Relativity Generative AI Pro
Study Guide
RelativityOne 2025
Last modified: October 1, 2025
Contents
Exam Overview..................................................................................... 3
Exam description......................................................................... 3
Study guide format...................................................................... 3
Study guide and exam updates................................................. 3
Exam Details.......................................................................................... 4
Exam Structure............................................................................. 5
Study Resources................................................................................... 6
About Generative AI.................................................................... 7
Study Links................................................................................... 8
Exam Items............................................................................................ 9
Exam Items................................................................................... 10
Content Distribution Disclaimers....................................................... 14
Change Log........................................................................................... 15
2
© Relativity. All rights reserved.
Exam Overview
This document is designed to be a study guide for the Relativity Generative AI Pro exam. Using this study guide, you can
take a deeper dive into the key points of Generative AI and prepare with exam questions and additional resources.
Exam description
The Relativity Generative AI Pro exam is designed for individuals who are curious about and want to learn more about
Relativity’s approach to Generative AI, overarching AI principles, and when the Relativity aiR suite can be utilized in various
use cases. Anyone who has an interest in applying AI-powered problem solving to the legal industry can benefit from
achieving this certification.
Study guide format
This study guide is divided into three parts:
• Exam Details: This section contains information about the exam structure and format and instructions on how to prepare
for the exam.
• Study Resources: This section contains study content and links that are crucial to both understanding the product and
successfully passing the exam.
• Exam Items: This section includes all exam items. Use the study resources and links to find and identify correct answers.
Study guide and exam updates
The Relativity Generative AI Pro exam currently tests on RelativityOne. Review the Exam Update Schedule for upcoming
exam changes. Plan accordingly when studying for the exam.
To view changes made to the exam from one version to the next, review the Change Log.
3
© Relativity. All rights reserved.
Exam Details
Exam Structure
The following table outlines the exam objectives.
About Generative AI
Using and describing generative AI basics
Understanding Relativity AI principles
aiR Features & Functionality
Recognizing and comparing benefits of the aiR
products
Articulating Relativity’s aiR privacy and security
approach
Using aiR products responsibly and quality
checking with human review
Understanding technical considerations and best
practices
Understanding foundations of prompt criteria
Use Cases
Identifying use cases for each aiR product
Exam Format
You will be tested on 45 multiple choice questions covering the objectives listed in the table. All questions are multiple
choice. You will have 35 minutes to answer the questions, and you must achieve a score of 80% or higher to pass.
Exam Preparation
There are no prerequisites for this certification; however, in order to prepare, we recommend following:
• Read and understand this Study Guide. This guide contains the necessary materials to prepare for the exam, including
topics, key points, resources, and exam items.
• Review the RelativityOne documentation. The help site is a detailed and useful resource for the exam.
• Familiarize yourself with the on-demand and live training content. We offer a variety of training resources to enhance
your knowledge, which can be found on Relativity Learning.
5
© Relativity. All rights reserved.
Study Resources
About Generative AI
Generative AI and aiR
Generative AI is an exciting new realm of artificial intelligence that takes text as input in order to create an output. Relativity
aiR products leverage the power of generative AI to expedite processes in RelativityOne while keeping data secure and
providing trustworthy and validated results. Each product was developed specifically to meet a legal challenge with detailed
explanations of every output decision.
Relativity aiR
aiR for Review
aiR for Review quickly and efficiently performs document review powered by generative AI. aiR for Review utilizes a
large language model (LLM) to predict responsive documents, locate specific legal issues, and pinpoint key documents
at the start of a case. It can then explain the detailed rationale behind its decisions by providing excerpts from identified
documents. Consistent recommendations reduce the coding discrepancies that come with traditional document review.
• Types of aiR for Review analysis:
− Relevance review: analyzes whether documents are relevant to a case, such as documents responsive to a
production request.
− Issues review: locates material relating to different legal issues.
− Key documents: finds key documents important to a case.
• Typical aiR for Review workflow:
1. Select the documents to review.
2. Create an aiR for Review project and choose analysis type.
3. Write and submit prompt criteria per Relativity best practices. Prompt kickstarter can optionally be used to create
prompt criteria from up to five existing case documents.
