Wednesday , November 18 2026
Breaking News

ChatGPT and Prompt Engineering Professional Certification Exam Answers 2026 (All 100 Questions)

ChatGPT and Prompt Engineering Professional Certification exam answers 2026: This page lists all 100 questions of the official ChatGPT and Prompt Engineering Professional Certification exam by Itronix Solutions, each with its four answer options and the correct answer highlighted in green. Use it to revise every topic before you attempt the test and earn your verifiable certificate.

About this certification: Prove your skills in writing effective prompts for ChatGPT and other AI assistants: roles, context, examples, formats and iteration.

Certification ChatGPT and Prompt Engineering Professional Certification
Provider Itronix Solutions
Category AI & Machine Learning
Total questions 100 multiple-choice questions
Marks 1 mark per question (100 marks)
Passing score 80% (80 out of 100)
Retake Available if you do not pass
Certificate Verifiable online certificate

Take the exam: ChatGPT and Prompt Engineering Professional Certification on Itronix Solutions. See every other Itronix exam in our complete list of Itronix certification answers.

ChatGPT and Prompt Engineering Exam Questions and Answers

Questions 1 to 25

1. What is prompt engineering?

A. Designing clear inputs to get better outputs from AI models ✔
B. Repairing network cables
C. Writing antivirus software
D. Building computer hardware

Correct answer: A. Designing clear inputs to get better outputs from AI models

2. Which prompt is likely to give the best result?

A. Write a 150-word product description for a steel water bottle aimed at college students, in a friendly tone ✔
B. Tell me stuff
C. Write something
D. Bottle

Correct answer: A. Write a 150-word product description for a steel water bottle aimed at college students, in a friendly tone

3. Few-shot prompting means:

A. Giving the model a few examples of the desired input and output ✔
B. Asking only a few questions per day
C. Sending the prompt few times
D. Using a small screen

Correct answer: A. Giving the model a few examples of the desired input and output

4. Zero-shot prompting means:

A. The model is offline
B. The model gets no instructions
C. The prompt is empty
D. Asking the model to do a task without giving examples ✔

Correct answer: D. Asking the model to do a task without giving examples

5. Assigning a role such as You are an experienced HR manager helps because:

A. It is required for login
B. It sets context, tone and expertise level for the answer ✔
C. It makes the answer always correct
D. It changes the model’s hardware

Correct answer: B. It sets context, tone and expertise level for the answer

6. Chain-of-thought prompting encourages the model to:

A. Use only emojis
B. Answer in one word
C. Ignore the question
D. Work through the problem step by step ✔

Correct answer: D. Work through the problem step by step

7. If the AI’s answer is too long, the best next step is to:

A. Ask it to shorten the answer to a specific length or number of points ✔
B. Restart the computer
C. Repeat the same prompt exactly
D. Close the app

Correct answer: A. Ask it to shorten the answer to a specific length or number of points

8. Specifying an output format like table or bullet list helps to:

A. Get structured, easy-to-use responses ✔
B. Slow down the model
C. Change the model version
D. Reduce accuracy

Correct answer: A. Get structured, easy-to-use responses

9. What should you avoid putting in prompts on public AI tools?

A. Word limits
B. Passwords and sensitive personal or company data ✔
C. Examples
D. Clear instructions

Correct answer: B. Passwords and sensitive personal or company data

10. Iterative prompting means:

A. Using many accounts
B. Refining the prompt step by step based on the outputs ✔
C. Typing very fast
D. Using one prompt and never changing it

Correct answer: B. Refining the prompt step by step based on the outputs

11. Delimiters such as triple quotes in a prompt are used to:

A. Encrypt the message
B. Hide the prompt
C. Clearly separate instructions from the content to be processed ✔
D. Decorate the text

Correct answer: C. Clearly separate instructions from the content to be processed

