User Guide
Learn how to create pedagogically sound exam questions with AI using TentaGen
Why TentaGen?
Generate pedagogically sound exam questions in minutes instead of hours. AI handles the time-consuming work of creating varied and well-formulated questions.
AI-generated questions based on your specific course material. Review and edit all questions before export to ensure perfect alignment with your needs.
Support for multiple export formats: Wiseflow JSON (New/Legacy), QTI 2.1, QTI 2.2, and Word (.docx). WISEflow and Blackboard have accepted the QTI 2.1 file in our tests.
Fill basic info
Enter subject, topic, cognitive level, and number of questions
Add context (optional)
Upload PDF, Word or PowerPoint files, images or web addresses, or paste a YouTube link, for questions from specific material
Generate & Export
Review questions, save to library, and export to Wiseflow JSON, QTI, or Word
Main Features
Quick Navigation
Basic Fields
- Subject: The main subject for questions (e.g., Biology, Mathematics)
- Topic: Specific area within the subject (e.g., Cell division, Algebra)
- Cognitive Level: Choose a pedagogical framework and one or more cognitive levels. Default: NBME (medical gold standard) at the Application of Knowledge level. NBME has three subsections: (1) Cognitive Level — Recall or Application; (2) Task Competency — 10 medical domains, multi-select or 'All'; (3) Vignette Structure — 5 formats, multi-select or 'All'. Also supports Bloom's (simplified/full), SOLO, and Feisel-Schmitz. Multiple selections are distributed evenly across generated questions.
- Question types auto-filtered: When NBME or Feisel-Schmitz is selected, incompatible question types are automatically disabled (e.g. NBME focuses on SBA format — choice matrices and cloze types are hidden). An info note appears below the question type selector when filtering is active.
23 Question Types
⭐ Core
📚 Extended
🔬 Specialized
🖼️ Image-based
Enable in Settings. Requires image upload — you can upload images directly in the preview or library after generation.
Generation is output-bound: the time depends almost entirely on how much text the model writes, not how much it reads. Three choices affect speed the most:
- Cognitive level: Biggest effect. Recall writes no clinical vignettes and is ~5× faster than Application, which builds a full vignette per question.
- Number of questions: Time scales roughly linearly — twice as many questions takes about twice as long.
- Framework: SOLO and full Bloom's require elaborate per-question rationales (solution paths) and are slower than NBME Recall or simplified Bloom's.
Examples (measured)
- 2 MCQ, NBME Recall: ~5 seconds
- 2 MCQ, NBME Application (vignette): ~23 seconds
For the fastest generation (e.g. a demo): choose MCQ, NBME Recall, and 1–2 questions. Per-question quality is unaffected by these choices — only how much extra text (vignette/rationale) gets written.
Supported File Formats
- PDF: documents with text (a scanned PDF: upload its pages as images, JPG or PNG)
- Word: .docx documents
- PowerPoint: .pptx presentations
- Images: JPG, PNG, WEBP, GIF — AI vision extracts all text and describes diagrams, tables, figures, and other visual content
- Video: YouTube links in the form youtube.com/watch?v=… or youtu.be/… are read through the video's captions when you paste them in the video section, not under Web URLs. Video upload is temporarily paused.
- Web: Addresses of web pages, such as Wikipedia and articles: the page's text is read. Download a file in one of the formats above and upload it. Paste YouTube links in the video section.
Limits: up to 50,000 characters of each document, image and web page are used, about 12–18 pages of running text. Longer text is cut to its start, and a warning says, when it is read, what share is used. A video's captions or transcript are used whole. The material and the text from the Course plan and frameworks box, that is the ticked documents and anything you pasted there, may come to at most 300,000 characters per generation together. Above that, Generate Questions waits until you remove or untick something. A book works best split into chapters of about 15 pages.
Context Priority
When uploading documents or URLs, choose how AI should handle context:
- Respect Subject & Topic:AI can create questions from its own knowledge about the subject AND from your uploaded material. You may get questions about things not in the material.
- Prioritize Uploaded Context:AI creates questions only from your uploaded material — nothing else.
- Hybrid:AI creates questions from your uploaded material, but uses the subject to find matching content in the material. Good for larger uploads with mixed topics. Selected automatically when you upload material.
Under Course plan and frameworks in the form you upload or paste the course plan, the learning objectives, national guidelines or a framework. The documents are kept apart from the material the questions are made from.
The AI is instructed to write questions that test the learning objectives in the documents, or that keep within what the documents cover when they have none. If the documents say something different from the AI's own knowledge base, the AI is instructed to follow the documents. Put the course plan and learning objectives in the box Course plan and frameworks, and add chapters or presentations as source material. The AI is then instructed to aim the questions at the learning objectives and take its facts from the source material. The mode you choose in the form decides whether the AI may also use its own knowledge base. Without source material, the AI takes its facts from its own knowledge base.
When you make questions with documents that have learning objectives or headings, the AI is instructed to state for each question which learning objective or area it tests. You see it in the box Learning objective or area on the question card, both in the preview and in the library. A question without a learning objective or area has no such box. Press the pencil on the question card if you want to change or remove the learning objective, or write your own.
The documents stay with the questions in this tab, as the material does. Saved documents stay in your account until you remove them.