4. Run the analysis and review the results, which consists of citations, rationale, and considerations.
aiR for Privilege
aiR for Privilege analyzes documents to consistently identify them as privileged during a first-pass privilege review while
citing excerpts from each document. Its provided data points can drive quality control (QC) checks, final review decisions,
and privilege log creation. The tool utilizes information such as social network graphs and external profiles to develop a
more complete understanding of the individual communicating and their role/privilege status.
aiR for Case Strategy
aiR for Case Strategy extracts critical facts, a timeline of events, and context of all key aspects of case to build a complete
case narrative to more efficiently prepare for depositions and trial. After document review, this tool can immediately begin
strategizing to present a clear overview of the story of a case.
Privacy and Security
• aiR products are integrated with Microsoft Azure’s OpenAI services. All data stays secure within RelativityOne, and any
customer data used for analysis is never retained by Relativity or Microsoft.
• Throughout every component of aiR, Relativity’s AI Principles are put into practice to ensure confidence and
trustworthiness in leveraging these powerful generative AI tools.
7
© Relativity. All rights reserved.
Study Links
Use the following links to prepare for the exam by locating answers for the exam items in the following section. We
recommend using the Status column to mark each link as complete as you study.
Link Status
Relativity Website Pages
• Transform e-discovery with AI
• Relativity aiR
• Relativity’s AI Principles
• aiR for Review
• aiR for Privilege
• aiR for Case Strategy
Relativity Documentation
• aiR for Review
• Best practices
• Creating an aiR for Review project
• Developing Prompt Criteria
• Using prompt kickstarter
• Exporting Prompt Criteria
• Running the analysis
• Analyzing aiR for Review results
• aiR for Review Prompt Criteria validation
• Review Center
• aiR for Privilege
Relativity Community Resources
• From Beginning to Breakthrough: Navigating Document Review’s AI Evolution
• A Focus on Security and Privacy in Relativity’s Approach to Generative AI
• Workflows for Applying aiR for Review
• Selecting a Prompt Criteria Iteration Sample for aiR for Review
• The AI Advantage: Building Responsible & Defensible GenAI Solutions with Relativity aiR
Relativity Learning Courses
• AI Fluency: AI and Relativity Products
8
© Relativity. All rights reserved.
Exam Items
Exam Items
This section includes all exam items on the Relativity Generative AI Pro exam. You will receive the 45 questions below on
exam day. Use the Study Links to locate and identify the correct answer to each question. In order to maintain the integrity
of the exam, Relativity will not share or confirm answers to any of the questions.
Note: Exam questions are subject to change without notice and in Relativity’s sole discretion. All question changes will
be reflected in the items below, as well as in the Change Log.
As an additional study tool, the Generative AI Pro Practice Quiz is available in Relativity Learning. You can use it to review
the items below.
About Generative AI
1. What are hallucinations? 5. What does grounding refer to?
a. Results in which a large language model (LLM) corrects a. Preventing a prompt from running because its content
incongruencies in the prompt. exceeds the limits of the context window.
b. Results in which a large language model (LLM) ignores b. The practice of requiring an AI model to
questions asked in the prompt. provide responses based on verifiable source information
to mitigate hallucinations.
c. Results in which a large language model (LLM) presents
factually incorrect information. c. The basic training or instructions given to an AI model in
a particular subject.
2. Which of the following is an advantage of generative AI
compared with other types of machine learning AI? 6. What is Relativity’s approach to privacy in AI product
development?
a. Generative AI does not need any additional model
training. a. Relativity prioritizes implementing the latest technologies
in privacy.
b. Generative AI is always faster at making predictions.
b. Relativity implements privacy initiatives after product
c. Generative AI is less computationally expensive. development to ensure compatibility.
c. Relativity ensures privacy is prioritized and protected for
every AI product.
3. How is a Support Vector Machine (SVM), the underlying AI
model of Review Center prioritized queues, different than
generative AI?
