ATLAS.ti
ATLAS.ti is a computer-assisted qualitative data analysis (QDA) software designed to support researchers in coding, organizing, and interpreting unstructured data from sources such as text documents, audio, video, images, and social media.[1] It integrates advanced tools for thematic analysis, visualization, and collaboration, enabling users to uncover insights efficiently while combining human expertise with AI-driven features like automatic coding and sentiment analysis.[1] Developed initially by Thomas Muhr under Scientific Software Development GmbH (now ATLAS.ti GmbH), the software's first commercial version was released in 1993, marking it as one of the pioneering tools in the field of QDA.[2] Over the years, ATLAS.ti has evolved to support cross-platform use on Windows, Mac, and web environments, with capabilities for handling diverse data formats including PDFs, Word documents, HTML files, survey data, and reference manager outputs.[1] Its AI enhancements, powered by models like OpenAI's GPT, allow for rapid summarization and opinion mining, reportedly reducing analysis time by over 90% in some workflows.[1] In 2024, ATLAS.ti was acquired by Lumivero, LLC, a leader in research software, which has bolstered its innovation in SaaS offerings and global support.[2] The tool is widely adopted by academic institutions, corporations like Microsoft, and international organizations such as the United Nations for qualitative and mixed-methods research projects.[2] Headquartered in Berlin, Germany, with regional offices in Canada, Spain, and Ecuador, ATLAS.ti continues to emphasize user-friendly design and real-time team collaboration features.[2]Overview
Description
ATLAS.ti is a computer-assisted qualitative data analysis (CAQDAS) tool designed for researchers to manage, code, and analyze unstructured data, including text, multimedia, and survey materials.[3] It serves as a powerful platform that supports the systematic organization and examination of qualitative information, enabling users to handle complex datasets efficiently.[1] At its core, ATLAS.ti bridges human interpretive skills with computational tools to uncover patterns, themes, and insights within qualitative research.[3] This integration allows researchers to focus on conceptual depth while the software streamlines repetitive tasks, such as data sorting and initial pattern recognition, thereby enhancing the overall analytical process.[1] The software supports a wide range of primary documents as input, including text files, PDFs, images, audio and video recordings, social media content, and survey data.[1] These diverse formats enable comprehensive analysis of varied qualitative sources without the need for extensive manual preparation.[3] ATLAS.ti plays a pivotal role in disciplines such as the social sciences, market research, and academic studies, where it facilitates the exploration of non-numerical data to generate meaningful interpretations.[1]Primary Applications
ATLAS.ti is widely applied in qualitative research across diverse disciplines, including anthropology, sociology, psychology, education, health sciences (such as nursing and psychiatry), and business sectors like market research and customer feedback analysis.[4][5][6] In these fields, researchers leverage the software to manage and interpret non-numerical data from sources like interviews, field notes, and multimedia content, enabling deeper insights into social phenomena, human behaviors, and organizational dynamics.[1] For instance, anthropologists use it for ethnographic studies, sociologists for examining social structures, and health scientists for analyzing patient experiences or policy impacts.[7][8] Common use cases include thematic analysis of interview transcripts to identify recurring patterns in participant narratives, grounded theory development from field notes to build theoretical models inductively, content analysis of social media posts to evaluate public discourse, and mixed-methods integration where qualitative findings complement quantitative data for comprehensive evaluations.[7][9] In business contexts, it supports market research by coding customer feedback to uncover sentiment trends.[4] These applications highlight ATLAS.ti's versatility in handling varied data types without imposing rigid analytical frameworks.[1] The software's benefits in these applications stem from its capacity to process large datasets for efficient pattern recognition, facilitate collaborative research through shared projects and real-time editing, and promote reproducible analysis via detailed audit trails and exportable reports.[1][2] By organizing complex information into structured codes and visualizations, ATLAS.ti reduces manual effort, allowing researchers to focus on interpretive depth rather than administrative tasks.[10] It is employed internationally in academic theses, policy evaluations, and corporate strategy development, with adoption across numerous universities and organizations worldwide.[2]History
Origins and Founding
ATLAS.ti originated as part of the ATLAS project, initiated in 1989 at the Technische Universität Berlin in Germany by Thomas Muhr, a psychology graduate who shifted into computer science.[11] The project aimed to create an archive of modern everyday culture through a database of verbal data, drawing inspiration from historical ethnographic works like Fray Bernardino de Sahagún’s documentation of Aztec culture, while developing software to support qualitative text interpretation.[11] This interdisciplinary effort involved psychologists, linguists, and computer scientists, focusing on tools that would enable researchers to handle non-numerical data beyond manual methods, with the ".ti" file suffix denoting "text interpretation."[11] The initial development emphasized hypertext-based linking to facilitate the analysis of qualitative data, rooted in hermeneutic principles from the social sciences that prioritize closeness to the data and iterative interpretation.