12. A system prompt or custom instruction is used to:

A. Change the font
B. Set overall behavior and rules for the assistant ✔
C. Delete chat history
D. Install the software

Correct answer: B. Set overall behavior and rules for the assistant

13. Which phrase adds a useful constraint to a prompt?

A. Whatever you think
B. Be nice
C. Do your best
D. Use simple language suitable for a 12-year-old ✔

Correct answer: D. Use simple language suitable for a 12-year-old

14. Why should you verify facts in ChatGPT responses?

A. Because verification is illegal
B. Because it only speaks French
C. Because it never answers
D. Because models can produce incorrect or outdated information ✔

Correct answer: D. Because models can produce incorrect or outdated information

15. Providing background context in a prompt usually leads to:

A. No change ever
B. A system crash
C. Less relevant answers
D. More relevant and specific answers ✔

Correct answer: D. More relevant and specific answers

16. Which is a good prompt for learning?

A. Code
B. Explain recursion with a simple real-life analogy and one Python example ✔
C. Tell
D. Recursion

Correct answer: B. Explain recursion with a simple real-life analogy and one Python example

17. Asking the AI to ask you clarifying questions first helps when:

A. You want random output
B. The internet is down
C. You want no answer
D. The task is unclear or needs more details ✔

Correct answer: D. The task is unclear or needs more details

18. Negative instructions like Do not use jargon help to:

A. Turn off the AI
B. Translate the text
C. Exclude unwanted styles or content ✔
D. Add more jargon

Correct answer: C. Exclude unwanted styles or content

19. A prompt template is useful because it:

A. Blocks the model
B. Deletes old prompts
C. Can be reused for similar tasks with consistent results ✔
D. Works only once

Correct answer: C. Can be reused for similar tasks with consistent results

20. What is the main goal of prompt engineering?

A. Making prompts as long as possible
B. Avoiding AI completely
C. Using difficult words
D. Getting accurate, relevant and useful outputs efficiently ✔

Correct answer: D. Getting accurate, relevant and useful outputs efficiently

21. ChatGPT is developed by:

A. Meta
B. OpenAI ✔
C. Anthropic
D. Google

Correct answer: B. OpenAI

22. ChatGPT is based on:

A. Decision trees
B. Spreadsheets
C. Rule-based expert systems only
D. Large language models from the GPT family ✔

Correct answer: D. Large language models from the GPT family

23. GPT stands for:

A. Global Prediction Tree
B. Graphical Processing Tool
C. Generative Pre-trained Transformer ✔
D. General Purpose Technology

Correct answer: C. Generative Pre-trained Transformer

24. A prompt is:

A. The model’s training data
B. A database
C. The input instruction or question given to an AI model ✔
D. A GPU

Correct answer: C. The input instruction or question given to an AI model

25. Prompt engineering is:

A. Writing HTML
B. Training models from scratch
C. Building GPUs
D. Designing inputs to get more useful and reliable outputs from AI models ✔

Correct answer: D. Designing inputs to get more useful and reliable outputs from AI models

Questions 26 to 50

26. A clear prompt usually includes:

A. Random keywords
B. Emojis only
C. The task, context, constraints and desired format ✔
D. Only one word

Correct answer: C. The task, context, constraints and desired format

27. Giving the model a role, such as "You are a tax advisor", helps to:

A. Remove all errors
B. Set the tone, perspective and expertise of responses ✔
C. Guarantee accuracy
D. Increase cost

Correct answer: B. Set the tone, perspective and expertise of responses

28. Specifying output format, such as a table or bullet list, helps to:

A. Make answers longer always
B. Hide reasoning
C. Get responses that are easier to use ✔
D. Remove facts

Correct answer: C. Get responses that are easier to use

29. Zero-shot prompting means:

A. Asking the model to perform a task without examples ✔
B. Giving many examples
C. Using images only
D. Fine-tuning the model

Correct answer: A. Asking the model to perform a task without examples

30. Few-shot prompting means:

A. Giving no instructions
B. Using only one word
C. Training a new model
D. Providing a few examples of the desired input and output ✔