7. According to the Relativity AI principles, what is Relativity’s
a. An SVM uses example documents instead of review approach to security?
instructions to classify documents.
a. Relativity embeds independent security measures into
b. An SVM evaluates documents independently and does every layer of the design process.
not learn from previous documents.
b. Relativity relies on single security measures to ensure
c. An SVM is pretrained on billions of documents from open consistency throughout development.
databases and the web.
c. Relativity focuses on contractual measures for ensuring
privacy.
4. What are large language models (LLMs)?
a. AI models that interpret and generate responses based 8. Which AI principle highlights Relativity’s commitment to
on human language. ethical AI applications and mitigating bias?
b. AI models that are specifically trained to generate binary a. Building AI with purpose that delivers value for our
classifications for a given task. customers.
c. AI models that can predict future trends in continuous b. Ensuring fairness is front and center in our AI
data like metrics over time. development.
c. Placing the security of our customers’ data at the heart of
everything we do.
10
© Relativity. All rights reserved.
9. Which of the following is a primary purpose behind 15. Where is customer data retained when using aiR products?
Relativity’s AI principles? a. Microsoft Copilot.
a. To comply with FedRAMP Authorization requirements for b. Within the RelativityOne boundary.
AI-based products.
c. OpenAI GPT platform.
b. To emphasize our commitment to customers and guide
responsible product development.
c. To replicate best practices from AI-leading companies. 16. What content does Relativity permit to be scanned by
Microsoft Azure OpenAI Service’s abuse monitoring and
content filtering system?
aiR Features & Functionality
a. Relativity has opted out of Azure’s abuse monitoring and
10. What three analysis types or predictions can aiR for Review content filtering system.
make?
b. Audio and visual content.
a. Relevance, Key documents, and Issues
c. Text content.
b. Relevance, Key documents, and Privilege
c. Issues, Key documents, and Privilege
17. Which of the following is the current OpenAI LLM model that
powers aiR for Review?
11. What is a benefit of aiR products using a large language a. GPT-4 Turbo
model (LLM) through Microsoft’s Azure OpenAI service? b. GPT-4 Sora
a. Customer data can be used to improve the model for c. GPT-4 Omni
future projects.
b. Microsoft can read customer data to ensure that aiR
products are not used in a harmful way. 18. What data is retained by the large language model (LLM)
c. Customer data stays secure within the logical boundary when using aiR products?
of Microsoft Azure instead of having to be exported out a. Prompt data is retained and used by the large language
for analysis in a different location. model (LLM) to enhance products and models.
b. No data is retained by the large language model (LLM).
12. What is one advantage of using aiR for Review to review c. Document data is retained and used by the large
documents over other AI review methods? language model (LLM) to enhance products and models.
a. aiR for Review provides transparent responses to verify
the review predictions and provide additional value to
downstream legal work. 19. What is a responsible approach to integrating aiR for Review
with human quality control?
b. aiR for Review makes a binary (e.g., relevant or not
relevant) decision instead of providing multi-level scoring. a. Overriding human review decisions whenever aiR for
Review coding results conflict.
c. aiR for Review learns from human coding decisions to
improve its predictions that can be used on other projects b. Analyzing metadata across documents coded by aiR for
in the future. Review.
c. Checking human reviewer decisions against aiR for
Review results for conflicting coding decisions.
13. How do aiR for Review and aiR for Case Strategy use
generative AI to review and analyze documents?
a. They use a sample set of documents to train the model. 20. How does aiR for Review enhance the QC process in
b. They use prompt criteria provided by a subject matter traditional relevance review?
expert to review and analyze documents. a. By deciding which documents are subject to QC.
c. They use a subject matter expert’s coding decisions to b. By allowing senior reviewers to resolve coding conflicts
train the model. faster.
c. By removing humans entirely from the QC process.
14. What customer data is retained by Microsoft when using aiR
products? 21. Why is testing and iterating prompt criteria recommended
a. Document data is retained and used by Microsoft to before running aiR for Review?
enhance products and models.
a. To increase and predict the accuracy of aiR for Review’s
b. Prompt data is retained and used by Microsoft to analysis.
enhance products and models.
b. To eliminate the need for QC in the later stages of the
c. No customer data is retained by Microsoft. review.
c. To disclose the methods and assessment of the AI to
opposing parties at the outset.
