I Rebuilt A Product That Sold 2,000+ Copies. Here Is What It Was Missing.

Webinar Replay · Fightback WhatsApp Group

I took a tool that has sold more than two thousand copies on WarriorPlus, pulled it apart with Page Sniper, and rebuilt it live as something I would actually use. The result is Traffic Genesis, and it is free.

This was a live session for the Fightback WhatsApp group. The whole build is in the replay above, permission prompts, waiting around and all. What follows is the short version, plus the part I think actually matters.

Why I picked this particular product

I am not going to name it, because this is not a takedown. The person who built it has done something genuinely impressive, and I said so on the call: he has trained an audience to buy from him without hesitating. That is hard, it took years, and he did not start from nothing.

What made it worth studying is the numbers. The affiliate dashboard showed more than two thousand copies sold at a twenty four percent conversion rate, and that figure is the post-launch average. During launch week these offers often run far higher. It is a low priced front end, it sells, and it builds a buyers list. That is the model working exactly as intended.

So the product has proven demand. That is the only thing I needed to know before deciding it was worth rebuilding.

What the tool actually was

Underneath, it is a prompt generator. You put in a niche, it hands you prompts to copy into ChatGPT: prompts for keywords, then a blueprint, then the article, then a humanising pass, then images.

And here is the distinction that the whole session came down to. It sells the process, not the outcome. You get the steps. You still do all the work, in another tab, and you still end up wherever your own judgement takes you.

That is not automatically bad. Plenty of people want the process. But if a sales page promises articles that rank, the process is only half of what you would need.

I tested it properly before criticising it

This is the part I want to be careful about, because it would be easy and lazy to just assert that someone else’s tool is weak.

I ran it twice. Once with the prompts as supplied. Then again with the prompts improved, rewritten by Claude to be as strong as I could make them. I wanted to know whether the ceiling was the prompts or the approach.

Both times the output was too generic, and both times it was missing the things that actually matter in 2026. Not unreadable. Not embarrassing. Just the same article anybody with a ChatGPT account could produce in two minutes, which is precisely the problem. If everyone can generate it, it does not differentiate you, and it is not going to get picked up by Google or quoted by an AI engine.

So the ceiling was not the prompts. It was the approach.

How Page Sniper does the teardown

The method in the replay is the same one every time:

• Collect everything. Point Page Sniper at the sales page, the JV page, the affiliate page, any video with a transcript. Each click adds to what it has already gathered.
• Run the research. Six steps, point and click, which produces the angles, the claims and the positioning that made the offer convert.
• Flip it. Switch from “promote this offer” to building a better version. All that research carries over, so it already knows what the product does and where it stops.
• Find the bottleneck. This is the whole point. Not “what features are missing” but what actually holds the user back from the result they were sold.
• Take the build spec into Claude Code and build.

For this product the bottleneck was obvious once it was written down. You are handed prompts and left to do the work.

What Traffic Genesis does instead

The original flow was keyword, blueprint, article, humanise, images, all as prompts you copy elsewhere. Traffic Genesis runs the whole thing on one page and actually does the steps:

• Start from a niche and get real keyword candidates, not a prompt asking you to think of some.
• Score them for whether you can win before you commit to writing anything, which is the step the original skipped entirely.
• Pick an angle rather than accepting the first framing the model offers.
• Write the article in the tool, with answer-first sections, because that is what gets lifted and quoted.
• Humanise it, get image prompts, then copy out as Markdown or HTML.
• Save projects so you can come back to a piece, and bring your own AI key so you are not paying me for tokens.

It is free, as promised on the call, and it is on the site now.

What is still being built

Being straight about this, because the replay shows me building things that have not shipped yet.

During the session I started on a keyword and entity optimiser, the panel that highlights your target terms in green as the article actually uses them. That is not live yet. Neither is the stats and data research step, or the embeddable calculator idea I talked through near the end. They are the next things I want in it.

Which is also why I did not share the tool in the WhatsApp group when I first built it. It was not good enough to use myself yet, and that is the only bar I trust.

My Take: The lesson from this one is not that somebody else built a weak tool. It is the difference between selling the process and selling the outcome. A prompt generator sells you the process, and the sales page often sells you the outcome, and those two things are not the same product. I tested that tool with its own prompts and again with much better prompts that Claude helped me write, and both times the output was too generic and missing what matters in 2026. That told me the prompts were never the ceiling. So we built the outcome instead, with a lot more in it, and made it free like I said I would on the call. The question I would ask before shipping anything now is simple: did I build this to use myself, or did I build it for a launch? You can usually tell within five minutes of using someone’s tool which one it was.