Documentation

The AI ad breakdown rubric

Twelve questions to take apart a video ad, yours or a competitor’s, with a close look at where AI helps and where it misleads. They distil the questions the e-commerce consultancy Bluesense Digital keeps asking in its ad teardowns on YouTube. The scoring and the wording are ours.

The 12 questions

QuestionWhat good looks likeRed flag
1. Who does the hook attract?A stranger gets it in about 3 seconds; it speaks to the buyer; it isn’t a tired lineIt hooks everyone, buyers or not
2. Does the person on screen match the buyer?The age, situation and setting of the real buyerA presenter the buyers wouldn’t see themselves in
3. Is it one awareness stage throughout?Hook, body, on-screen text and copy speak to the same stageA curiosity hook bolted onto a product-aware body
4. When does the product appear?Late for cold viewers; early only in offer or reminder adsThe product in every cut of a cold ad’s first 3 seconds
5. Are the bridges smooth?The sense of being sold to rises gently to the CTAA staircase: product, pitch, “ours is better”, each a drop-off
6. Benefit first, mechanism before product?One benefit, and why it works before the product’s nameA feature list up front
7. Does it deal with the alternatives?It names what they’ve tried and why that fell shortIt ignores the obvious “I already have one”
8. What proof, and how specific?Shown on screen, measured, matched to the personaVague claims, invented numbers, one famous face carrying it all
9. Do offer, risk reversal and CTA fit the tone?A clear next step; softer for creator or educational adsNo CTA, a vague offer, or a hard sell at the end of a story
10. Pace and feel?A cut every 3 seconds or less, text overlays, movement; it looks native to the feedA static talking head, a glossy studio look, text too small for a phone
11. How is AI used?To show what a camera can’t: the problem, a metaphor, the harm of alternatives; the real product when it appearsAI showing the product doing what it can’t, AI people presented as customers, AI-script tells
12. Does the landing page continue the angle?Same promise, same words, ideally the same video at the topA generic product page, or a broken link
  • AI-script tells, per the channel: “it’s not X, it’s Y” lines, the same clipped rhythm in every sentence, no contractions, stock metaphors (video).
  • Good AI use, in its teardowns: AI for the hook and the problem, then a cut to the real product, a real demo and real proof (video).
  • The red line: an AI demo of things the product can’t do, which the channel calls close to fraud: it breaks trust and brings refunds (video).

How to score

  • Give each question 0 (red flag), 1 (partly) or 2 (good), for a total out of 24.
  • 20 to 24: ready to test. 14 to 19: fix the lowest scores first. Under 14: rewrite before you produce.
  • A red flag on question 11 (faked capability, an AI person claiming results, invented figures) means you don’t run the ad, whatever the total.
  • Score the script before production, when changes are cheap, then score the finished video.

Copyable checklist

Ad breakdown sheet
AD: ____________   STAGE: ____________
Score: 0 = red flag, 1 = partly, 2 = good

1. Hook attracts the buyer, clear in 3 s ......... _/2
2. Person on screen matches the buyer ............ _/2
3. One awareness stage in every layer ............ _/2
4. Product timing fits the stage ................. _/2
5. Smooth bridges, no staircase .................. _/2
6. Benefit first, mechanism before product ....... _/2
7. Alternatives dealt with ....................... _/2
8. Specific, real proof .......................... _/2
9. Offer, risk reversal and CTA fit the tone ..... _/2
10. A cut every 3 s or less, looks native ........ _/2
11. AI shows, never fakes; no AI-script tells .... _/2
12. Landing page continues the angle ............. _/2

TOTAL: __/24   Red flag on 11? yes / no (yes = don’t run)

With Dokiri Studio

Score a Dokiri script on its review page, where the cast, shots and lines appear before you pay, and use Change something to fix the weak answers (review and rewrite). The free virality score gives a quick automated read out of 100 on 8 criteria; this rubric is the slower, manual one. Question 12 is outside Dokiri, which makes the video, not the page. People in Dokiri ads are drawn by AI: never present them as real customers or testimonials. More in AI ads and Meta’s rules.