Assessment Authoring with AI-Assisted Item Writing

Author Aide is a specialist AI authoring tool designed specifically for question and test creation workflows. It integrates directly into secure item authoring systems and item banks, allowing organizations to generate, review, edit, and save assessment content within a controlled assessment environment.

Author Aide is built by assessment experts and provides AI-generated questions inside a secure authoring environment with integrated item banking workflows, role-based access controls, and human review processes. Questions can be saved directly into item banks after approval, helping organizations streamline assessment development while maintaining governance and security.

Customer prompts, uploaded documents, and generated outputs are not used to train AI models. Uploaded materials are securely stored, and access to Author Aide is protected through authentication and authorization controls.

No. Author Aide does not access learner or test-taker data, and learners do not interact with the tool.

Yes. Users can review, modify, regenerate, discard, or save AI-generated questions before they are added to an item bank or assessment. Human review is recommended before production use.

Yes. Author Aide is a standalone licensed product that can be enabled or disabled independently.

Yes. Questionmark supports integrations with systems including Workday Learning and  Cornerstone OnDemand, along with standards such as LTI, SCORM, xAPI, and AICC.

Author Aide is covered by ISO 27001 certification for information security management.

Item writing is the process of creating assessment questions for exams, quizzes, certifications, and tests. This includes writing question prompts, answer options, scoring guidance, and learner feedback.

Item banking is the process of storing, organizing, managing, and reusing assessment questions in a centralized repository or database.

Item banking helps organizations manage large volumes of assessment content, reuse approved questions, organize assessment materials, and support scalable test development workflows.

Effective test questions are clear, aligned to learning objectives, appropriate for the intended audience, and reviewed for quality and accuracy before use.

Yes. AI-assisted authoring can generate draft assessment questions from prompts and uploaded source materials, helping organizations accelerate assessment development workflows.

Yes. Users can upload PDFs and other materials to help guide AI-generated question creation using Retrieval-Augmented Generation (RAG).

Yes. Author Aide can generate questions individually or in groups of up to 25 at a time.

Yes. Users can control how questions are generated and tailor question complexity in line with Bloom’s taxonomy.

Author Aide supports translation to and from 44 languages.

Author Aide uses multiple large language models and AI services through Microsoft Azure and AWS Bedrock, selecting the most appropriate models for different AI tasks.

No. Customer prompts, uploaded materials, and generated outputs are not used to train AI models.

Customers retain ownership of the content generated by Author Aide to the extent recognized by applicable laws.

Questionmark provides customized pricing based on customer requirements, licensing, and usage needs. Organizations can contact Questionmark directly for pricing information or a product demo.

Author Aide is powered by Learnosity’s API innovation. Specifically, GPT-4 is used for item generation. However, our intention is to be foundation-model agnostic to adopt and adapt models that best fit their intended purpose.

Author Aide brings the innovation without the data privacy risk. Create tests and write prompts safe in the knowledge that your proprietary information remains proprietary to you and is not stored or used as training material for any AI model.

All of our current AI architecture uses LLMs, which at this time rarely cite their sources well or consistently. Because of this, all AI-generated content should be treated as a draft in need of human review – especially if you aren’t using your own specialist content for it to create questions etc. from.

We don’t rely on just one LLM. Instead, we use the best LLM for each specific task. It’s also subject to change as LLMs inevitably get better in terms of affordability or their ability to successfully handle source material better. When choosing the best LLM for a task, we always adopt a data-driven approach, often comparing AI functionality to human ability. 

Yes, selecting the Bloom’s Taxonomy option makes a huge difference in what questions you can generate in terms of complexity.

By design and in its current state, Author Aide isn’t automatically fed your items and therefore can’t know if it’s duplicating content. As part of our ongoing AI work, we are looking at the possibility of using AI to search for duplicate items.

Whether you’re interested in a demo or just have general questions, you can reach out to us using this form and we’ll be happy chat.