[12] Muhr served as the lead developer during the prototype phase from 1989 to 1992, which emerged from post-Chernobyl research involving in-depth interviews and field notes, and was influenced by surveys on qualitative data analysis needs as well as concepts from artificial intelligence, such as semantic networking and annotation.[12] The prototype's relational and non-hierarchical architecture allowed for flexible coding and mapping of segmented texts, addressing the limitations of early computer-assisted qualitative analysis tools.[13] In 1993, Muhr transitioned the academic prototype into a commercial product, founding Scientific Software Development GmbH in Berlin to release the first Windows version of ATLAS.ti, specifically targeting qualitative researchers seeking efficient alternatives to paper-based methods.[2] This marked the software's entry into the market as a dedicated tool for hermeneutically informed analysis of non-numerical data, with Muhr continuing as the primary developer to refine its hypertext capabilities for broader academic use.[12]Key Milestones and Versions
The development of ATLAS.ti traces back to the late 1980s, when the initial prototype emerged from a research project at Technische Universität Berlin aimed at supporting qualitative text analysis. The first commercial version was released in 1993 by Scientific Software Development, founded by Thomas Muhr, initially as a Windows-based tool for grounded theory and hermeneutic analysis.[2][11] In the late 1990s, ATLAS.ti 4.1 was launched in November 1997, expanding support to Windows platforms and introducing enhanced visual qualitative data management capabilities. During the 2000s, the software evolved under ATLAS.ti Scientific Software Development GmbH, with versions like 5.0 in 2003 emphasizing knowledge workbench functionalities for interdisciplinary research, while maintaining a focus on desktop environments for academic and professional users.[14][15] The 2010s marked significant expansions in accessibility and integration. Version 7, released in 2013, introduced mobile app compatibility for iOS and Android devices, enabling on-the-go data collection and synchronization with desktop projects to accommodate field-based qualitative research.[16] Entering the 2020s, ATLAS.ti 22 launched in December 2021, initiating a shift toward cross-platform capabilities by integrating a web-based version alongside traditional desktop applications, facilitating collaborative workflows in remote and team settings. This version laid groundwork for AI enhancements, with subsequent releases like version 23 in 2023 introducing automated coding powered by OpenAI. In September 2024, ATLAS.ti was acquired by Lumivero, LLC, a leader in research software, bolstering innovation in SaaS offerings and global support.[17] As of November 2025, the latest iteration, version 25 released in December 2024, features advanced cloud synchronization and ongoing subscription-based updates, reflecting adaptations to digital humanities trends through improved multimedia handling and global accessibility.[18][19]Core Features
Data Management and Coding
ATLAS.ti facilitates project creation by allowing users to bundle primary documents—such as text files, PDFs, images, audio, video, and geospatial data—into a single project file for organized analysis.[20] Users initiate a project through the Welcome Screen or File menu, then import documents via the Document Manager, where consistent naming conventions and added comments serve as metadata for enhanced searchability across the project.[21] Document groups enable further categorization based on attributes like data type, participant demographics, or thematic relevance, with documents assignable to multiple groups to support flexible retrieval.[20] The software's coding mechanisms support qualitative paradigms such as open, axial, and selective coding, enabling researchers to iteratively develop themes from data. Open coding involves creating initial independent codes to label data segments, while axial coding organizes these into hierarchical categories with subcodes, and selective coding refines the structure by integrating core themes.[22] Free codes represent unused labels available for future application, in vivo codes capture exact phrases from the data itself for literal representation (e.g., highlighting "sustainable lifestyle" to name and apply the code), and code families function as organizational folders grouping related codes thematically to streamline analysis.[23][24] Quotation management in ATLAS.ti centers on selecting and annotating specific segments of data, such as text snippets, image regions, or video clips, which are then linked to one or more codes for thematic association.[25] Through the Quotation Manager, users can review, rename, filter, and retrieve these segments by associated codes, with multi-value coding allowing multiple labels per quotation to reflect nuanced interpretations. Memos provide space for researcher notes, linked directly to quotations, codes, or other memos via drag-and-drop for reflective documentation and contextual elaboration.[26][27] Hyperlinking and networks enhance relational mapping by connecting elements across the project. Hyperlinks create directed relations between quotations (e.g., "explains" or "contradicts"), built by dragging from source to target quotations in the manager or margin area, de-linearizing data to reveal rhetorical connections.[28] Networks allow visual construction of conceptual models, linking codes, documents, quotations, memos, and multimedia via drag-and-drop with predefined or custom relations, increasing node density to illustrate interconnections and support theoretical development.[29] ATLAS.ti supports Unicode encoding, enabling seamless handling of multilingual and non-Western language data without character set limitations.[30] Full-text search extends across all project elements, including documents, codes, and memos, using tools like Project Search for pattern matching, synonyms, and wildcards to locate relevant content efficiently.[31]Analysis and Visualization Tools