11
© Relativity. All rights reserved.
22. What is the purpose of a post-review validation? 28. What training is required to write prompt criteria in aiR
a. Preventing new legal issues from being introduced when products?
documents are added to the workspace. a. Extensive experience working with ChatGPT on a variety
b. Proving the AI classifier functions correctly before using it of tasks.
to review documents. b. An understanding of how to write clearly and concisely is
c. Verifying and measuring the accuracy of the review coding all that is required.
decisions. c. A strong understanding of the special syntax that large
language models (LLMs) expect.
23. How does Relativity suggest aiR for Review be configured to
mitigate bias during prompt validation? 29. What purpose does the Case Summary section serve within
a. Human reviewers should see aiR predictions while aiR products’ prompt criteria?
coding, but be advised to not rely on them as they may be a. It displays a two-page summarization, written by AI, of all of
inaccurate. the documents in the Relativity workspace.
b. Human reviewers should not see the aiR predictions while b. It provides a place to define key people, organizations, and
coding. terms, as well as give a general overview of the matter.
c. Human reviewers should only be able to see the c. It provides an adaptive question and answer interface,
issue predictions, because those are not as critical as where the user answers questions about the facts of the
responsiveness or key predictions. case.
24. Which of the following documents would be most appropriate 30. In addition to the Relevance Field and Relevance Choice,
to run in aiR for Review? what is required when setting up an aiR for Review project
a. A 200-page email thread with 30KB of extracted text for Relevance or Relevance and Key Documents analysis
types on the Relevance tab?
b. A relevant brief with 300KB of extracted text
a. The criteria of what makes a document relevant.
c. A collection of compressed 50KB JPEG images
b. At least ten coded documents to serve as examples.
c. A multi-choice field that represents the issues in the case.
25. What is a key component of the prompt criteria development
phase?
a. Running your prompt criteria across a larger population. 31. What is a common cause of documents being erroneously
predicted as Not Relevant by aiR for Review?
b. Refining the prompt criteria on small sets of documents.
a. Prompt criteria did not define key jargon, project
c. Validating the prompt criteria across a large set of codenames, nor internal company terminology.
randomized documents.
b. The documents were more than one page, and aiR for
Review only works on one-page documents.
26. What is Relativity’s recommended approach to identifying a c. The definition of relevance in the prompt criteria was
population for developing your prompt criteria? written too broadly, causing overinclusive results.
a. Choosing a small sample which represents the documents
of your matter.
32. What is an advantage of testing prompt criteria on smaller sets
b. Selecting statistical samples across your document of documents?
population for validation.
a. Your results are more detailed, with longer rationale and
c. Providing all email thread duplicates and non-inclusives to considerations.
confirm consistency.
b. aiR for Review is more accurate when run on fewer than 50
documents.
27. What is the recommended workflow for using aiR for Review? c. You can iterate quickly, improve the prompt criteria, and
see the impact.
a. Run on a larger population, develop the prompt criteria,
and validate on a statistical sample.
b. Validate on a statistical sample, develop the prompt 33. What is the maximum number of issue choices that can be
criteria, and run on a larger population. added to an aiR for Review Issues analysis job?
c. Develop the prompt criteria, validate on a statistical a. 5
sample, and run on a larger population.
b. 10
c. 15
12
© Relativity. All rights reserved.
34. Why does aiR for Review maintain past job results even after Use Cases
re-running?
40. Which of the following is a use case of aiR for Review?
a. To allow for version comparison and rollback.
a. Grouping documents into a tree hierarchy based on
b. To reduce costs of running already analyzed documents. conceptual content.
c. To reduce compute usage. b. Answering questions interactively over a collection of
documents.
