How we verify our content
Last full verification pass: August 2026 — every module re-read · applies to both the SAT and ACT question banks
A student using this site has no tutor standing behind them to catch a mistake. That raises the bar: a wrong answer key here isn’t an inconvenience, it teaches a child the wrong thing. So we treat accuracy as an engineering problem, not a promise. This page explains exactly what we check, how, and — just as important — what we have not yet proven, because a trust page that overstates is worth nothing.
math answer re-derived from scratch by an independent computer-algebra engine — not by trusting our own worked solution.
separate AI systems, built by different companies, review the content. Every error either flags is verified by a human before anything changes.
known wrong items remain on either exam. Everything found in every pass has been fixed.
1. Math: answers proven, not assumed
For every math question on both tests, the correct answer is re-derived independently by a computer-algebra system that reads only the words and numbers printed in the question — it never looks at our answer key or our explanation. If its result disagrees with the keyed answer by even a rounding step, the item is stopped and rewritten. Across the whole bank this independent solver reproduced our keys with zero disagreements. A right answer reached by wrong working teaches the wrong method, so on the ACT every math explanation was also read in full by a human and every per-choice rationale checked against the value that choice actually holds.
2. A validator that blocks whole classes of mistake
Before any change ships, an automated checker runs over both banks and refuses to pass a file that contains a structural fault. It catches, among others:
- two answer choices that are the same value written differently (e.g. “25%” and “0.25”, or the same number in and out of scientific notation);
- a per-choice explanation that names an answer by letter while the site shuffles the choices — so the explanation can never point at the wrong option on screen;
- a rationale array that doesn’t line up one-to-one with the choices, an empty field, or an answer index out of range.
This is why a mistake we fix once cannot quietly come back: the machine that would have caught it is now part of every build.
3. A second, independent reviewer
We ask a second AI — a different model, trained by a different company — to attack the content adversarially: to assume an error exists and hunt for it, to argue for every wrong choice, and to fact-check every date, figure and scientific claim against authoritative sources on the open web. That is the closest thing to an independent examiner a free project can get. Crucially, we treat its output as leads, not verdicts. A second model catches a different slice of problems than we do — and it also raises false alarms. So every single flag is re-checked by hand, and we change content only when the error is real. (In our most recent pass, several “critical” flags turned out to be correct items the reviewer had misread; we did not change them, and we hardened the validator so the pattern can’t confuse anyone next time.)
4. Passages: fact-checked and original
Reading and science passages are checked two ways. Every real-world claim — a date, a mechanism, a measurement — is verified against authoritative sources. And distinctive sentences are searched against published text to confirm the writing is original: our passages describe real topics in our own words and attribute to invented authors and works, never copying a real source. The two famous quotations that appear are public-domain and framed as genuine.
5. Security and privacy, audited too
The same standard applies off the page. The site has been through a full security red-team (real attacks run against a live copy) and a children’s-privacy (COPPA) audit. No advertising, no trackers, no sale of data, and no third-party cookie of any kind. Details are in our Privacy Policy and Children’s Privacy Notice.
What we honestly do not yet claim
“Verified” means different things for different content, and we won’t blur them:
| Content | Level of verification |
|---|---|
| Math answer keys (SAT & ACT) | Machine-proven — every solvable item independently re-derived by computer algebra, zero disagreements. |
| ACT math explanations | Read in full by a human, every per-choice rationale checked against the value it describes. |
| ACT English, Reading & Science | Every item adversarially re-read, module by module — keys, distractors, rationales and passage facts — with every flag verified by hand and fixed. |
| SAT Reading & Writing | Every item adversarially re-read, module by module, with every flag verified by hand and fixed. |
No AI — ours or anyone’s — can prove a body of writing is 100% perfect. But every question on both tests has now been re-read adversarially, module by module — a reviewer actively trying to break each item rather than approve it — with every flag checked by hand before any change. Combined with machine-proven math keys and a validator that blocks whole classes of error, that is a stronger claim than any “reviewed by experts” badge: every math answer machine-verified, every question structurally validated, and every item independently re-read with each flag verified by hand. We keep re-checking as students report anything that looks off.
Found something wrong? Tell us.
The last line of defense is you. If any answer, explanation, or fact looks off — even slightly — please report it here. We read every report, verify it, and fix real errors fast. That feedback loop is part of the method, not an afterthought.
This is an independent study tool. Our accuracy process is our own; it is not endorsed by or affiliated with the test makers.