Correct answer: D. Providing a few examples of the desired input and output

31. Chain-of-thought style prompting encourages:

A. Skipping logic
B. Step-by-step reasoning for complex problems ✔
C. Random guesses
D. Shorter answers only

Correct answer: B. Step-by-step reasoning for complex problems

32. Breaking a complex task into smaller steps (prompt chaining) helps to:

A. Reduce accuracy
B. Improve quality and control over each stage ✔
C. Confuse the model
D. Increase hallucinations

Correct answer: B. Improve quality and control over each stage

33. Delimiters such as triple quotes or XML-style tags help to:

A. Encrypt data
B. Clearly separate instructions from content ✔
C. Hide the prompt
D. Shorten output

Correct answer: B. Clearly separate instructions from content

34. Iterative prompting means:

A. Never reviewing outputs
B. Using one prompt forever
C. Deleting history
D. Refining prompts based on previous outputs ✔

Correct answer: D. Refining prompts based on previous outputs

35. A system prompt or custom instructions set:

A. Persistent behaviour and preferences for the conversation ✔
B. The GPU speed
C. The model’s weights
D. The internet connection

Correct answer: A. Persistent behaviour and preferences for the conversation

36. Temperature controls:

A. Model size
B. Internet access
C. Randomness and creativity of outputs ✔
D. Response language

Correct answer: C. Randomness and creativity of outputs

37. A low temperature generally produces:

A. More focused and deterministic responses ✔
B. More random responses
C. Images
D. Longer answers always

Correct answer: A. More focused and deterministic responses

38. Tokens are:

A. Passwords
B. GPUs
C. Payments
D. Chunks of text that models process, such as word pieces ✔

Correct answer: D. Chunks of text that models process, such as word pieces

39. The context window is:

A. The training dataset
B. The browser tab
C. The screen size
D. The maximum amount of text the model can consider at once ✔

Correct answer: D. The maximum amount of text the model can consider at once

40. If a conversation exceeds the context window:

A. Earlier content may be truncated or forgotten ✔
B. The model learns permanently
C. Nothing changes
D. The answer becomes perfect

Correct answer: A. Earlier content may be truncated or forgotten

41. A hallucination is:

A. A type of prompt
B. A correct citation
C. A system error code
D. A confident but incorrect or fabricated output ✔

Correct answer: D. A confident but incorrect or fabricated output

42. Hallucinations can be reduced by:

A. Asking for longer answers
B. Providing source material and asking for citations or verification ✔
C. Raising temperature
D. Removing context

Correct answer: B. Providing source material and asking for citations or verification

43. AI outputs should be:

A. Trusted blindly
B. Published without review
C. Reviewed and fact-checked before important use ✔
D. Never edited

Correct answer: C. Reviewed and fact-checked before important use

44. Asking the model to say "I don’t know" when unsure helps to:

A. Remove context
B. Increase hallucinations
C. Reduce fabricated answers ✔
D. Slow responses

Correct answer: C. Reduce fabricated answers

45. Grounding means:

A. Training on random data
B. Turning off the model
C. Using no context
D. Basing responses on provided or retrieved trusted information ✔

Correct answer: D. Basing responses on provided or retrieved trusted information

46. Retrieval-Augmented Generation (RAG) combines:

A. Two GPUs
B. Document retrieval with text generation ✔
C. Images and audio only
D. Spreadsheets and email

Correct answer: B. Document retrieval with text generation

47. Custom GPTs allow users to:

A. Host websites
B. Mine crypto
C. Change OpenAI’s base model weights
D. Create tailored assistants with instructions, knowledge and tools ✔

Correct answer: D. Create tailored assistants with instructions, knowledge and tools

48. ChatGPT can analyse uploaded files using:

A. Only manual reading
B. Only voice
C. Only images
D. Built-in data analysis or code execution tools ✔

Correct answer: D. Built-in data analysis or code execution tools

49. Multimodal models can process:

A. Only text
B. Only code
C. Text, images and sometimes audio ✔
D. Only numbers

Correct answer: C. Text, images and sometimes audio

50. Prompting for images with DALL·E or similar tools should describe:

A. Only file size
B. Only one adjective
C. Subject, style, composition and lighting ✔
D. Only colours

Correct answer: C. Subject, style, composition and lighting

Questions 51 to 75

51. Prompt injection is:

A. A way to speed up models
B. A debugging tool
C. A training method
D. Malicious instructions hidden in content to manipulate an AI system ✔

