The SnowPro Specialty: Gen AI exam, version GES-C02, gives you 55 questions in 85 minutes. Two of its four domains carry 67% of the score: Gen AI Functions at 38% and Gen AI Governance at 29%. The second number is the one people underrate. Many candidates prepare as if this were a prompt-writing test, and then meet a block of questions on roles, privileges and data protection.
This note turns Snowflake's published weights into an estimate of where the questions come from, and then into a study plan you can follow. Every exam figure below comes from Snowflake's GES-C02 certification page, which I read on 27 October 2026. Check it again before you book, because fees and outlines change.
GES-C02 at a glance
- Exam version: GES-C02
- Questions: 55, multiple choice and multiple select
- Time limit: 85 minutes, which is about 93 seconds per question
- Passing score: 750 on a scaled score from 0 to 1000
- Fee: $225 USD, or $180 USD for candidates registering in India
- Language: English
- Prerequisites: none
- Delivery: online proctoring or an onsite testing center
Snowflake also notes that the exam may include unscored items. They are not marked on the form, they do not affect your score, and extra time is built in for them. So treat every question as if it counts.
Who the exam is written for
There is no formal prerequisite, but the candidate profile is specific: one or more years of Gen AI experience with Snowflake in an enterprise environment. Snowflake adds that successful candidates may be able to write Python, and that data engineering and SQL knowledge is assumed.
Read that last line carefully. If you have not written Snowflake SQL before, start there. Almost every Gen AI feature on this exam is called from SQL, so weak SQL turns easy questions into hard ones.
Where the 55 questions are likely to come from
Snowflake publishes domain weights, not question counts. Multiplying each weight by 55 gives a fair estimate. Treat it as a guide only, because the unscored items are part of the 55.
| Domain | Weight | Estimated questions |
|---|---|---|
| 1.0 Snowflake for Gen AI Overview | 18% | about 10 |
| 2.0 Snowflake Gen AI Functions | 38% | about 21 |
| 3.0 Snowflake Gen AI Governance | 29% | about 16 |
| 4.0 Snowflake Document Processing | 15% | about 8 |
Domain 1: Gen AI overview (18%)
This is the "which tool for which job" domain. Snowflake's overview says the exam tests whether you can define and apply its Gen AI principles, capabilities and best practices. It also expects you to build and fine-tune open-source models with Snowpark Container Services and the Snowflake Model Registry. Be clear on when a managed Cortex function is enough and when a custom model is worth the extra work.
Domain 2: Gen AI functions (38%)
The largest domain is about using Cortex AI features and functions, including LLMs, for real use cases. Snowflake's documentation lists functions such as these:
- AI_COMPLETE generates a completion from text or image input with an LLM you choose.
- AI_CLASSIFY sorts text or images into categories you define.
- AI_FILTER returns true or false for each row, so it fits in a filtering clause.
- AI_AGG returns insights across many rows of text.
- AI_EMBED and AI_SIMILARITY create embeddings and compare them.
- AI_SUMMARIZE, AI_SENTIMENT and AI_TRANSLATE cover the common text tasks.
Reading about these is not enough. Load a small table of support tickets or reviews and run each function on it. Note what goes in, what comes out, and which functions work row by row and which work across rows. That difference is easy to test with one question.
Domain 3: Gen AI governance (29%)
Almost a third of the exam is about control: who may call the functions, and how data is protected. A good starting point is the access model in Snowflake's documentation. A role needs the USE AI FUNCTIONS account-level privilege plus one of two database roles, CORTEX_USER or AI_FUNCTIONS_USER.
Practise it. Create a test role, call a function before and after the grant, and read the error you get. Then look at AI_REDACT, which removes personally identifiable information from text, and at AI_COUNT_TOKENS, which tells you how large a job is before you run it. The study guide lists the full objectives for this domain, so check your notes against it.
Domain 4: Document processing (15%)
The smallest domain is about building, managing and optimizing document parsing pipelines. Three pieces matter most. TO_FILE references a file in a stage, AI_PARSE_DOCUMENT extracts text or layout from staged documents, and AI_EXTRACT pulls specific fields out of them. Build one small pipeline end to end: stage a few PDFs, parse them, extract two or three fields, and store the results in a table.
A 40-hour plan that follows the weights
If you have about 40 hours, give each domain the share of time it has on the exam:

- Gen AI Overview: 7 hours
- Gen AI Functions: 15 hours
- Gen AI Governance: 12 hours
- Document Processing: 6 hours
Spread over four weeks of about 10 hours each, it looks like this:
- Week 1: Read the exam study guide and the overview objectives. Set up a Snowflake trial or sandbox account and run AI_COMPLETE and AI_CLASSIFY on a sample table.
- Week 2: Work through the rest of the functions, including embeddings, similarity, filtering and aggregation. Keep a one-line note for each one.
- Week 3: Governance first: privileges, database roles, redaction and token counts. Then build the document pipeline.
- Week 4: Take Snowflake's official SnowPro Practice Exam: Gen AI. Sort your misses by domain, review the weakest one, and finish with timed sets at about 93 seconds per question.
Timing on exam day
At 85 minutes for 55 questions you have about a minute and a half per item. Multiple select questions take longer, because every option needs a decision. If one stalls you, make your best choice and move on. Spending three minutes on a single item costs you two later ones.
Keeping the credential current
Snowflake lists continuing education options for renewal. You can pass another eligible SnowPro exam from the Specialty or Advanced series, or complete an eligible instructor-led course. The Snowflake Gen AI course is on that list. Plan this early if your employer pays for training.
Frequently asked questions
1. How many questions are on the GES-C02 exam?
There are 55 questions, multiple choice and multiple select, with a time limit of 85 minutes. Some items may be unscored, but you cannot tell which ones.
2. What is the passing score for SnowPro Specialty: Gen AI?
You need 750 on a scaled score from 0 to 1000.
3. How much does the GES-C02 exam cost?
The registration fee is $225 USD. Candidates registering in India pay $180 USD.
4. Do I need SnowPro Core before GES-C02?
No. Snowflake lists no prerequisites. It does assume data engineering and SQL knowledge, and it recommends one or more years of Gen AI experience with Snowflake.
5. Which domain should I study most?
Gen AI Functions (38%), then Gen AI Governance (29%). Together they make up about two-thirds of the exam.
Sources: Snowflake's GES-C02 certification page on learn.snowflake.com and Snowflake's documentation on Cortex AI functions, both read on 27 October 2026.

Responses (0)