The learning objective in the export
- New Wiseflow JSON and Legacy JSON: a tag of its own on the question.
- QTI 2.2: in the question's metadata, as a learning objective.
- Word: the line Learning objective or area, before the grading guide.
- CSV: the column Curriculum Reference.
- QTI 2.1: the learning objective is not included.
Fact-check against Swedish or international sources
When Fact-check searches the web you first choose which sources the question is checked against: Swedish medical sources (Socialstyrelsen, SBU, Läkemedelsverket, FASS, Janusinfo, Folkhälsomyndigheten, Vårdhandboken, Läkartidningen, Internetmedicin and 1177), international sources, or both. If you choose Swedish sources only and no Swedish source is used, the verdict is Uncertain. If the questions were made with Hybrid and the material is still in the tab, the verdict stands, and the card says that no Swedish source was used or that no search was made. Questions made with Prioritize Uploaded Context are checked against your material at once while it is in the tab, without a web search and without a choice of sources. In the library, and in a tab without the material, the check works as for other questions, with a choice of sources. Your last choice is preselected next time in the same tab.
Automatic Tags
The following tags are added automatically to all questions:
Custom Tags
- Term/Period: E.g., T3
- Semester: E.g., HT25
- Exam Type: E.g., Regular
- Course Code: E.g., BIO101
- Additional Tags: Comma-separated (e.g., Eyes, Macula, LO1)
Filter & Sort
- Filter questions by tag, type, difficulty, or date
- Sort by newest or oldest first
- Edit points and difficulty directly in library
Edit Questions
- Add or remove tags (including TentaGen tag)
- Edit question text, answer options, and correct answer
- Adjust points (1-10) and difficulty level
Each question has a review status to help you track quality assurance:
Default status for new questions. Not yet reviewed.
Question has been reviewed but awaits final approval.
Question is quality-assured and ready for use in exams.
How to use it
- Click the status badge next to a question in the library to cycle between statuses
- Select multiple questions and change status for all at once via the toolbar
- Status is saved automatically with timestamp and user information
Wiseflow JSON Formats
- New Wiseflow JSONCurrent format with labels that have ID fields. Recommended for most exam centers.
- Legacy JSONOlder format with tags array instead of labels. Used by older exam centers.
- QTI 2.1Blackboard's QTI 2.1 dialect; WISEflow and Blackboard have accepted the file in our tests. Multiple choice, true/false, multiple response, fill-in-the-blank and free text arrive with their answer key; matching, ordering, choice matrix, cloze questions with dropdowns or drag-and-drop and image questions arrive as free-text questions without a key, and you are warned before the export.
- QTI 2.2QTI 2.2 format, built against the standard's published schema.
- Word (.docx)Professional Word document with questions, answer options, correct answers, and grading guides. Perfect for printing, review, or archiving.
Tip
You can choose export format both in the preview after generation and in the library. Word export includes TentaGen branding, metadata, and grading guides.
How questions that TentaGen has written are marked
When you export questions that TentaGen has written, every file states that they are AI-drafted. The marking is never part of the question text itself. In WISEflow each question gets the label AI-drafted or AI-utkast. The New Wiseflow JSON format also puts it on a line under the question's title. In WISEflow, students see neither the label nor the line. In the QTI files the marking is among the file's details about the questions, and in QTI 2.2 each question also gets the label. In the Word files it is on a line of its own and in the file's properties, and in the CSV file in the last column. When you convert questions to WISEflow, TentaGen adds no marking. Tags already in the original file are kept.
Profile Settings
- Tutor Initials: E.g., AB, JD, or id:pma. Automatically added as a tag to all your questions.
- UI Language: Choose between Swedish or English for the interface (doesn't affect question language)
Default Question Types
- Choose which question types should be pre-selected when generating questions. Saved automatically and loaded next time you log in.
- All 23 question types organized in three tiers: Core, Extended, and Specialized.
EU AI Act
Our assessment under the EU AI Act (2024/1689): TentaGen is a preparatory AI tool under Article 6(3) and not high-risk, since no questions automatically affect grades or admissions.
- Marking (Article 50) — questions TentaGen has written are marked as AI-drafted in every export file
- Human oversight (Article 14) — all questions require teacher review
- Review status (Draft → Reviewed → Approved) ensures quality control
Read more on our EU AI Act page.
Privacy & GDPR
- GDPR-compliant data processing with clear legal basis (Article 6.1)
- Cookie consent with option to accept or decline
- Uploaded files are deleted from our storage once their text has been read; images that belong to questions are stored on our servers
- Your material is not used to train OpenAI's or Anthropic's models (under their API terms)
See our privacy policy and terms of service for complete details.
For Best Results
- •Be specific in subject and topic
- •Upload PDF, Word or PowerPoint files or web addresses, or paste a YouTube link, for more focused questions
- •Use 'Additional Context' for special instructions, and 'Course plan and frameworks' for learning objectives, the course plan and guidelines
- •Always review questions before export - AI can make mistakes
- •Use tags consistently to easily find and organize your questions
- •Save all generated questions to the library for future use
- •Select the NBME framework with multiple cognitive levels (Recall + Application) for a clinically grounded mix — each question is automatically assigned the correct level badge
- •Configure default question types in Settings to save time on each generation