What are you actually deciding when you search for an AI skills test?
An AI skills test is any assessment that scores how well you understand or apply AI tools. The category splits into three products that look similar and solve different problems: free awareness quizzes, paid certifications you can put on a CV, and per-candidate hiring assessments. Buying the wrong one is the usual mistake.
The three buy paths cost wildly different amounts. Free quizzes cost nothing and prove nothing. Paid certifications run from US$100 to US$199 for a single exam. Hiring assessment platforms start around US$35 a month and rise past US$400.
Before comparing anything, answer one question. Do you need a number for yourself, a credential for someone else, or a filter for candidates? Those are three different purchases.
This page compares what each option actually costs in 2026, what each one measures, and where our own free test stops being useful.
What does each AI skills test cost in 2026?
Free self-assessments including Preface's AI-Q Test and UD's AI IQ Test cost nothing. Individual certifications cost US$100 for Uxcel's Certified AI Practitioner or AWS Certified AI Practitioner, US$149 for the Anthropic Claude Developer Certification, and US$199 for the OpenAI GPT API Developer Certificate. Hiring platforms are billed monthly or per candidate.
Free self-assessments
--- Preface AI-Q Test: free, Hong Kong. Covers prompt writing, AI scenario comprehension and practical application. Returns an AI-Q score plus your percentile among other Hong Kong users. Launched as part of Preface's "Just Start" initiative with over 30 partner organisations including Google, Microsoft and Meta, per South China Morning Post.
--- UD AI IQ Test: free, no account required. 15 true-or-false questions, 8 industry tracks, 4 result bands. Available at ai-staff.ud.hk/iq-test.
Paid individual certifications
--- Uxcel Certified AI Practitioner: US$100, or US$70 with the 30% Uxcel Pro discount. 65 questions in 65 minutes. Positioned for product, design and analysis roles rather than engineering.
--- AWS Certified AI Practitioner (AIF-C01): US$100 at foundational level, with regional pricing variation. Passing any AWS exam earns a 50% voucher valid 12 months, which can bring a follow-up exam to US$50.
--- Anthropic Claude Developer Certification: US$149.
--- OpenAI GPT API Developer Certificate: US$199.
Hiring and team assessment platforms
--- Software subscriptions: roughly US$35 to US$400 or more per month depending on seats and reporting depth.
--- Per-candidate pricing: around US$1 to US$5 for basic hard-skills testing, and US$10 to US$50 or more for comprehensive assessments with personality, culture-fit and AI-generated insight reports.
Two notes on these figures. Certification prices are list prices and vouchers move them meaningfully, so check for a discount before paying full price. Free tests carry a different cost, which is that most of them exist to start a conversation with you.
What does each test actually measure?
Free quizzes measure awareness, meaning whether you hold correct beliefs about what AI can and cannot do. Paid certifications measure recall and applied judgement against a defined syllabus, under timed exam conditions. Hiring assessments measure task performance, asking candidates to produce work rather than answer questions about work.
That distinction explains why a free quiz cannot substitute for a certification. A 15-question true-or-false quiz can tell you that you still believe something outdated. It cannot tell you whether you can build a working prompt chain under time pressure.
It also explains why a certification cannot substitute for a hiring assessment. Uxcel's 65-question format tests whether you know the concepts. It does not show you what a candidate produces when handed a messy brief and a model.
The one measurement none of them provide is the most valuable: whether your daily workflow has improved. No test scores your actual output, and that gap is worth remembering before you pay anything.
Where does UD's AI IQ Test fit, and where does it stop?
UD's AI IQ Test is a free 15-question true-or-false quiz across 8 industry tracks: catering, retail, real estate, education, medical, logistics, insurance and HR. It returns one of four bands, from Digital Dinosaur at 0 to 25 percent correct up to AI Pioneer at 76 to 100 percent. It needs no signup and takes a few minutes.
It is genuinely useful for one job. If you are a Hong Kong manager or owner who is unsure whether AI applies to your specific operation, the industry-specific framing puts the question in your own vocabulary rather than in generic tech language.
Where it stops, stated plainly
--- It measures awareness, not proficiency. Fifteen true-or-false questions cannot establish that you can do anything.
--- There is no certificate. Nothing comes out of it that you can attach to a CV, a proposal or a procurement submission.
