codebook Reuse a codebook from another dashboard Tag responses by hand, merge, recode Not documented for TextAI fields (CSV export documented) Not documented in pages read QuestionPro BI dashboards Not checked
Forsta Machine-learning categorization plus sentiment Not documented in pages read Forsta analysts shape the model with you Not documented in pages read Not documented in pages read Reporting templates Not published in pages read
Caplena Yes: AI topics, or import your codebook Yes: copy topics from another project, or append waves Yes: review and adjust Not listed: Excel and CSV, including a column per topic Yes: automatic translation to one main language Dashboards and driver analysis No: annual credits, "Book a call"
Ascribe Yes: from manual to fully AI, your choice Yes: codebooks saved for future use Yes: coders, with team management Yes Yes Tabulation data output No: demo on request
NVivo AI Assistant suggests child codes from coded content Codebook export and import across projects Yes: the researcher edits Yes: a coding matrix can be saved as a new .sav Not checked Matrix queries, not banner tables Licence-based; no prices in pages read
MAXQDA AI Assist add-on applies your codes and suggests subcodes Not documented in pages read Yes: you accept each suggestion Indirect: turn codes into document variables, export those as .sav Not checked Statistics in MAXQDA Stats (Analytics Pro) Not checked
Displayr Yes: builds a code frame from a sample, applies it to all Coding can run automatically as new data arrives Not documented in pages read Could not confirm from public pages Not checked Yes: tables, crosstabs and automated testing Yes: from EUR 2,919 per user per year; limited free tier
ChatGPT or Excel by hand ChatGPT: whatever you prompt; Excel: none Only if you keep and reapply it yourself You review everything You build the columns yourself Yes (ChatGPT) Pivot tables Yes
IBM SPSS Text Analytics for Surveys Automatic categorization (desktop software) Categories import and export as Excel Manual refinement Excel or IBM SPSS Data Collection file Not checked In SPSS Statistics No longer sold

Survey Coder

Survey Coder is built for one job: coding the open-ended questions of a survey so the codes go back into the data. The AI proposes a codebook from your responses, or applies the codebook you already use, and then codes every response, with several codes where the answer carries several ideas, plus nets. Your team reviews only the answers the AI marks as doubtful instead of reading everything. The codebook carries over to the next wave, so a tracker stays comparable. It works across languages. The codes come out as columns per respondent in your .sav or in Excel, and if you uploaded a .sav you also get a tabulation book with banner and significance testing. It reads .sav, Excel and CSV, so it does not matter which platform the fieldwork ran on. The point is to give your coders leverage, not to replace them.

Best when: you run survey programmes (monthly NPS, brand trackers, concept tests) across different survey platforms, and the deliverable is a coded .sav and a table book.

Not the best fit when: you want a feedback dashboard over reviews, tickets and social posts, or you need to manage a large team of human coders working by hand. Caplena and Ascribe are stronger there.

Details: how coding works and what the export contains. Pricing is per coded response in one-off credit packs.

Qualtrics Text iQ

Text iQ is the text analysis module inside Qualtrics. You can add and remove topics on any response by hand, and Advanced Text iQ clients get recommended topics once there are at least 500 comments without topics in a single language. Topics and their parents can be exported for use in other surveys. Topics and sentiment are included in response exports, and Qualtrics lists SPSS (downloaded as .sav) among its export formats. Both Basic and Advanced support 14 languages, including English, Spanish and Portuguese; Advanced is requested through your account executive. Qualtrics publishes one plan that includes Text iQ: Strategic Research, at $420 per month ($5,040 per year), with a 30-day trial.

Best when: your surveys are fielded in Qualtrics and Text iQ is already in your licence. Nothing to move, and the topics travel with the export.

Not the best fit when: your studies come from several platforms, or you need a banner tab book built from the coded data outside Qualtrics.

Sources: Topics in Text iQ, Sentiment analysis, Export formats, Text iQ functionality, plans. Our longer comparison: Survey Coder vs Qualtrics Text iQ.

QuestionPro TextAI

TextAI turns open-ended feedback into themes, sub-themes and sentiment. You can upload your own code frame as a template file so TextAI follows that structure, or have a new dashboard follow the codebook of another dashboard or report, which QuestionPro says avoids rebuilding the same codebook each time. In Theme Configuration you can create themes, merge sub-themes, tag responses by hand and remove tags, then recode. The output can be imported into a QuestionPro BI dataset as Theme, Sub Theme and Sentiment fields. The TextAI export page documents a dashboard-level CSV; we found no documented route for TextAI fields into an SPSS file. QuestionPro does export the survey's raw data to SPSS, open-ended text included if you switch it on. QuestionPro also announced in April 2026 that it had acquired Fathom, a text analytics platform for open-ended survey data.

Best when: your surveys already run on QuestionPro and you want themes in its BI dashboards.

Not the best fit when: you need the codes as variables in a .sav, or your data comes from other platforms.