35. You have case documents that you want to leverage to build c. Determining which documents in a collection contain four
prompt criteria for an aiR for Review project. What is the important issues.
fastest way to accomplish this?
a. Use Copy Existing to import existing prompt criteria from
41. Which aiR product is able to identify and create fact objects
another aiR for Review project.
that it finds in documents?
b. Use prompt kickstarter’s Draft with AI button to build the
prompt criteria based on your existing case documents. a. aiR for Review
c. Copy your review protocol document and paste it into the b. aiR for Case Strategy
Matter Overview field. c. aiR for Privilege
36. What are the limitations of using aiR for Review’s prompt
42. How can aiR for Review be used to identify and correct
kickstarter when creating prompt criteria?
problems in your outgoing production?
a. You can upload up to 5 files in .DOCX format only with no
a. By applying issues analysis on the other party’s production.
size limit.
b. By applying issues analysis post-production.
b. You can upload up to 3 files with a maximum total size of
50MB. c. By applying relevance analysis pre-production.
c. You can upload up to 5 files with a maximum total
character count of 150,000.
43. When QCing aiR’s coding decisions, how can you locate
additional material related to legal issues and gain case
37. How can you collaborate on building and iterating prompt knowledge early?
criteria with other stakeholders that have no access to your
a. By instructing aiR to re-draft the prompt criteria based on
RelativityOne workspace? aiR’s relevant coding decisions.
a. Add the names of the stakeholders in the People and b. By checking for documents that aiR predicted as relevant,
Aliases field of the Case Summary tab. but which may have been missed by human reviewers.
b. Use Import/Export to export the prompt criteria as a PDF. c. By removing information from the first level review based
c. Export your prompt criteria to a MS Word file from on aiR’s non-relevant coding decisions.
the Prompt Criteria panel.
44. Which of the following use cases are not recommended for
38. How should legal jargon and complex legal phrases be
aiR for Review?
handled in the prompt criteria for an aiR for Review job?
a. Locate material relating to different legal issues.
a. Quote relevant statutes and include “any and all” to cover
every legal nuance in the prompt. b. Identify privileged communications within a dataset.
b. Avoid using legal jargon or filler phrases and stick to c. Find hot documents important to a case or investigation.
simple terminology.
c. Use extensive legal terminology to ensure understanding
45. Which aiR for Review use case requires formal validation?
of the formal context of the case.
a. Review for production
b. Early case insights
39. Which metric is shown in the aiR for Review job confirmation
modal? c. Opposing party production analysis
a. Estimated Total Review Time
b. Estimated Per Document Cost
c. Estimated Time to Start
13
© Relativity. All rights reserved.
Content Distribution Disclaimers
Relativity is aware that a number of apps are available to the public which purport to be study guides for various Relativity
exams. These apps are published by third parties that are not affiliated with Relativity, and Relativity does not monitor their
content and cannot ensure its accuracy. As such, you should only use Relativity-approved study materials. Your use of
any other third party study resources is at your own risk. For approved Relativity study resources please visit the Relativity
Community or contact certification@[Link] if you have any questions.
Relativity may, in its sole discretion, provide certain materials, including but not limited to workbooks, presentations,
outlines, webinars, practice examinations, on demand videos and tutorials, and study plans (collectively, the “Materials”),
as part of its training, certification, and implementation services. The Materials are unique to each training session and may
only be used by individuals who have attended the training session in which they were distributed (“Attendees”). Attendees
may request access to Materials that were not distributed during a training session, and Relativity may grant such access
in Relativity’s sole discretion. Attendees shall not redistribute any Materials to any third party or to any person who has not
attended the training session in which the Materials were distributed without Relativity’s prior written authorization.
14
© Relativity. All rights reserved.
Change Log
1. July 8, 2024 - Posted document.
2. February 5, 2025 - Update to Question 11.
3. April 22, 2025 - Added additional Study Links.
4. August 11, 2025 - Study Link table updated. Total exam length updated to 45 questions. Questions 24, 27, and 30
updated. Questions 17 and 33-39 added. Generative AI Pro Practice Quiz updated to reflect exam content updates.
5. October 1, 2025 - Study links updated on page 8.
15
© Relativity. All rights reserved.