Show IH: AI Ad Generator – Reverse-engineering winning ads into a logical process
Hi IH,
Like many of you, I’ve found that the biggest friction in launching a new SaaS isn't the engineering—it's the distribution. We often treat ad creative as a black box and rely on guesswork, or worse, we ask an LLM to "be creative" and get back generic, unusable fluff.
I built AI Ad Generator to turn creative research into a repeatable, logical engineering process.
The Core Logic:
Instead of just prompting an LLM, the system follows a 3-step pipeline:
- Context Scraping: It automatically extracts USPs and core value props directly from a landing page URL—no more manual copy-pasting.
- Benchmark Deconstruction: It analyzes high-performing ad frameworks that are already proven to scale in the wild (the "DNA" of winning ads).
- DNA Reconstruction: It maps your product's specific variables into these proven structures to generate platform-specific hooks and copy that actually follow marketing logic.
The goal is to provide indie hackers and small teams with the same strategic depth usually reserved for high-end creative agencies, without the $10k price tag.
I’m currently focusing on refining the Reverse-Engineering engine and would love your feedback on the output quality or the logic behind the deconstruction.
The "ask an LLM to be creative and get back generic, unusable fluff" diagnosis is the right pain point — most of us treat ad copy as a one-shot prompt when it's actually a pipeline problem. As a solo indie shipping a tiny iOS memo app (a Captio replacement, very small scale), my hand-written 3-sentence App Store pitch consistently outperformed anything I got from generic LLM prompting — but only because that pitch named the exact failure mode my app solves. That hints at where your Context Scraping step might leak: if a landing page is vague (and a lot of indie ones are), the rest of the pipeline confidently amplifies the vagueness. Does your system flag "your USP is too generic to ad-test yet" as a preflight, or does it just generate plausible-looking ads on top of weak inputs?
If you're curious about the under the hood logic of how we actually achieve a 90-second analysis without losing depth, I've just published a full technical breakdown on our blog: https://ai-ad-generator.com/blog/analyze-winning-meta-tiktok-ads-in-90-seconds
This is a strong direction because the real value is not “generating ads.” It is turning creative research into a repeatable decision process.
That is also where I think the current name may cap the product.
AI Ad Generator is clear, but it makes the product sound like another generic prompt tool in a very crowded category. What you’re describing is closer to creative intelligence, benchmark deconstruction, and structured ad logic.
If the reverse-engineering engine becomes the core asset, a cleaner brand like Beryxa.com would probably age better than a descriptive AI tool name. It gives the product more room to become a serious creative strategy platform, not just an ad copy generator.
This is your final warning. I've noted your systematic link-farming across all my technical threads. If I see one more automated link dump, I'm escalating a formal ban request to the admins with all documented evidence.
Understood. I’ll stop commenting on your posts.
My intent was to give positioning feedback, not spam your threads.
I’ll remove myself from the conversation.
A couple of relevant links in the right context can be helpful, but plastering them across every single thread starts to feel like noise rather than value. We put a lot of genuine effort into crafting these posts to actually help the community. If you're creating your own content, you probably get where I'm coming from.
Anyway, let's just keep things clean moving forward. Best of luck with your project.
Stop spamming your links on every single one of my posts. This is a space for real engineering, not for your parasitic SEO farming. Reported for spam.