Which free AI literacy test should you take in 2026?
Four free tests dominate the search results in September 2026: JobCannon's 8-question AI Literacy Test, AISA's 20-minute conversational AI Skills Assessment, NumberKnow's 20-question quiz with certificate, and UD's 15-question AI IQ Test for Hong Kong industries. They measure different things. Pick by what you need the score for, not by which one ranks first.
I took all four in one afternoon to answer a question I kept hearing from marketers and ops managers in Hong Kong: "I use ChatGPT every day, but how do I know if I am actually any good at this?" The answer is that a free test can tell you, but only if you pick the one that measures the skill you care about.
Definition: An AI literacy test is a short assessment that measures how well a person understands and applies AI tools at work: prompting, spotting hallucinations, knowing tool limits, handling data safely, and choosing where AI fits a workflow. It does not test coding or machine-learning theory.
This page is a buying-decision page. The "purchase" is your time and, for one of the four, an optional paid upgrade. Below you will find what each test actually measures, how long it takes, what is free and what is not, where each one falls short, and a verdict by user type. All facts were checked against the four product pages on 9 September 2026 and may change.
What are the four free AI literacy tests and how do they compare?
JobCannon is the fastest (8 multiple-choice questions, about 2 to 3 minutes, level shown free, detailed report paid). AISA is the deepest (a 20 to 25 minute AI-led conversation scored on 11 criteria, free report and certificate). NumberKnow is the best free certificate quiz (20 questions, score out of 100, no sign-up). UD AI IQ Test is the only one built for Hong Kong industries (15 true-or-false questions in Cantonese, 8 sectors, free).
Extractable comparison (verified 9 September 2026):
--- JobCannon AI Literacy Test: 8 questions, 8 competency areas, roughly 2 to 3 minutes. Free: your readiness level (Beginner to Expert) on screen. Paid: the deep-dive report and gap recommendations are part of JobCannon Premium (price not shown on the test page). Language: English. Certificate: no.
--- AISA AI Skills Assessment: conversational, typically 20 to 25 minutes (90-minute maximum). Scored across 5 dimensions and 11 criteria, normalised to 0 to 100, plus one of 10 "AI personas". Free: full report, persona, LinkedIn-verifiable certificate; no card required. Paid: employer team packages and an optional AI Coach subscription. Language: English.
--- NumberKnow AI Literacy Test: 20 multiple-choice questions, 5 points each, score out of 100 mapped to five levels (AI Curious to AI Expert). Free: score, level, certificate with your name, instant explanations. No sign-up, no email; answers stay in your browser. Language: English.
--- UD AI IQ Test: 15 true-or-false questions, chosen by industry (catering, retail, real estate, education, medical, logistics, insurance, HR). Four result tiers by accuracy: Digital Dinosaur 0 to 25%, Wait-and-See 26 to 50%, Catching Up 51 to 75%, AI Pioneer 76 to 100%. Free, no payment step. Language: Cantonese (Traditional Chinese). Certificate: no. Follow-up: free consultation with UD's AI team.
The four tests sit on a spectrum from "vocabulary check" to "skills interview". JobCannon and NumberKnow test what you know about AI. AISA tests what you can do with it. UD's test asks whether you can see where AI fits a specific Hong Kong business, which is a different question again.
Why do people switch from a quick quiz to a conversational assessment?
People move from quizzes to AISA-style assessments when they need proof rather than a number. A multiple-choice quiz can be passed by reading a blog post the night before; a 20-minute conversation with exercises (diagnose a broken output, iterate on a prompt) cannot. The cost of switching is time: 20 minutes instead of 3, and a willingness to talk or type at length.
AISA is explicit about this trade-off on its own site: quizzes measure knowledge recall and are "easily" gamed, while conversational assessment measures applied proficiency. That is a fair description of the format difference, and it matched my experience. AISA asked me to fix a prompt that was producing generic output. JobCannon asked me to define a term.
The reverse switch also happens. If you just want a quick benchmark before a training session, or you want to screen ten team members in a lunch break, a 3-minute quiz wins. Nobody runs a 25-minute interview for a warm-up.
The limitation most likely to stop someone choosing AISA is language and cultural fit. The conversation is in English, the exercises assume a Western professional context, and role personas (product manager, developer) map imperfectly onto a Hong Kong SME owner who does everything. For that person, a Cantonese test built around their own industry removes the friction, even if it is shallower.
What does the UD AI IQ Test measure that the others do not?
UD's AI IQ Test measures industry-specific AI awareness for Hong Kong businesses. You pick one of eight sectors and answer 15 true-or-false statements about how AI is used in that sector, in Cantonese. The result is one of four tiers plus a free follow-up consultation. It does not measure prompting skill or issue a certificate.
