Searching for an intelligence assessment today returns hundreds of browser-based options, most of them advertised as free and most promising a result within minutes. The format has almost entirely displaced the paper booklets of a generation ago. What has not changed is the underlying question people bring to them: is this number telling me anything real, and if so, what?
The honest answer sits somewhere between the dismissive and the credulous. A carefully constructed web assessment can produce a usable estimate of reasoning ability. A badly built one produces a flattering number designed to make you share a link. Telling them apart is not difficult once you know what to look at.
The shift online was driven by economics as much as convenience. A traditionally administered battery requires a trained examiner, a private room, a physical kit of blocks and cards, and roughly ninety minutes of one-to-one time. That is expensive to deliver and impossible to scale.
A web version removes almost all of those costs. Items are delivered by the browser, responses are recorded automatically, timing is handled by the clock on the page, and scoring happens instantly against a stored reference distribution. One well-built assessment can serve millions of people without a single examiner being involved.
The trade-off is control. Everything the examiner used to guarantee — that the participant was rested, alone, undistracted and unaided — becomes an assumption. That single change explains most of the difference in quality between clinical and online results.
Despite enormous variation in presentation, the item pool is remarkably consistent across the better free tests. Three families do most of the work.
The dominant format is the progressive matrix: a grid of shapes with one cell missing, and a set of candidate answers below. The participant identifies the rule governing the grid and selects the option that satisfies it. Rules combine rotation, reflection, addition, subtraction, shading changes and positional shifts, layering as the test progresses.
Matrices are popular online for good reason. They require no reading, translate across languages without adaptation, and correlate strongly with broader reasoning performance. Anyone unfamiliar with the format should look at how abstract reasoning items are constructed before sitting a scored attempt, because the first few minutes are otherwise spent learning the conventions rather than solving.
Sequence items present a run of figures and ask for the next term. The arithmetic involved is deliberately simple; the difficulty lies in identifying the generating rule, which may involve alternating operations, nested patterns or differences between differences. These items lean on everyday numerical reasoning rather than mathematical training, which is why people with modest formal maths often perform well on them.
Verbal items — analogies, odd-one-out, relationship statements — appear less frequently online because they are language-dependent and complicate international norming. When they do appear, they add useful breadth, since verbal reasoning captures something matrices miss.
No credible assessment reports a percentage of correct answers. The conversion runs through a norm distribution: your raw total is compared against the recorded performance of everyone else who has taken the same items, and your position within that distribution is expressed on a scale where the average is set at 100.
This is where free online tests diverge sharply in quality. A serious one norms against a large, filtered sample, discards suspiciously fast or incomplete sittings, and reports a confidence interval rather than a single figure. A weak one applies a fixed lookup table that no one has validated, or worse, applies a curve deliberately tuned to return generous numbers.
The self-selection problem affects even the well-built ones. People who voluntarily seek out an assessment are not a random slice of the population — they skew towards the curious, the educated and the already confident. Norming against that pool rather than a representative sample shifts the whole scale, usually making scores look lower than a clinical equivalent would. Some publishers correct for this; many do not disclose whether they have.
The word carries several different meanings across the sites offering it, and the distinction matters before you invest thirty minutes.
None of these models is inherently dishonest. What separates a reasonable site from a poor one is whether the model is stated before you begin rather than revealed at the end. A straightforward free iq test that shows its terms up front respects your time in a way that a surprise paywall does not.
Immediate scoring is the headline feature of nearly every online option, and it is genuinely useful. Feedback delivered while the experience is fresh helps you connect specific items to your performance in a way a report posted three weeks later never will.
The risk is that instant delivery encourages instant belief. A number appearing within seconds of your final answer carries an air of computational authority that the underlying statistics may not support. Treat speed as a convenience feature, not as evidence of rigour.
Most web assessments impose an overall limit rather than timing each item separately, which shifts the burden of pacing onto the participant. That design choice has consequences worth understanding before you start.
An overall limit rewards people who abandon difficult items quickly and return to them if time allows. It punishes the natural instinct to finish each question before moving on, because a single stubborn matrix can absorb five minutes and cost half a dozen easier items further down. Since later items are usually harder, running out of clock at the end costs less than it would elsewhere, but running out with unattempted items in the middle is expensive.
A minority of tests advertise no limit at all. These measure reasoning depth more purely, but they also make results far less comparable between people, since the person who spent three hours and the person who spent twenty minutes are being scored on the same scale.
Online assessment is easy to criticise, but it does several things well. Access is the obvious one: anyone with a connection can sit a reasoning test at no cost, which was simply not possible before.
Repeat measurement is another. Because the marginal cost is zero, you can track performance across months, which is far more informative than a single isolated figure. Improvement in your own results over time, under consistent conditions, is meaningful even if the absolute number is imprecise.
Sample sizes also work in their favour. A widely used web assessment accumulates response data on a scale no clinical instrument can match, which allows item-level statistics — difficulty, discrimination, misfit detection — to be far more precise than the sample size behind many traditional batteries.
Set expectations accordingly. Most free web tests measure a narrower slice of ability than a full battery, often relying almost entirely on non-verbal pattern items. Working memory and processing speed, both central to formal assessment, are usually absent or crudely approximated.
Conditions are unverified. Nobody knows whether you paused, looked something up, sat with a friend, or restarted after a poor opening. Practice effects are unmanaged: the second attempt at the same item pool will almost always produce a higher figure, and that rise is familiarity rather than growth.
Most importantly, no online result carries formal standing. It cannot support an educational assessment, a diagnosis, or an application to any organisation that requires supervised testing — the admission standards used by high-IQ societies make that condition explicit. The point applies with even more force to assessment of children, where an unsupervised figure is close to meaningless. If the number needs to be recognised by someone else, the sitting has to be supervised, a distinction covered in more depth in the material on accuracy and official assessment.
If you intend to take your score at all seriously, control what you can. Sit in one uninterrupted block at a time of day when you are alert, not late at night after a long shift. Close everything else on the machine and put the phone out of reach. Use a desktop or laptop rather than a phone, since matrix items on a small screen cost accuracy for reasons that have nothing to do with reasoning.
Read the instructions properly before starting the clock, and do not treat the first scored item as a warm-up. If you want the number to mean anything, resist retaking the same test within a few weeks. Practical detail on all of this sits in the material on how to test your IQ and on sustaining attention under pressure.
Handled this way, a free assessment gives you a reasonable estimate and a repeatable baseline. Handled carelessly, it gives you a number that tells you more about your evening than your reasoning.
Continue with the rest of the series on assessment and reasoning: