Free to read · no account needed

16 min · foundations

How to spot a bad AI answer in under a minute

Goal: Use a short checklist before you paste AI text into Slack, email, docs, or class work.

Best for: Anyone burned by fluent, confident wrong answers.

This pairs with the branded 1-minute verify ritual for claims. Here the focus is *the whole answer*: relevance, fluff, defensibility, and whether the next step is human-doable.

Eval means you score outputs against a frozen checklist—not against whether you enjoyed reading them. Ten minutes of eval beats an hour of vibes.

Why beginners skip this

Skipping the hard part of “How to spot a bad AI answer in under a minute” feels fine until a live thread pays the bill. Hold: Checklist before paste: facts, tone, scrub, human detail—check it once this week, or it stays a bookmark Mid-week cost shows up as rework, not as a missing bookmark.

Step 1 — Sixty-second checklist

  • Does it answer the ask you typed—or a neighboring question?
  • Any numbers, names, dates, or links you did not provide? Verify or delete.
  • Would you defend every sentence in a meeting with your name on it?
  • Is the next step something a human can actually do Monday?
  • Did it invent a softer reality (fake discounts, fake citations) to please you?

Step 2 — Force a self-check

Do this step with a real artifact open—not from memory.

Re-read your last answer.
List: (1) claims I did not give evidence for (2) fluff sentences (3) one risk if I send this as-is.
Then return a tighter version with [?] on unverified claims. Do not add new facts to look complete.
Example

Filled prompt

Almost-sent LinkedIn post with a metric I cannot source.
Checklist: facts / tone / scrub / human detail. Kill or fix before post.

What good output looks like

  • Labels trust/check/stop correctly for “Use a short checklist before you paste AI text int”
  • Does not treat the model as a database or authority
  • Context pack stays ≤ one screen when asked
  • Pass/fail: you can explain the limit in one line

Step 3 — Risk tiers (reuse)

Low risk: subject lines, reordering your bullets. High risk: client numbers, class citations, health/legal/money—mandatory verify.

Example

Filled prompt

Draft I almost pasted to Slack about launch timing.
Run checklist aloud: facts checkable? tone mine? secrets scrubbed? human detail added?
Return pass/fail per item. I will fix fails before send.

What good output looks like

  • Does not treat the model as a database or authority
  • Context pack stays ≤ one screen when asked
  • Pass/fail: you can explain the limit in one line
  • Labels trust/check/stop correctly for “Use a short checklist before you paste AI text int”

Save this card

Card: Checklist before paste: facts, tone, scrub, human detail—check it once this week, or it stays a bookmark. Keep this beside the tool; read before you start.

1. What I shipped: ___ 2. What I refused: ___ 3. What I will never-do again: ___ (date: Asia/Shanghai)

  • [ ] Tools/writes paused
  • [ ] Refuse-line dated
  • [ ] Artifact path saved
  • [ ] Next link chosen (or stop)

Before you start

Sit with three real asks from this week. You will label trust/check/stop for “Use a short checklist before you paste AI”—bring the asks, not slogans. Guide: foundations-eval-your-own-outputs. If you only have a blank practice box, stop and fetch the real job first.

  • Goal in one line: ___
  • Allowed tools: ___
  • Forbidden: ___
  • Human approve needed for: ___

How to use this week

This week: before pasting anything out, run facts / tone / scrub / human-detail on one live answer. Log one catch on a live artifact—not a blank demo. Write one refuse-line and date the note (Asia/Shanghai). If you only skimmed foundations-eval-your-own-outputs, put 25 minutes on the calendar and redo once.

Score before you fall in love

Freeze the rubric, generate once, score, then decide. Liking first is not eval.

Keep it real

Run this guide’s Practice now on work you will actually ship this week—not a fictional scenario. Dated notes beat perfect theory.

Do not bookmark-only

Finish Practice now, then tick this guide on your path. Titles unread are not skills.

Close the loop

After Practice now, write one line: what you will refuse next time. That line is the habit.

Example

Filled prompt

AI draft for customer email claims "we'll refund automatically."
Checklist aloud. I never approved refunds.
Fix or kill before send. Log what failed.

What good output looks like

  • Context pack stays ≤ one screen when asked
  • Labels trust/check/stop correctly for “Use a short checklist before you paste AI text int”
  • Pass/fail: you can explain the limit in one line
  • Does not treat the model as a database or authority

Red flag

Forwarding without the one-minute check; “reads right” replacing spot-checks—re-check names and numbers.

Pitfalls

PitfallFix
Only checking toneCheck claims
Asking “any issues?” vaguelyUse the numbered self-check
Sending because it sounds smartMeeting-defense test
Example

Filled prompt

Temptation: skip the habit for foundations-eval-your-own-outputs to “save five minutes.”
Job still: Use a short checklist before you paste AI
Do the smallest correct pass: Done visible · one checkable artifact · Unknowns labeled [?].
Boundaries: draft only; no silent send; no invented metrics.
Return: the artifact + the refuse-line you said aloud.

What good output looks like

  • Names the shortcut you refused (specific to foundations-eval-your-own-outputs)
  • Still delivers a checkable artifact for “Use a short checklist before you paste AI”
  • Unknowns labeled; draft-only respected
  • Pass/fail: refuse-line written and dated (foundations-eval-your-own-outputs #4)

Practice now

Live card: apply “How to spot a bad AI answer in under a minute” to a mistake from this week; write the fix habit (Asia/Shanghai).

Today: before sending one draft, run the one-minute check (names, numbers, next action). Fix one miss you catch (Asia/Shanghai).

Today card (copy) — foundations-eval-your-own-outputs: 1. Artifact I will produce: ___ 2. Pass/fail check: ___ 3. Refuse-line if tempted to skip verify/send: ___ (date: Asia/Shanghai) 4. One never-do from this guide: ___

  • [ ] Artifact path written
  • [ ] Refuse example would say "yes"
  • [ ] Dated Asia/Shanghai

One sentence takeaway

Checklist before paste: facts, tone, scrub, human detail—check it once this week, or it stays a bookmark

  • You can point to a dated refuse-line
  • You have one artifact path from tonight
  • You can teach the red flag in one sentence

Next

1-minute verify · Rubrics · Claim–evidence matrix · Verify before you share · Reject AI slop

Pair with today: open the first Next link (foundations-when-ai-sounds-sure-but-wrong) only after you dated a refuse-line for this guide (Asia/Shanghai).

← Back to home · Learn · Tools · Compare · Start