--- There is no percentile. Unlike Preface's AI-Q Test, it does not tell you how you rank against other Hong Kong users.
--- The bands are coarse. Four bands across 15 questions means two wrong answers can move you a whole band, so treat the label loosely.
--- The content and interface are in Cantonese and framed for business owners. An English-first marketer at a multinational will find the framing off.
--- There is no team view. You cannot assess 12 staff and get a comparison report, which is exactly what a training budget decision needs.
If any of those six lines describes what you need, this is not the right tool and a paid product is the honest answer. The companion AI Ready Check is closer if your question is about the business rather than yourself. It runs 10 questions across 8 industries in about 3 minutes and returns a readiness report on which roles to automate first.
Which AI skills test should you choose?
Choose by what the result has to do for you. If it has to convince someone else, buy a certification. If it only has to inform you, take a free test. If it has to compare people, neither free tests nor individual certifications will work and you need an assessment platform.
Verdict by situation
--- Freelancer or consultant who needs a credential line on a proposal: Uxcel Certified AI Practitioner at US$100, or US$70 with the Pro discount, if your work is marketing, product or analysis. AWS Certified AI Practitioner at US$100 if your clients are cloud-adjacent.
--- Developer or technical practitioner: Anthropic Claude Developer Certification at US$149 or the OpenAI GPT API Developer Certificate at US$199, chosen by which platform you actually ship on.
--- Marketer, ops manager or educator who wants to know where they stand: Preface AI-Q Test, because the percentile against other Hong Kong users is the only benchmark on this list that tells you something relative.
--- Hong Kong owner or manager asking whether AI fits their operation: UD AI IQ Test, then AI Ready Check for the business-level version of the same question.
--- Team lead sizing a training budget for 10 or more people: none of the free tests. Budget for an assessment platform at US$35 to US$400 a month, or run a structured internal exercise on real work.
--- Anyone who mainly wants their outputs to get better: skip all of them this month and fix one workflow instead.
How do you assess your own AI skill without buying anything?
Run a structured self-audit with a model you already pay for. A test scores your beliefs, but an audit examines your actual workflow, which is the thing that determines your output. The prompt below interviews you about how you really work and names the gap that is costing you the most time.
Try this prompt:
Act as an AI workflow auditor. Interview me to assess how effectively I use AI at work. Ask exactly one question at a time and wait for my answer before continuing.
Cover these six areas in order: the tasks I use AI for weekly, how I structure prompts, whether I reuse prompts or rewrite them each time, how I verify outputs, what I have tried and abandoned, and what I still do manually that feels repetitive.
Rules:
--- Ask follow-up questions when an answer is vague. Do not accept "I use it for writing" without asking what kind and what happens next.
--- Do not teach or suggest anything until the interview is finished.
--- After the interview, give me: one paragraph describing my current level in plain language, the three highest-value gaps ranked by time saved per week, and one specific change to make this week.
--- Be direct about weaknesses. Do not open with praise.
Begin with question one.
This takes roughly 15 minutes and produces something no scored test does, which is a ranked list tied to your own tasks. Do it before you spend US$100 on a certification, because it often reveals that the gap is workflow design rather than knowledge.
One caveat. The model will be agreeable unless you tell it not to be, which is why the prompt forbids opening praise. If the output still reads like flattery, paste it back and ask for the harshest defensible reading.
The takeaway
Free tests are worth the few minutes they take, and they are worth exactly that. They will not certify you, rank you against your peers in most cases, or tell you whether your work has improved.
Paid certifications are worth US$100 to US$199 when a named credential unlocks something specific: a client, a role, a procurement box. When nothing specific depends on it, that money buys you a PDF.
The honest sequence is to audit your workflow first, take a free test to check your assumptions, and pay for a certification only when someone else needs the proof. If you want the fastest free starting point, our test is 15 questions and takes a few minutes. If you want a workflow that measurably runs better, that is a different conversation and we are happy to have it. We understand AI. We understand you better. With UD by your side, AI doesn't feel cold.
Reviewed by the UD AI team.
Start with the free test, then build the workflow
Fifteen questions, no signup, eight industry tracks. It tells you where your assumptions are out of date.
Then, if you want the workflow behind the score, we'll walk you through every step, from tool selection to deployment.