ATLAS.ti's analysis and visualization tools enable researchers to query, explore, and present insights derived from coded qualitative data, facilitating the identification of patterns, relationships, and themes without relying on advanced statistical modeling. These tools support exploratory functions such as frequency analysis and overlap detection, as well as output-oriented visualizations for communicating results effectively. By processing quotations, codes, and documents, they help uncover conceptual connections central to methods like grounded theory and content analysis.[32] Query tools in ATLAS.ti include code co-occurrence tables, which display the frequency and overlap of codes applied to the same or adjacent quotations, allowing users to quantify relationships between concepts. For instance, researchers can generate tables showing how often pairs of codes appear together, providing a basis for pattern recognition in thematic analysis. Word clouds visualize word frequencies across selected documents, codes, or quotations, with word size proportional to occurrence, and options to filter by parts of speech like verbs for targeted exploration. Retrieval functions, such as the Quotation Reader or TreeMaps, enable users to fetch and review relevant segments of data based on code frequencies or overlaps, highlighting prominent themes through interactive displays.[33][34][32] Network views offer visual diagrams that map relationships among codes, quotations, documents, and memos, representing them as interconnected nodes to illustrate theoretical frameworks or hierarchical structures. Users can build these networks manually or automatically by adding linked elements, supporting the iterative development of grounded theory models. Hierarchies are depicted via TreeMaps, where the size of rectangles corresponds to the prominence of codes or documents based on quotation counts, aiding in the visualization of data distribution. Sankey charts, used in co-occurrence and code-document analyses, portray flows between entities with link thickness indicating strength of association, such as the magnitude of code overlaps across document groups.[35][32][36] Advanced querying capabilities allow for complex searches using Boolean operators like AND and OR to combine codes with document attributes, enabling precise retrievals such as all instances of "code A AND document type B." This supports sophisticated explorations in content analysis by narrowing results to specific contexts or expanding to related themes. The tools align with grounded theory by facilitating the examination of code interrelations and emergent patterns, while also accommodating content analysis through frequency-based pattern identification.[37][32] The Word Frequencies tool, formerly known as the Word Cruncher, performs frequency-based text mining on selected data scopes, generating lists, clouds, or TreeMaps of word occurrences without full statistical computation. It allows filtering by documents, codes, or quotations to focus on relevant subsets, providing a simple quantitative lens for qualitative text exploration.[38][18] Export options encompass customizable reports in Word, PDF, or Excel formats, capturing analysis outputs like code tables, networks, or query results for publications and team collaboration. Visualizations such as Sankey diagrams or word clouds can be saved as images, while Excel-compatible tables support further processing in external tools for additional charts like scatter plots. These features ensure seamless sharing of analytical insights across projects or with non-ATLAS.ti users.[39][40][32]AI Integration
Automated Coding and Transcription
ATLAS.ti's automated coding features leverage artificial intelligence to accelerate the initial processing of qualitative data, particularly for text and transcripts, by identifying key elements such as sentiments, themes, and entities. The AI Coding tool, powered by OpenAI's GPT models, automatically generates open and descriptive codes, detecting patterns and insights that would otherwise require extensive manual effort.[41] This includes sentiment analysis, which classifies emotions in text as positive, negative, or neutral using natural language processing techniques.[42] Theme detection occurs through AI-driven topic modeling, grouping related content to reveal underlying patterns, while named entity recognition identifies and codes specific elements like names, places, and organizations.[43][44] These capabilities were first introduced with named entity recognition in version 22 (2022) and expanded with full AI Coding in version 23 (2023).[45][18] The software's transcription tools provide automatic speech-to-text conversion for audio and video files, transforming recordings into editable, searchable text that integrates directly into projects for further coding and analysis.[46] This feature supports over 30 languages and dialects, includes automatic speaker detection for multi-participant discussions, and enables synchronized playback to verify accuracy.[18] Introduced in version 25 (December 2024), it processes files up to 20 times faster than manual methods while maintaining high precision, though users often perform minor edits for context-specific nuances.[18][46] Underlying these functions are pre-trained machine learning models optimized for qualitative tasks, such as GPT for coding and advanced AI for transcription, which can be customized through user-defined parameters like fine-tuning sliders and feedback loops to refine outputs over time.[41] AI-generated codes and transcripts serve as suggestions that researchers review and adjust, ensuring interpretive control remains with the human analyst and complementing traditional manual coding approaches.[47] This integration reduces initial workload by up to 90% for coding while preserving the depth of qualitative inquiry.[41]Conversational AI Capabilities