Correct answer: D. Malicious instructions hidden in content to manipulate an AI system

52. To reduce prompt injection risk, applications should:

A. Share secrets in prompts
B. Disable validation
C. Execute all instructions found in documents
D. Treat external content as data and restrict tool permissions ✔

Correct answer: D. Treat external content as data and restrict tool permissions

53. Sensitive personal or confidential data in prompts should be:

A. Stored in prompts permanently
B. Shared publicly
C. Pasted freely
D. Avoided or handled according to policy ✔

Correct answer: D. Avoided or handled according to policy

54. Bias in AI outputs can come from:

A. Fast internet
B. Patterns in training data ✔
C. Short prompts only
D. Using punctuation

Correct answer: B. Patterns in training data

55. Responsible use of ChatGPT includes:

A. Ignoring policies
B. Passing off unchecked output as expert advice
C. Disclosing AI assistance where required and checking accuracy ✔
D. Plagiarising

Correct answer: C. Disclosing AI assistance where required and checking accuracy

56. Asking for multiple options or variations helps to:

A. Confuse the model
B. Reduce quality
C. Explore ideas and choose the best result ✔
D. Increase cost only

Correct answer: C. Explore ideas and choose the best result

57. Asking the model to critique or improve its own draft can:

A. Delete the draft
B. Remove all bias
C. Guarantee perfection
D. Surface issues and improve quality ✔

Correct answer: D. Surface issues and improve quality

58. Providing an example of the desired tone helps the model to:

A. Ignore instructions
B. Change language randomly
C. Shorten output
D. Match style more closely ✔

Correct answer: D. Match style more closely

59. Negative instructions alone (only saying what not to do) are often:

A. Less effective than stating what to do ✔
B. Always best
C. Required
D. Ignored by design

Correct answer: A. Less effective than stating what to do

60. Word or length limits in prompts help to:

A. Control response length ✔
B. Change the model
C. Remove hallucinations
D. Improve accuracy always

Correct answer: A. Control response length

61. Persona, Task, Context, Format is:

A. A common structure for writing effective prompts ✔
B. A model name
C. A coding language
D. A database schema

Correct answer: A. A common structure for writing effective prompts

62. A prompt template is:

A. A model checkpoint
B. A GPU setting
C. A file format
D. A reusable prompt with placeholders for variables ✔

Correct answer: D. A reusable prompt with placeholders for variables

63. APIs let developers:

A. Edit model weights directly always
B. Integrate language models into their own applications ✔
C. Bypass usage policies
D. Host ChatGPT locally always

Correct answer: B. Integrate language models into their own applications

64. API usage is typically billed by:

A. File size only
B. Number of users only
C. Screen time
D. Tokens processed in input and output ✔

Correct answer: D. Tokens processed in input and output

65. Function calling or tool use allows models to:

A. Request structured calls to external tools or APIs ✔
B. Run on any hardware
C. Access any system freely
D. Modify their training

Correct answer: A. Request structured calls to external tools or APIs

66. Structured output (e.g. JSON mode) helps to:

A. Hide data
B. Increase tokens
C. Produce machine-readable responses ✔
D. Make answers poetic

Correct answer: C. Produce machine-readable responses

67. Fine-tuning differs from prompting because it:

A. Is free always
B. Only changes the prompt
C. Needs no data
D. Adjusts model weights using example data ✔

Correct answer: D. Adjusts model weights using example data

68. Embeddings are:

A. Images
B. Prompts
C. Passwords
D. Numeric vector representations of text meaning ✔

Correct answer: D. Numeric vector representations of text meaning

69. Semantic search uses embeddings to:

A. Encrypt text
B. Sort alphabetically
C. Find content by meaning rather than exact keywords ✔
D. Count words

Correct answer: C. Find content by meaning rather than exact keywords

70. AI agents built on LLMs can:

A. Never use tools
B. Only answer one question
C. Only generate images
D. Plan and take multi-step actions using tools ✔

Correct answer: D. Plan and take multi-step actions using tools

71. Evaluation of prompts should use:

A. No testing
B. One random try
C. Test cases with expected outputs and criteria ✔
D. Gut feeling only

Correct answer: C. Test cases with expected outputs and criteria

72. Prompt versioning helps teams to:

A. Increase costs
B. Delete history
C. Track changes and compare prompt performance ✔
D. Hide prompts

Correct answer: C. Track changes and compare prompt performance

73. ChatGPT’s knowledge may be outdated because:

A. Models are trained on data up to a cutoff date ✔
B. They never learn
C. They read the internet live always
D. They use only books

Correct answer: A. Models are trained on data up to a cutoff date

74. Browsing or search features help by:

A. Removing citations
B. Retrieving current information from the web ✔
C. Changing model weights
D. Slowing answers only

Correct answer: B. Retrieving current information from the web

75. Copyright considerations mean users should:

A. Ignore terms
B. Remove attributions
C. Copy any text freely
D. Avoid reproducing protected works and respect licences ✔

Correct answer: D. Avoid reproducing protected works and respect licences

Questions 76 to 100

76. Using AI for code generation requires:

A. Trusting all output
B. Skipping tests
C. Deploying without review
D. Reviewing, testing and securing generated code ✔

Correct answer: D. Reviewing, testing and securing generated code

77. Summarisation prompts work best when they specify:

A. Only the file name
B. Nothing
C. Audience, length and key points to focus on ✔
D. Only emojis

Correct answer: C. Audience, length and key points to focus on

78. Translation prompts can improve by specifying:

A. Random words
B. Nothing
C. Only the source language
D. Target audience, tone and regional variant ✔

Correct answer: D. Target audience, tone and regional variant

79. Classification prompts benefit from:

A. A defined list of allowed labels ✔
B. Images only
C. No instructions
D. Unlimited labels

Correct answer: A. A defined list of allowed labels

80. Extraction prompts are used to:

A. Pull specific fields from unstructured text ✔
B. Delete data
C. Generate stories
D. Translate code

Correct answer: A. Pull specific fields from unstructured text

81. Brainstorming prompts benefit from:

A. Asking for one idea only
B. Asking for many diverse ideas with constraints ✔
C. Using no context
D. Setting temperature to zero always

Correct answer: B. Asking for many diverse ideas with constraints

82. Socratic prompting means:

A. Forcing short answers
B. Asking the model to guide learning through questions ✔
C. Removing explanations
D. Giving only facts

Correct answer: B. Asking the model to guide learning through questions

83. Prompting in steps for maths problems helps because:

A. It removes tokens
B. Language models can make arithmetic errors without structured reasoning ✔
C. Maths is never needed
D. Models are calculators

Correct answer: B. Language models can make arithmetic errors without structured reasoning

84. Using a calculator or code tool for precise maths is recommended because:

A. LLMs never err
B. Tools are slower always
C. It is required by law
D. LLMs can make numerical mistakes ✔

Correct answer: D. LLMs can make numerical mistakes

85. Conversation memory features let ChatGPT:

A. Remember preferences across chats when enabled ✔
B. Read your email always
C. Change its weights
D. Access private files without permission

Correct answer: A. Remember preferences across chats when enabled

86. Temporary or incognito chats are useful when:

A. You do not want the conversation saved to history or memory ✔
B. You need images
C. You need longer answers
D. You want permanent records

Correct answer: A. You do not want the conversation saved to history or memory

87. Model selection matters because:

A. All models are identical
B. Models never change
C. Bigger is always cheaper
D. Different models trade off speed, cost and reasoning ability ✔

