You do not need a video host, an analytics tool, or another app for this. VSL Conversion Diagnosis is already in JeremyAI. You give it the offer, the opening of the VSL, where it sends people, and the numbers for a clearly defined period. It says whether the data points at the script at all, separates facts from assumptions, and tells you what to measure before you rewrite anything.
This sample is not a proven winner. One 14-day period of views, bookings, and sales is a starting point. It is not proof the VSL is working, and it is not proof the VSL is broken.
Watch the walkthrough
The walkthrough is one run inside JeremyAI. The worker opens, the sample VSL and numbers go in, and the review comes back in the same chat. If the player does not load, open the walkthrough here.
This example shows a first-pass diagnosis. Check its assumptions against your show data, your retention curve, and your sales pipeline before you change the script. Your answer may differ.
What this does
Takes one VSL: the offer, the opening, the next step, and the numbers for one bounded period.
Says first whether the data supports a constraint in the VSL at all.
Separates the facts in the numbers from the assumptions you are running on.
Maps where the count drops, stage by stage, instead of judging the script from the top of the funnel.
Names what to preserve: the offer, delivery, audience, destination, and substantiated claims. Does not add unapproved claims.
Rewrites one section only if the script is the constraint. If the constraint is the page, the offer, the traffic, or missing proof, it says so and leaves the script alone.
Gives you what to measure next, and the result that would support or kill the hypothesis.
Stays in JeremyAI. It does not change the page.
When to use it
Use it when a VSL is getting views but not enough bookings, applications, or sales, and you are about to rewrite the script. Run it first so the rewrite goes at the real constraint, not the first line that feels weak.
If you only have some of the counts, bring what you have and mark the rest unknown. The worker should keep the diagnosis provisional when the data is thin.
Before you start
Signed in at
jeremy.utari.ai.VSL Conversion Diagnosis visible in Workers. It is a shared JeremyAI worker, so you do not build it.
The offer, the price, who it is for, and what the customer actually receives.
The opening of the VSL, or the section you are worried about.
Where the VSL sends people, and the action you want.
Numbers for one completed period: views, retention, clicks, bookings, shows, sales, and spend. Mark unknown fields as unknown.
Only the proof you can substantiate. If you have none, say no proof yet.
Do it
Go to Work, then Workers, and select VSL Conversion Diagnosis. Open Chat. Do not open settings.
Paste the VSL and the context. The walkthrough uses the sample prompt below. For your own VSL, keep the shape and replace the fields.
Send it. The review comes back in the same chat.
Read the review before you rewrite anything. Write down whether the data supports a script constraint, what is fact, and what is assumed.
If the reply says the script is not the constraint, that is the result. Do not ask for a rewrite over that answer. Run the measurements first.
How to read the sample
The walkthrough uses a sample, labeled as sample. The offer is a $2,500 coaching program for agency owners, delivered over 12 weeks with weekly calls. It is coaching, not software and not done-for-you ads. The VSL goes to cold traffic, agency owners doing under $30k a month, and sends people to a booking page for a sales call.
The opening says: You do not need more clients. You need a way to see which accounts are actually losing money. In the next 12 minutes I will show you the audit we run before we touch ads, offers, or follow-up. If you want the same review on your account, book a call at the end.
For one completed 14-day period, the supplied data is 400 VSL views, 140 watched past 25 percent, 36 clicked to the booking page, 11 booked calls, 2 sales, and $1,800 in spend. Shows, refunds, and a comparison period are unknown. No proof is supplied.
Those numbers describe one window. They do not establish profit, and they do not explain why anyone booked or bought. The worker should keep the diagnosis provisional and should not invent testimonials, guarantees, scarcity, or delivery details.
What JeremyAI returned
The reply opened with the call: "The data doesn't convict the script, so I'm not rewriting anything yet." It did not write a draft section. For this prompt, which asked for a rewrite only if the script was the issue, that is the right outcome.
It then worked through the sample in this order:
Facts. 140 of 400 views were still watching at the 25% mark, which is 35%. The 36 clicks are 9% of views and about 26% of the engaged viewers. 11 bookings from 36 clicks is a 31% booking page conversion. 2 sales from 11 bookings is an 18% close, but only if every booking showed. That is $164 per booked call and $900 per sale.
Assumptions. That all 11 bookings came from this VSL inside the window. That 2 sales is the final count and not a lag, since some of the calls may not have happened yet. That the 35% retention says something about the opening and not a mid-video drop. That the 36 clickers were the engaged viewers. With shows and refunds unknown and no comparison period, the real close rate and return on spend are not known yet.
Leak map. The reply flagged the step from engaged viewer to booking click, 140 down to 36, as the leak to watch. Whether those 36 clickers were among the 140 engaged viewers is not known yet. Click to book at 31% looked fine to the worker. The 18% close is below the floor the worker wants on showed calls, but at 11 bookings one more sale moves it to 27%, so it treated that as noise until show data lands.
Three candidate constraints. The viewer-to-click step, the show and close side, or the lack of any proof supplied for this review. It said there is not enough signal to pick one yet.
Proof. The reply treated the missing proof as a gap to close whichever constraint it turns out to be. It called that a proof-building problem, not a script rewrite, "so the draft stays parked."
The hook. It flagged the opening line without touching it. In the reply's view, telling a sub-$30k agency owner they do not need more clients contradicts what that buyer feels, and there is no proof in the video to carry the premise. It did not call the hook broken. It said that if the retention curve shows an early cliff, that line is the first suspect.
