Rate My YouTube Script: A 10-Point Rubric You Can Score in 10 Minutes
"Rate my script" is a request for a number, and most answers give an opinion instead. This rubric gives the number. Ten checks, each scored 0, 1 or 2, for a total out of 20, scored by you in about ten minutes with the script and the title in front of you. Where a check rests on something we measured, the study is linked. Where it rests on judgement, it says so.
The score describes the draft, not the audience. It is a relative read of where the writing is likely to lose attention, and it cannot know how published viewers will behave.
The rubric
| # | Check | 0 | 1 | 2 |
|---|---|---|---|---|
| 1 | Title match: does the first 90 seconds deliver what the title promises? | The promise is paid off after minute five, or never | Paid off, but the viewer has to wait through setup | The first lines name the promise and start paying it |
| 2 | First line: does it open on a claim, a question or a scene? | A greeting, a channel intro, or "in this video" | A topic statement | A specific claim, number, or image the viewer did not expect |
| 3 | Reason to stay: by the end of the opening, does the viewer know what they will get and why it matters to them? | Neither | One of the two | Both, in the viewer's terms |
| 4 | Time to the first payoff: how long until the viewer receives something they came for? | More than a quarter of the script | Within the first quarter | Within the opening |
| 5 | Structure: can you name the job of each section in one line? | Sections blur into each other | Most sections have a job | Every section has a job and an order you can defend |
| 6 | Pace: words per minute at your natural speaking rate, against the runtime you want | Off by more than 25% | Off by 10 to 25% | Within 10% |
| 7 | Repetition: does any point appear twice in different words? | Three or more repeats | One or two | None |
| 8 | Specifics: are claims carried by numbers, names and examples? | Mostly adjectives | Some specifics, unsourced | Specific and sourced, or clearly first-hand |
| 9 | Exit: does the last section earn the ending, or trail off into a summary and a subscribe ask? | Summary plus ask | Ending with one new thing | A turn, a result, or a question the next video answers |
| 10 | Policy: does any passage match a category in YouTube's advertiser-friendly or inauthentic content guidelines? | A clear match you have not addressed | A borderline passage | Nothing, or the passage is framed with the context the guidelines ask for |
Reading the total
These bands are the rubric's own scale, chosen so that one weak opening cannot hide behind a strong middle. They are not a retention prediction.
- •17 to 20: record it. Spend the remaining time on delivery.
- •12 to 16: one more pass on the checks that scored 0. A single 0 on checks 1 to 4 costs more than two 1s elsewhere, because the opening is where the audience is largest.
- •Under 12: the draft has a structural problem, usually a late payoff or a missing reason to stay. Rewrite the opening before polishing anything else.
Notes on the checks that rest on data
Checks 1 to 3 come from what we measured in the first 45 seconds of 349 videos: openings of flops were 60% more likely to give no reason to keep watching than openings of hits, hits matched their title's promise more often (87% against 78%), and hits carried specific numbers more often (63% against 52%). Details and the raw data are in the 349-video hook study. The small-channel replication on 941 videos from channels with 1,000 to 20,000 subscribers did not reproduce those differences, with one exception: greeting openings leaned toward flops (23.3% against 16.6%). So check 2's 0 for a greeting is the one opening signal that held across both studies, and the rest of checks 1 to 3 are a rubric, not a law.
Check 4 uses our finding that hits and flops reach their point at nearly the same word (14 against 15); the difference was what the point was, not how fast it came. How fast should a video get to the point covers that.
Check 6 uses the corpus figure that a ten-minute video runs about 1,810 words at the median of 3,923 measured videos (caption words divided by full runtime, median 178 words per minute). Measure your own pace with the talking speed test, then convert with the words to minutes converter. Does talking faster help is the reason check 6 scores the match to your target length, not the speed itself.
Check 10 is a read against published guidelines, not a prediction of a monetization decision. The inauthentic content checker and the policy pre-flight in the free script check both report what a passage matches, and does YouTube allow AI-written scripts quotes the relevant policy text.
A paste-in prompt for any assistant
If you would rather have ChatGPT, Claude or Gemini score the rubric, paste the script with this prompt. The two instructions that matter are the quoting rule and the ban on general praise.
Score this YouTube script on the ten checks below, each 0, 1 or 2, and give the total out of 20. For every check scored 0 or 1, quote the exact sentence from the script that caused the score and write the replacement sentence. Do not give general praise. Do not summarise the script. Title: [paste title]. Target runtime: [minutes]. My speaking pace: [words per minute]. Checks: 1 title match in the first 90 seconds; 2 first line opens on a claim, question or scene, not a greeting; 3 reason to stay is stated in the viewer's terms; 4 time to first payoff; 5 each section has one nameable job; 6 word count matches target runtime at my pace; 7 no point made twice; 8 claims carried by numbers, names, examples; 9 the ending earns itself rather than summarising; 10 any passage that matches YouTube's advertiser-friendly or inauthentic content categories. Script: [paste script]
Expect the assistant to be generous on checks 5, 8 and 9 and to miss check 6 unless you give it the pace. Count the words yourself with the script word counter.
What the rubric cannot see
A rubric scores the page. It cannot hear delivery, see the edit, or know the thumbnail, and a 20 out of 20 script can still be the wrong topic for the channel. Why did my YouTube video flop gives the checking order for that larger question: topic first, packaging second, script third.
For a scored read with the problem text quoted and a rewrite returned, the free script check does checks 1 to 10 in one pass and returns a relative attention-risk map over the draft with the biggest flagged passage rewritten. The hook analyzer scores checks 2 and 3 alone from the first three sentences. How to get feedback on a YouTube script covers the human routes.
Try it on your own script
Paste your draft below. You get your hook, structure, and pacing scores, a script-level attention-risk map, and the single biggest issue quoted from your own lines. Free, no login.
Free · No login · See a sample audit first if you prefer.
Frequently asked questions
What is a good score for a YouTube script?
On this rubric, 17 to 20 out of 20 means record it; 12 to 16 means fix the checks that scored 0, starting with the opening; under 12 means a structural rewrite. The bands are the rubric's own scale and not a prediction of retention.
Can ChatGPT rate my YouTube script?
Yes, if you give it a rubric and force specifics. The prompt in this article asks for a score per check, the exact sentence behind each low score, and a replacement sentence, and bans general praise. Supply your speaking pace, or it will skip the runtime check.
Which check matters most?
Checks 1 to 4, the opening. The audience is largest in the first 30 seconds, and the one opening signal that held across both of our hook studies was that greeting openings leaned toward flops. A 0 on any of those four costs more than two 1s later in the script.
Does the free Prepublish check use this rubric?
It covers the same ground: hook, structure and pacing scores, a relative attention-risk map over the draft, a policy pre-flight, and the biggest flagged passage with its rewrite. It reads text only and does not know how published viewers will behave.
Related Articles
We Analyzed 349 YouTube Hooks. Most Hook Advice Did Not Survive.
The first 45 seconds of 349 videos from 36 channels, 254 million combined views, overperformers against flops. Speaking pace and time to payoff showed no difference. Concreteness did.
We Bet Our Own Hook Study Would Replicate on Small Channels. It Did Not.
A preregistered replication on 96 small channels between 1,000 and 20,000 subscribers, 941 analyzed videos. We locked our predictions in public before collecting a single hook. None of the three patterns held up, and we are publishing that.
AI vs Manual Script Review
A direct comparison of AI script analysis and manual review across speed, cost, consistency, and specificity, plus the hybrid workflow that beats both.
Free tools to put this into practice
Want to analyze your own scripts?
Start Script Analysis