Most AI aptitude tools are a chat window. Some accept a screenshot you upload, and ChatGPT can share your screen in supported modes, but either way you are working in the same window as the test and doing the describing yourself. TestSolve is built the other way round: it reads the question where it already is, and puts the worked answer on your phone.
It handles numerical, verbal, deductive, inductive and abstract reasoning, and error checking. On our own 50-question benchmark it answered 46 correctly, which is 92%. That number is measured, and how it was measured is described below.
First question free. No account, no subscription, no card.
Windows and macOS, or no install at all in the browser version. See how it works
A general assistant is good at a question you can hand it cleanly. Most aptitude questions are not that. A numerical reasoning item is a table of figures with a note underneath saying the values are in thousands, and the question turns on which column you read. An inductive item is six shapes in a row. You can screenshot either and upload it, but you are leaving the test window to do it, and on a clock that is the expensive part.
And describing a figural series in words, rather than capturing it, means doing the hard half of the question yourself before you ask. Aon's scales numerical gives 37 tasks in 12 minutes, which averages out at roughly nineteen seconds a task. There is no room in that for a detour.
Reading the screen removes that step. The question arrives as it actually appears, including the parts a person would forget to mention.
| Chat assistants and GPT wrappers | TestSolve | |
|---|---|---|
| How the question gets in | You retype it, describe it, or upload a screenshot | Read from the screen as displayed |
| Data tables and charts | Uploading a screenshot works; transcribing by hand often does not | Read directly, including the footnotes |
| Abstract and figural series | Possible by screenshot, impractical to describe in words | Read as images |
| Time cost per question | Window switching and uploading, on a clock | Seconds, without leaving the test window |
| Where the answer appears | In the same window as the question | On a second device |
| Cost | Free, or a monthly subscription | $14.99 for 30 questions, paid once |
Three steps, and only the middle one is ours.
One. Run the desktop app, or open the browser version and share your screen. The browser version installs nothing.
Two. When a question is on screen the solver reads it, works out what kind of question it is, and routes it to the method that suits it. A numerical item gets arithmetic with the figures checked back against the table. A figural item gets pattern analysis. A verbal item gets the passage weighed against the claim.
Three. The answer appears on your phone with the working, so nothing is drawn over the window you are reading.
92%. Forty-six of fifty questions, on 2026-09-18.
The method matters more than the number. We keep a harness that renders questions through the same code the practice tests use, sends each one through the live solving pipeline exactly as a real capture would go, and marks the result against the key. It is not a hand-picked sample and it is not a demo. It is the product, measured.
The figure has moved before and will move again, and we would rather publish one we can defend than a rounder one we cannot. Earlier copy on this site quoted 94% and 96%. Nobody had measured either, so both were removed in September 2026.
Where it is strongest: numerical data interpretation and figural series. Where it is weakest: a question that is a picture with no readable text at all, where it will tell you it could not read the item rather than guess at one.
Aptitude tests are built from a small number of formats and the publishers recombine them. What matters is the format in front of you, not whose logo is on the page.
| Format | What it looks like | How it goes |
|---|---|---|
| Numerical reasoning | A table or chart, then a calculation, usually five options | Strongest format. Reads the figures and shows the arithmetic |
| Verbal reasoning | A passage, then a claim to judge true, false or cannot say | Solid, including the cannot say judgement |
| Deductive reasoning | Statements, then which option must or cannot follow | Solid. Five options is the common shape |
| Inductive and abstract | Shapes in a series, or a matrix with one cell missing | Good, and the clearest case for reading the screen |
| Error checking | Reference codes against copies, under time pressure | Good. Mostly a speed test |
| Personality questionnaires | Rate how much you agree with a statement | Not solvable. There is no right answer to find |
| Game-based assessments | Balloons, cards, reaction tasks | Not solvable. They score how you play, not what you answer |
It is a solver, not a leaked answer key. It does not look your test up in a database of past papers. Those databases are wrong about as often as they are right, and they go stale the moment a publisher rotates its item pool, which the adaptive ones do continuously.
It will not help on a personality questionnaire or a strengths inventory, because there is nothing there to solve. It will not help on a game-based assessment, which measures the pattern of your play rather than a set of answers. Anyone selling you answers to those is selling you nothing.
And it is not magic on a bad screen. If the question is a low resolution image with no text in it at all, it will say so rather than invent something.
The first question is free and needs no account. After that it is $14.99 for 30 questions or $19.99 for 50. One payment. There is no subscription and the credits do not expire.
Most people take the 50 after trying the free question, which covers a full graduate battery with room to spare.
One free question, no card, no signup. Then decide.
Or run it in the browser with nothing to install. Would rather practise first? Take a free timed test.
Software that works out the answer to an aptitude question rather than looking it up. TestSolve reads the question off your screen, solves it, and shows the answer with the working on a second device.
TestSolve measures itself at 92% on a 50-question benchmark run through its live pipeline. General-purpose chat assistants do well on questions you can type out, and lose accuracy on data tables and figural series where the information is visual and cannot be pasted in.
It can solve a question you can retype accurately, which mostly means word problems. It cannot see your screen, so a data table has to be transcribed figure by figure and an abstract series of shapes cannot realistically be described at all. On a timed test that transcription usually costs more time than the question is worth.
It reads the question as displayed instead of asking you to describe it. That matters most on the formats where the information is visual: tables with footnotes, charts, figural series and matrices.
Yes, and that is its strongest format. It reads the figures from tables and charts, does the calculation and shows the working.
Yes. Shapes in a series and matrices with a missing cell are read as images, which is the clearest case for reading the screen rather than typing a description.
Personality questionnaires and strengths inventories, which have no right answers, and game-based assessments, which score the pattern of your play rather than a set of answers.
Yes, a desktop app for both, plus a browser version that installs nothing and shares your screen instead.
The first question is free with no account. After that it is $14.99 for 30 questions or $19.99 for 50, paid once, with no subscription.