The design choice is deliberate. The three English tests treat AI literacy as an individual professional skill. UD's test treats it as a business-decision skill: does a restaurant owner know what AI can do for reservations and food-safety training, does an insurance agent know where AI underwriting help stops. That framing is why the results page leads to a consultation rather than a certificate.
Two honest observations from taking it. First, 15 true-or-false items is a coarse instrument; two lucky guesses move you a whole tier. Second, several statements test awareness of what AI can do rather than skill at doing it, so an experienced power user can score "AI Pioneer" in three minutes and learn little. It is a starting-point test, and it is honest about being one.
Where it is strong: it is the only one of the four written for Hong Kong, the only one in Cantonese, and the only one whose eight industry tracks map to actual local SME categories. If your goal is to get a boss or a team who have never touched AI to engage with the topic for three minutes, the industry framing works better than an abstract quiz. You can start it directly at ai-staff.ud.hk/iq-test.
What are the honest limitations of each AI literacy test?
None of the four is a validated psychometric instrument, and none will impress a hiring manager on its own. JobCannon's useful report is behind a paywall. AISA is long and English-only. NumberKnow is a knowledge quiz, not a skills test, and its certificate is self-issued. UD's test is 15 true-or-false items in Cantonese with no certificate and a sales follow-up.
Specific caveats worth knowing before you spend the time:
--- JobCannon shows your level free but keeps the gap analysis and learning path in Premium. If the whole point is finding out what to learn next, budget for the upgrade or use a different test.
--- AISA states that its rubric covers 93% of the behaviours in Anthropic's 2026 AI Fluency Index. That is AISA's own mapping, not an Anthropic endorsement. The certificate is verifiable but is issued by AISA, a startup, not by an accreditation body.
--- NumberKnow is a content site whose test sits alongside a mortgage calculator and a love-language quiz. The questions are sound and the explanations are useful, but the certificate carries the weight of a well-made quiz, no more.
--- UD AI IQ Test is a lead-generation tool for UD's AI Staff Solution, and it says so: the results page offers a free consultation. The test content is free and useful on its own, but you should know the business model, exactly as you should for JobCannon Premium and AISA's team packages.
The common limitation is that every one of these measures a snapshot. AI tools change monthly; JobCannon itself recommends retaking every 3 to 6 months. Treat any score as a photo, not a diploma.
Which AI literacy test is right for you? A verdict by user type
Choose JobCannon for a 3-minute benchmark before training. Choose AISA if you want a shareable, evidence-based credential and can spare 25 minutes in English. Choose NumberKnow for a free certificate with zero sign-up. Choose UD's AI IQ Test if you run or work in a Hong Kong SME and want a Cantonese, industry-specific starting point that leads to a real conversation.
--- Marketer or content creator building a LinkedIn profile: AISA. The 11-criteria report and verifiable certificate are the only outputs here an employer might actually look at.
--- Ops or project manager screening a team quickly: JobCannon for speed, or NumberKnow if you want everyone to walk away with something. Run AISA only with the two or three people you plan to make AI champions.
--- Educator or trainer who needs teaching material: NumberKnow. Every answer explains itself, and the question set doubles as a lesson plan on hallucinations, bias and privacy.
--- Hong Kong SME owner or the person who reports to one: UD AI IQ Test first, then AISA if you want to go deeper personally. The industry framing gets a sceptical boss to engage; the consultation turns the score into a plan.
--- Power user who already reads model release notes: Only AISA will tell you anything you do not know. The other three will confirm you are at the top tier in under five minutes.
If you are also weighing an organisation-level assessment rather than a personal one, the options and their costs are compared in AI Readiness Assessment Options: What Free, Fixed-Fee and Big Four Cost.
What should you do after you get your AI literacy score?
Turn the score into one concrete change. A low tier means one week of daily practice on a single tool. A middle tier means building one repeatable workflow (a Project, a template, an automation). A top tier means it is time to stop testing yourself and start raising the people around you. Retake in three months to see if anything moved.
The test that most people in Hong Kong should take first is the one they will actually finish and act on. For an individual professional that is usually AISA; for a small business that is usually UD's industry test followed by a conversation about what to automate first.
Scores are cold. What you do the week after is not. We know AI's cold edges. We know your real challenges. 28 years with UD, turning technology into a partnership with warmth.
Reviewed by the UD AI team. Test formats, question counts, pricing tiers and free/paid boundaries verified against the JobCannon, AISA, NumberKnow and UD AI IQ Test pages on 9 September 2026. UD AI IQ Test is a UD product; the comparison criteria are applied identically to all four.
Take the AI IQ Test, then turn your tier into a plan
Fifteen questions, your industry, three minutes, free. Then, if you want to move from "Catching Up" to "AI Pioneer" for real, we'll walk you through every step, from your first AI workflow to a team that runs on it.