Correct answer: D. Different models trade off speed, cost and reasoning ability

88. Reasoning-focused models are best for:

A. Simple greetings only
B. File storage
C. Complex multi-step problems ✔
D. Image compression

Correct answer: C. Complex multi-step problems

89. Rate limits restrict:

A. How many requests can be made in a period ✔
B. Answer quality
C. Language
D. Formatting

Correct answer: A. How many requests can be made in a period

90. Enterprise AI governance includes:

A. No rules
B. Ignoring compliance
C. Policies on data, access, review and acceptable use ✔
D. Sharing all data

Correct answer: C. Policies on data, access, review and acceptable use

91. Prompt length should be:

A. Random
B. As long as needed to give clear context, but no longer ✔
C. Always maximum
D. Always one line

Correct answer: B. As long as needed to give clear context, but no longer

92. Asking the model to cite sources from provided documents improves:

A. Creativity only
B. Token cost reduction
C. Speed only
D. Traceability and trust ✔

Correct answer: D. Traceability and trust

93. Reframing a failed prompt by adding context or examples often:

A. Makes no difference
B. Deletes history
C. Improves results ✔
D. Breaks the model

Correct answer: C. Improves results

94. A good prompt engineer focuses on:

A. Secret magic words
B. Longest possible prompts
C. Avoiding evaluation
D. Clarity, testing, iteration and responsible use ✔

Correct answer: D. Clarity, testing, iteration and responsible use

95. The main goal of prompt engineering is to:

A. Bypass safety
B. Reliably get useful, accurate and safe outputs from AI ✔
C. Use more tokens
D. Trick the AI

Correct answer: B. Reliably get useful, accurate and safe outputs from AI

96. Top-p (nucleus sampling) controls:

A. The range of likely tokens considered when generating ✔
B. Context window
C. Language
D. Model size

Correct answer: A. The range of likely tokens considered when generating

97. Stop sequences tell the model to:

A. Start over
B. Summarise
C. End generation when a specific text appears ✔
D. Translate

Correct answer: C. End generation when a specific text appears

98. Meta-prompting means:

A. Asking the model to help write or improve a prompt ✔
B. Training the model
C. Using images
D. Deleting prompts

Correct answer: A. Asking the model to help write or improve a prompt

99. Specifying the audience, such as "explain to a 10-year-old", helps:

A. Increase errors
B. Remove context
C. Adjust complexity and vocabulary ✔
D. Change model

Correct answer: C. Adjust complexity and vocabulary

100. Max tokens setting limits:

A. The prompt quality
B. The model version
C. The length of the generated response ✔
D. The temperature

Correct answer: C. The length of the generated response

How to Pass the ChatGPT and Prompt Engineering Exam

  • Read every question carefully; most options look similar, so look for the most precise answer.
  • Revise the questions above topic by topic rather than memorising letters, because option order can change in the live test.
  • You need 80 correct answers out of 100 to pass, so aim to understand at least 85 to stay safe.
  • If you do not clear it on the first attempt, review the questions you missed and use the retake option.

Frequently Asked Questions

How many questions are in the ChatGPT and Prompt Engineering Professional Certification exam?

The exam has 100 multiple-choice questions, each worth 1 mark.

What is the passing score for the ChatGPT and Prompt Engineering Professional Certification?

You need at least 80% (80 out of 100) to pass and receive the certificate.

Can I retake the ChatGPT and Prompt Engineering Professional Certification exam?

Yes. If you score below 80%, you can retry the test after reviewing the topics.

Is the Itronix certificate verifiable?

Yes. Every certificate issued by Itronix Solutions has a unique ID that can be checked on the Itronix verify-certificate page.

More AI & Machine Learning Certification Answers

Disclaimer: These answers are provided for learning and revision. Questions and answers are checked regularly; if you find a mistake, let us know in the comments.

About Priya Dogra

Priya Dogra runs priyadogra.com, where she shares free certification courses, internships, jobs, scholarships and exam preparation guides for Indian students and freshers. Follow her on YouTube, LinkedIn and Instagram for the latest opportunities.

Check Also

Unix/ Linux Certification exam answers 2026

Unix/ Linux Certification Exam Answers 2026 (All 100 Questions)

All 100 Unix/ Linux Certification exam questions with correct answers highlighted - updated 2026.

Leave a Reply

Your email address will not be published. Required fields are marked *