How every answer is scored.
Every task on this site is scored on the same five axes, and the total decides the verdict. No verdict is ever picked first and justified afterwards, and no verdict is for sale.
The five axes
Five axes, each scored 0–2.
| Output | Can a current model produce the artefact at all? |
|---|---|
| Inputs | Can the reader actually supply what the task needs? |
| Verification | Can the reader check the result without the expertise they were avoiding? |
| Liability | Who carries the consequences if it is wrong? |
| Effort delta | How much of the work does the model genuinely absorb? |
The verdict
- YES: total 8–10 with no axis at 0.
- PARTLY: total 4–7, or 8+ with exactly one axis at 0.
- NO: total 0–3, or Verification and Liability both 0.
- Tasks whose output legally requires sign-off by a regulated professional cap at PARTLY regardless of score, and the answer says so.
How answers are produced
The founding answers were researched and scored by hand, each backed by a stored, dated model output. As the site grows, most new answers are assessed by an automated pipeline that applies the same rubric axis by axis; each of those pages says so, with the model that assessed it and the date, stamped at the bottom of the page.
Before an assessed page publishes it has to pass a set of mechanical checks: the verdict must match its axis scores under the thresholds above, numeric claims must trace to supplied data rather than the model's memory, British English throughout, and health, legal and money pages must carry the right signposting. A second model from a different company then re-reads every page and flags anything inconsistent or overclaimed; flagged pages wait for a person.
Humans stay in the loop: a sample of every publish batch is re-read by a person, and any page the checks or the second model flag waits for review before it can go live. Health, legal and money pages carry mandatory signposting, checked mechanically on every one.
Verdicts age, so every answer carries its date, and answers past ninety days without a re-check are listed openly on the staleness register.
Found an error?
Email team@canaidoit.ai with the page and what is wrong. Reported pages are re-checked by a person, and a corrected answer records the change in its history.