GAF Study Guide 2026: How to Pass on Your First Attempt

What You Are Actually Studying For

Generative AI Foundations is Pearson's entry-level credential for people who need to understand and use generative AI tools responsibly. It is developed and administered through Certiport, a Pearson VUE business, as part of the Critical Career Skills (CCS) family. If you have seen other certifications that share the same three-letter acronym, set them aside. Nothing about their sponsors, fees, or content applies here. This guide covers only Pearson's Generative AI Foundations exam.

The exam is brand-agnostic. You are not being tested on one vendor's chatbot interface or on a particular product's menus. You are being tested on concepts that transfer across tools: how generative models produce output, how to write and improve prompts, and how to think about the ethical, legal, and social consequences of using them. That shapes how you should prepare. Memorizing screens will not help. Being able to reason about a scenario will.

If you are still orienting yourself, the overview of what the certification is and the eligibility and prerequisites page cover the background. This article focuses on what to do between deciding to sit the exam and walking out with a passing score.

Exam Mechanics You Must Know Before Booking

Knowing the format removes a surprising amount of anxiety. Here is what the issuer documentation establishes.

The question-count discrepancy

Two official-adjacent sources disagree on how many questions you will see. Certiport's scoring policy explains that totals differ by exam form, so the sensible response is to prepare for the higher end of the range and not to build a pacing plan around a single number. With 50 minutes on the clock, you have roughly a minute per question at the high end. That is enough to read carefully and not enough to agonize, which is exactly why the flag-and-review feature matters.

What a 700 really means

The passing score is a scaled score, not a percentage of questions answered correctly. A 700 does not translate to "70 percent," and you should not try to reverse-engineer how many questions you can miss. The practical consequence is simple: aim for solid command of every domain instead of a minimum-viable margin. The passing score breakdown explains scaled scoring in more depth.

Voucher and Delivery Channel: Avoid the Costly Mismatch

The most avoidable first-attempt failure has nothing to do with knowledge: buying a voucher that cannot be redeemed where you plan to test. The issuer's terms draw clear lines.

• Certiport single-exam voucher: USD 72 in U.S. pricing, or USD 83 for a bundle that includes a retake. A Certiport Authorized Testing Center (CATC) may add its own proctoring charge on top.
• Where it works: Certiport vouchers are valid only for CATC delivery, either in person or remotely. They cannot be redeemed through OnVUE or at a Pearson VUE Testing Center.
• Pearson Mindhub voucher: USD 85, with Mindhub Pro listing both web and member prices at USD 85. These have country restrictions and must match the country you select at checkout.

Pearson's October 2024 launch announcement mentions an OnVUE delivery route. However, the public Mindhub scheduling link currently redirects to an unrelated Adobe scheduling page, so a current, GAF-specific OnVUE booking process could not be verified. Do not assume it works. Before paying, contact your testing channel and confirm in writing that the specific voucher type you intend to buy can be used with the specific delivery method you intend to use.

For the full fee picture, including retakes, see the certification cost breakdown, and for scheduling logistics see the exam dates and scheduling guide.

Domain 1: Generative AI Methods and Methodologies

This is the conceptual foundation. The official outline publishes no percentage weights, so you cannot assume this domain is the largest or smallest, but its vocabulary underpins everything that follows. If you do not understand what a generative model does, prompt questions will feel like guesswork.

Expect scenario-style questions that ask which approach suits a described need, or which statement about generative AI is accurate. Because the exam is brand-agnostic, answer from principle, not from memory of a specific product's behavior. If two options both sound plausible, prefer the one that acknowledges limitations and human oversight.

For a deeper tour of each area, the domains guide walks through all four content areas.

Domain 2: Basic Prompt Engineering

This domain tests whether you can turn a goal into an effective instruction. It is the most hands-on part of the outline, and it rewards candidates who have actually typed prompts into a generative tool and watched what came back. The issuer's expectation of about 150 combined hours of instruction and hands-on use makes sense here: you cannot learn prompting purely by reading.

A reliable habit for practice: take a one-line request, then rewrite it three ways, adding context, then format, then constraints. Compare the results. You will internalize what the exam is probing far faster than by rereading definitions. Exam questions in this area often present a weak prompt and ask which revision would improve it most, so train your eye on diagnosing what is missing.

Domain 3: Prompt Refinement

Where Domain 2 is about writing the first prompt, Domain 3 is about the iteration that follows. Real work with generative AI is rarely one-shot. You read the output, judge it, and adjust. This domain tests whether you can diagnose a disappointing response and choose the right correction.

Domain 4: Ethics, Law, and Societal Impact

Candidates from technical backgrounds sometimes underestimate this domain, and candidates from non-technical backgrounds sometimes find it the most intuitive. Either way, it carries real exam presence and deserves dedicated study time because the right answer often hinges on careful reading of a described situation.

When a question describes a risky scenario, the best answer usually combines verification, transparency, and protection of sensitive information. Be cautious with options that treat AI output as automatically authoritative or that suggest sharing confidential data with an unvetted tool. Do not rely on memorizing specific statutes; the exam outline frames this domain around understanding principles and consequences, so focus on reasoning about responsible use.

A Four-Week Plan Mapped to the Four Domains

This is the one section on pacing, and it is tied to the outline rather than generic advice. The order matters: concepts first, then the two prompt domains back to back while the skills are fresh, then ethics, then integration. Adjust the length to your own baseline. If you already use generative tools daily, you may compress Weeks 2 and 3.

For a compact refresher in the final days, the one-page cheat sheet consolidates the must-know facts, and the training overview covers structured learning options.

Readiness Check and Exam-Day Rules

Judging whether you are ready

Practice products exist for this exam, including Certiport's Critical Career Skills practice test and a MeasureUp Generative AI Foundations practice test. Use them for diagnosis, not just score-chasing. After each timed set, sort every missed question into one of the four domains and look for a pattern. A scattered set of misses suggests careless reading; a cluster in one domain tells you where to spend your last study days. You can also try the free GAF practice test to check your timing against the 50-minute limit, and the practice question bank helps you drill domain by domain.

Be honest about difficulty without catastrophizing. The exam is a foundations-level assessment, but it is proctored and timed, and scenario wording can be subtle. The difficulty guide and the pass rate discussion give a measured view; note that no verified pass-rate figure is published, so be skeptical of any site quoting a precise one.

Testing-center rules to respect

• Bring valid identification; CATC rules require it.
• Candidates under 18 need parental consent.
• Books and personal devices must remain inaccessible during the exam.
• Arrive early or log in early, since the tutorial and administrative time sit outside the 50 timed minutes but still occupy your seat time.

After You Pass: Validity and Career Use

A certification earned on or after June 18, 2025 is valid for five years. Recertification requires purchasing a new voucher and passing the then-current exam, so plan for that if you want the credential to stay active. Earlier nonexpiring certifications keep their existing status.

As for what the credential does for you, treat it as evidence of baseline competence and responsible-use awareness, which is useful for students, early-career professionals, and people in roles where generative AI is entering daily workflows. It is a foundations credential, not a specialist engineering qualification, so set expectations accordingly. The ROI analysis, the earnings overview, and the jobs page explore how it fits into a career path without overstating what a single entry-level credential can do.

Frequently Asked Questions