Sources: Use Text AI output in datasets, Theme configuration, Codebook in advanced text analytics, Export TextAI dashboard data, SPSS export. On Fathom: what the acquisition changes.

Forsta

Forsta is a survey, CX and research platform. Its text analytics data sheet describes machine-learning categorization with automated sentiment in every approach, and a service model: Forsta's experts apply or create the categorization model from your approved framework, and its analysts work with you to shape the algorithm. Its January 2025 product news adds an AI summarize feature that distils large volumes of data into themes. Qualtrics announced on 18 May 2026 that it had completed its acquisition of Press Ganey Forsta. We did not find public documentation of how coded categories are exported to SPSS, so we leave that cell open.

Best when: you already run research on Forsta and want text analytics set up with the vendor's analysts.

Not the best fit when: you want to run coding yourself, outside a platform contract, and hand over a coded .sav.

Sources: Text analytics data sheet (dated 2022), Forsta AI capabilities, Qualtrics completes acquisition.

Caplena

Caplena, based in Zurich, is an AI feedback analysis platform with plans for brand teams and for market research agencies. You can let the AI generate topics or upload your own codebook as CSV or Excel, and copy topics from another project. For recurring surveys, Caplena recommends keeping one project and appending each wave with a wave column, so the same codebook applies. It translates automatically with DeepL and Google Translate, so a multilingual dataset is analysed in one main language. Exports cover Excel and CSV, including a tabular format with a column per topic and a format with one row per response; SPSS .sav is not named on its export page. Pricing is annual credits, with prices given on a call.

Best when: you combine surveys with reviews, tickets and other feedback and want dashboards and driver analysis, especially as an agency.

Not the best fit when: the deliverable is a .sav with the codes attached and a banner tab book.

Sources: Import, copy and export a codebook, Languages and translations, Exporting your data, Pricing. Our comparison: Survey Coder vs Caplena.

Ascribe

Ascribe calls its Coder "verbatim coding software and web-based team management system". It covers everything from manual coding to AI-assisted human coding to fully AI-powered coding, and coders control the level of automation. Codebooks can be saved for future use, its AI Coder codes in multiple languages, and managers can oversee a team of coders in real time. Its help centre documents a SAV file export "for loading into SPSS and/or Dimensions", plus Excel outputs and a column-binary output for tabulation. Prices are not published; you can request a demo.

Best when: you run a coding department, need to split and supervise work across human coders, or want a vendor to code for you.

Not the best fit when: you are a small insights team that wants the AI to do the first pass and only check the doubtful answers, without running a coding operation.

Sources: Ascribe Coder, Download data with scripted output. Our comparison: Survey Coder vs Ascribe.

NVivo

NVivo, from Lumivero, is qualitative and mixed-methods software. It imports Qualtrics surveys directly, storing respondents as cases, closed questions as case attributes and open-ended questions coded to codes; other survey exports come in as spreadsheets. Its AI Assistant, which uses OpenAI under an agreement where uploaded data is not used for training, suggests child codes from content you have coded, summarizes, and can autocode sentiment. The NVivo help says you can export a coding matrix "as an SPSS sav file", which gives you a new file rather than appending codes to your original one. A codebook can be exported and imported to reuse a code structure across projects. Lumivero now also offers NVivo AI Cloud for individual subscriptions, with team access announced for early 2027.

Best when: surveys are one source among interviews, focus groups and documents, and the analysis is interpretive.

Not the best fit when: you code thousands of answers per wave on a commercial deadline and the output has to be banner tables.

Sources: Framework and coding matrices, AI Assistant, Qualtrics import, Export codes and codebooks, NVivo product page. Our comparison: Survey Coder vs NVivo.

MAXQDA

MAXQDA imports survey data from Excel and SPSS files; open-ended columns come in as text, already coded with their column header. With the AI Assist add-on, its AI Coding feature applies a set of codes you have created to survey responses, explains each suggestion, and lets you decide which to accept; a companion feature suggests subcodes. To get codes into SPSS, MAXQDA can turn a code into a document variable (its "quantitizing" function, which can be switched to 1/0), and document variables export as a .sav. Statistics on code frequencies run in MAXQDA Stats, which needs an Analytics Pro licence.

Best when: you are a researcher who wants to stay in control of every code, with AI as a second pair of eyes, in a mixed-methods project.

Not the best fit when: you need the AI to draft the whole codebook and code a large wave, then reuse it on the next one.

Sources: AI Coding: segments and survey responses, Survey data from Excel, Transform a code into a document variable, Exporting data to Excel and SPSS, Analyze survey data with MAXQDA Stats.

Displayr

Displayr is an analysis and reporting platform for survey data. It says its AI can auto-generate code structures from a sample of responses and apply them across the full dataset in one workflow, and that category coding can run automatically as new data arrives. It also covers tables and crosstabs, automated statistical testing and reports, so the same dataset goes from raw responses to a shareable report without leaving the platform. Pricing is published: a free plan with limited AI text analysis, Professional from EUR 2,919 and Enterprise from EUR 3,649 per user per year, with a 14-day trial. We could not confirm from its public pages how coded variables export to a .sav, because its help centre did not load for us.