What to measure next
The reply listed these in order. Use the same period as the prompt.
Show rate on the 11 bookings. Find out where the other 9 sit in the pipeline. Check whether any calls or sales are still pending, since the sales count may be undercounted.
Retention curve at 10% intervals. This shows whether the drop is in the first 45 seconds or mid-video.
Whether the 36 clickers were engaged viewers. Tag whether they sat inside the 140 who watched past 25%, or outside it.
More volume before any verdict. For a sample this size, the reply recommended at least 3x the volume before calling anything. In its words, 400 views and 11 bookings "isn't a sample, it's a rumor."
What would support or kill the script hypothesis
The reply ended with a decision line. Hold the next move to it:
Supports the script as the constraint. In the reply's read, the script hypothesis is supported if the retention cliff lands inside the first 25%, the clicks are mostly non-engaged viewers, and booking conversion holds around 30% when engaged viewers reach the page. Only then would the worker rewrite the opening, with proof slots in brackets.
Kills it. The curve is smooth, clicks line up with engaged viewers, and booking conversion holds. In the reply's read, that weakens the script hypothesis and moves the next investigation to show rate and close rate, a sales process problem a rewrite would not fix.
Points at the offer-to-call match. The reply said that if shows come back at 70% or higher and close is still under 20%, the offer-to-call match is the problem to look at.
What not to do yet
Do not rewrite the VSL from this one reply. The reply found the data does not convict the script.
Do not change the hook before you have the retention curve. It is a suspect, not a finding.
Do not fill the proof gap with claims you cannot substantiate. Leave proof slots in brackets until you have real results.
Do not read the 2 sales as a close rate or the spend as a return. Shows and refunds are unknown.
The prompt from the walkthrough
Copy the block below for the same run as the video. For your own business, replace the offer, the opening, the period, and the counts. Mark anything you do not know as unknown.
I have a VSL I want reviewed. This is sample data, not a proven winner. Find the likely constraint, then rewrite one section only if the script is the issue.
Business and offer: a $2,500 coaching program for agency owners. It is a 12-week delivery with weekly calls. Not software. Not done-for-you ads.
Audience: cold traffic, agency owners doing under $30k a month.
Next step: the VSL sends people to a booking page for a sales call.
VSL opening, first 45 seconds:
You do not need more clients. You need a way to see which accounts are actually losing money. In the next 12 minutes I will show you the audit we run before we touch ads, offers, or follow-up. If you want the same review on your account, book a call at the end.
Performance for one completed 14-day period:
- VSL views: 400
- Watched past 25 percent: 140
- Clicked to the booking page: 36
- Booked calls: 11
- Shows: unknown
- Sales: 2
- Spend: $1,800
- Refunds: unknown
- Comparison period: none
Proof we can use: no proof yet.
First, say whether this data supports a constraint in the VSL. Separate facts from assumptions. Say what to preserve: the offer, the delivery, the audience, the destination, and any substantiated claim. Do not add unapproved claims.
If the script is the issue, rewrite one section only and label what changed. If the constraint is the booking page, the offer, the traffic, or missing proof, say that and do not rewrite the script.
Do not invent results, testimonials, guarantees, scarcity, time commitments, or delivery details. Leave missing claims in brackets.
Tell me what to measure next, and what result would support or kill the hypothesis.
Keep this to analysis and one draft section. Do not change the page.
A prompt for your own VSL
Use this when the sample is not your offer. Leave unknown fields as unknown.
I have a VSL I want reviewed. Find the likely constraint, then rewrite one section only if the script is the issue.
Business and offer: [what we sell, price, who it is for, and the delivery model]
Audience: [traffic source and who is watching]
Next step: [where the VSL sends people and the action we want]
VSL opening or section to review:
[paste the opening, or the section you are worried about]
Performance for [one completed period]:
- VSL views: [count]
- Watched past [percent]: [count or unknown]
- Clicked to the next step: [count or unknown]
- Booked calls, applications, or purchases: [count or unknown]
- Shows: [count or unknown]
- Sales: [count or unknown]
- Spend: [amount]
- Refunds: [count or unknown]
- Comparison period: [details or none]
Proof we can use: [verified results, or no proof yet]
First, say whether this data supports a constraint in the VSL. Separate facts from assumptions. Say what to preserve: the offer, the delivery, the audience, the destination, and any substantiated claim. Do not add unapproved claims.
If the script is the issue, rewrite one section only and label what changed. If the constraint is the booking page, the offer, the traffic, or missing proof, say that and do not rewrite the script.
Do not invent results, testimonials, guarantees, scarcity, time commitments, or delivery details. Leave missing claims in brackets.
Tell me what to measure next, and what result would support or kill the hypothesis.
Keep this to analysis and one draft section. Do not change the page.
Bring the numbers back
When you have the show rate and the retention curve, paste this into the same chat.
Here is the show rate on the bookings from that period: [data].
Here is the retention curve at 10% intervals: [data].
Here is how many of the clickers also watched past [percent]: [data].
Does the script hypothesis still hold, or is the constraint the show and close side, the offer-to-call match, or missing proof?
If the script is now the constraint, rewrite one section only, keep proof slots in brackets, and label what changed.
If the worker is missing
Refresh the worker list. If it is still not there, email [email protected] and include the account email. Do not build a replacement for this walkthrough.
The how-to for this worker is VSL Conversion Diagnosis Worker.
Support
Stuck on a step? Email [email protected]. Include the account email, the worker name, and a screenshot of the chat.