ATLAS.ti's Conversational AI introduces a chat-based interface known as the AI Assistant, enabling users to interact with project data through natural language queries. This feature, powered by OpenAI's ChatGPT technology, allows researchers to ask questions such as "What themes emerge in interview quotes about user experiences?" and receive generated summaries, code suggestions, or insights directly from the project's documents and codes.[48] The AI Assistant integrates seamlessly into both Desktop and Web versions, facilitating dynamic dialogue that accelerates qualitative analysis while maintaining traceability to original data sources.[49] In terms of document interaction, Conversational AI supports direct "conversations" with individual or multiple documents, automatically extracting key concepts, generating codable quotations, and suggesting thematic connections without requiring manual navigation. For instance, users can query a set of documents to identify recurring patterns or contradictions, with the AI providing context-aware responses that include citations for verification. This capability extends to enhanced analysis tools, where AI-powered co-occurrence exploration reveals relationships between codes, and predictive coding proposes new categories based on the project's existing context, building on foundational visualization methods for deeper interpretive insights.[50][51] The feature was rolled out in post-2022 updates, beginning with initial AI integrations in 2023 and evolving into full Conversational AI by version 24.2.0 in September 2024.[18] Regarding ethical considerations, ATLAS.ti positions Conversational AI as a supportive tool rather than a replacement for human expertise, offering an AI Privacy Mode to deactivate OpenAI integrations and ensuring transparency through traceable outputs and optional usage, thereby upholding researcher control and data security.[50][51]Usage and Platforms
Standard Workflow
The standard workflow in ATLAS.ti follows a structured, iterative process designed to facilitate qualitative data analysis from initial project creation to final reporting, emphasizing organization and reflexivity throughout.[37] This approach supports researchers in managing diverse data types while building toward interpretable insights, typically spanning four main phases that align with hermeneutic principles of ongoing interpretation.[20] Phase 1: Project Setup and Data ImportResearchers begin by creating a new project through the File menu or Home tab, which serves as a centralized container for all project elements including documents, codes, and memos.[37] During setup, users define research questions implicitly through planning document organization, such as naming conventions (e.g., prefixing files with interview dates like "IT_090622") and adding initial comments for context.[20] Data import follows via the Add Documents option, supporting a wide range of formats including text files, PDFs, images, audio, video, and geographic data; for instance, transcripts can be uploaded in VTT or SRT formats after preparation to ensure structure like speaker identification and anonymization.[37] Documents are then grouped—e.g., by participant type such as experts versus laypeople—to reflect research objectives and enable targeted analysis later.[20] Phase 2: Exploration and Initial Coding
Once data is imported, exploration involves reading through documents in the Document Manager, using tools like word clouds or word lists to identify recurring themes and patterns.[37] Initial coding proceeds by selecting segments (quotations) and assigning codes via drag-and-drop in the margin area or the Coding Dialogue (shortcut Ctrl+J), building a codebook incrementally with in-vivo codes derived directly from the data or descriptive labels.[37] For multimedia, such as audio transcripts, users segment content and code by speaker using focus group tools, ensuring comprehensive annotation without over-coding to maintain focus on emergent concepts.[20] This phase establishes the foundational code structure, often starting with open coding to capture broad ideas before refinement. Phase 3: Analysis Iteration
Analysis advances through querying and refining, where the Query Tool applies operators like AND, OR, or CO-OCCURS to retrieve code combinations, revealing relationships across the dataset.[37] Codes are iterated by merging duplicates, splitting overly broad ones, or creating smart codes in the Code Manager, while memos capture reflective notes on emerging interpretations.[37] Global filters allow focusing on subsets, such as specific document groups, to test hypotheses iteratively; this hermeneutic cycling—reviewing and recoding after analyzing a few documents—promotes deeper understanding and reduces bias.[37] For team-based work, inter-coder agreement checks using Krippendorff’s alpha (targeting ≥0.800) ensure reliability during iterations.[37] Phase 4: Reporting and Export
The workflow culminates in synthesizing findings for output, generating visualizations like networks, Sankey diagrams, or code co-occurrence tables to illustrate patterns.[37] Reports are exported in formats such as Word, PDF, or Excel via the managers' export options, including codebooks and query results for documentation.[37] Team collaboration is supported through shared project bundles (.atlproj files), allowing secure distribution and merging of contributions while maintaining version history.[20] ATLAS.ti facilitates mixed-methods approaches by exporting coded data—e.g., binarized code-document tables—to statistical software like SPSS via Excel or syntax files, enabling integration with quantitative analysis.[37] Best practices emphasize iterative hermeneutic cycles, where users repeatedly cycle through coding and querying to refine interpretations, alongside rigorous data security measures like password-protected projects and encrypted backups to protect sensitive information.[37] Version control is maintained via snapshots and master project merges, preventing data loss and tracking changes during collaborative efforts.[37] AI tools, such as automated coding, can accelerate initial phases but require user consent and should complement manual review.[37]