Best when: you want coding, tables and reports in one analysis tool and are happy to work inside it.

Not the best fit when: your team already tabulates elsewhere and only needs coded variables added to its own file.

Sources: Pricing, AI methods to code open-ended responses, SPSS vs Displayr.

ChatGPT, or coding by hand in Excel

For a few hundred answers, a spreadsheet works: one column per code, a 1 or a 0, and a pivot table. ChatGPT can speed up the first draft of a codebook and suggest codes for an uploaded file. What neither gives you is a process: you have to keep the codebook yourself and reapply it identically next wave, check every assignment, and rebuild the code columns and labels before the file goes to SPSS. That is manageable once; it becomes the bottleneck in a tracker.

Best when: a one-off study, a small sample, or exploring themes before you fix a codebook.

Not the best fit when: the study repeats, several people code, or someone downstream needs a clean .sav.

More detail: Survey Coder vs ChatGPT and Survey Coder vs manual coding.

IBM SPSS Text Analytics for Surveys

This is the tool most often named for "code open ends and export to SPSS", so it deserves a clear note. It was IBM's desktop application for categorizing survey text responses. Its documentation describes importing your own categories, or older code frames, from Excel; exporting categories back to Excel; and exporting categorization results to Excel or an IBM SPSS Data Collection file, as dichotomies or categories, with the multiple-response variable created automatically. The export carries respondent IDs and category labels, but not the open-ended text or the other variables, so you merge it back into your data yourself.

The key fact for a buyer today: in the IBM Community forum, IBM SPSS technical support stated that the product "is no longer sold" (February 2025) and that it reached end of support and was removed from the IBM catalogue (September 2024). The two replies give different dates, 2023 and 2020, but agree it is gone. IBM's suggested alternatives were the SPSS TEXTANALYSIS extension command, IBM SPSS Modeler Premium with Text Analytics, and IBM Watson Discovery; an IBM Champion added that the free extension "isn't a replacement".

Best when: you already have a licence installed and a working project.

Not the best fit when: you are choosing a tool now.

Sources: Exporting to Microsoft Excel, Importing and exporting predefined categories, IBM Community, February 2025, IBM Community, September 2024.

How to choose: a checklist

  1. Where does your data live? If every study is in one survey platform, its built-in module is one less tool. If studies arrive as .sav or Excel from several sources, you need a tool that reads files.
  2. What is the deliverable? A dashboard, a report, or a coded .sav that goes to tabulation. Ask the vendor to export a real file and open it in SPSS before you buy.
  3. Does a response get several codes? Take ten answers that clearly say two things and count the codes assigned.
  4. Can you bring your own codebook, and does wave 2 inherit wave 1's? In a tracker, the codebook is the instrument.
  5. Where does a human check the work? Reviewing everything defeats the purpose; reviewing nothing is not defensible. Look for a way to see the doubtful answers first.
  6. Which languages? Check yours on the vendor's own list, not in the demo.
  7. How is it paid? Seat licence, annual credits or per response changes the cost of a study that runs once a quarter.

Frequently asked questions

What is the best AI tool to code open-ended survey responses and export to SPSS?

It depends on where your data lives. Inside Qualtrics, Text iQ topics come out in the .sav response export. Ascribe and Survey Coder both export a .sav with the codes for data from any source; Ascribe adds human coding teams and a coding service, Survey Coder adds AI-first review of doubtful answers and a banner tab book from your .sav. NVivo can save a coding matrix as a new .sav.

Is IBM SPSS Text Analytics for Surveys still available?

According to IBM SPSS technical support in the IBM Community forum, it is no longer sold and has been removed from the IBM catalogue. IBM suggested the SPSS TEXTANALYSIS extension, SPSS Modeler Premium with Text Analytics, or Watson Discovery instead.

Can I just use ChatGPT to code open-ended responses?

For a small, one-off file, yes, as long as you check the result. For repeated waves, you have to keep the codebook stable yourself, and build the code variables for SPSS by hand.

What should the SPSS export of coded open ends look like?

One row per respondent, the respondent ID, and one variable per code (or a multiple-response set) with value labels, attached to the rest of the survey variables so the codes can be cross-tabulated straight away.

Does AI coding replace human coders?

No. It moves their time: the AI does the first pass across every answer, and people spend their judgement on the codebook and on the answers the AI is unsure about.

Next steps

To see how your own answers code, the free coding tool takes up to 100 responses, proposes a codebook from the actual text and returns an Excel, with no account needed. For the method behind good coding, read AI open-end coding: how it works and where it breaks; for what happens after, cross-tabulation analysis. All our side-by-side pages are in the comparison hub.