I built an AI that roasts landing pages for CRO — before I build more, tear the feature set apart

The competitor comparison feature is the hook — not the roast. "See how your headline compares to 3 competitors" is a concrete, slightly threatening question that makes agency clients immediately curious and slightly defensive. The roast is fun but 'your score is 62/100' doesn't create urgency. 'Your main competitor is scoring 78 and their CTA is outperforming yours' does.

On the agency trust question: the client-ready PDF is critical but it needs to look like the agency did the thinking, not the tool. If it reads like 'Swiplio says...' you'll have problem getting sign-off. If it reads like an analysis the agency ran with their own branding on it, it flies. That's the product decision that determines whether agencies buy it.

The before/after tracking is the Pro feature I'd build first. It creates an ongoing reason to pay — not just a one-time audit tool.

What's the main agency pain point you found in your early conversations — time to produce the report or confidence in the recommendations?

For an AI landing page audit tool, I’d be careful not to make the “roast” part the core value. It’s fun for attention, but the real value is whether the feedback helps someone make a better decision.

The strongest version would probably separate feedback into clear layers: positioning, clarity, trust signals, offer, CTA, friction, and proof. Then show what should be changed and why it matters for conversion.

Also, I’d avoid only giving generic CRO advice. The tool becomes much more useful if it understands the type of page: SaaS homepage, waitlist, paid ads landing page, ecommerce product page, or agency service page. Each one needs a different kind of critique.

Building EarningsScores (AI that scores earnings reports), so I think about the trust problem constantly.

On Q3 — would I trust an AI score in front of a paying client: not without showing my work. The score alone isn't the thing. What builds trust is seeing exactly which line of copy caused the problem and why. When I show users a score of 62/100, they push back until they see the specific sentences that dragged it down.

The 17-point rubric is the right instinct. I'd make the rubric visible in the free tier, not just the score. Let people see what they're being graded against before they pay. It answers the "why should I trust this" question before they even ask it.

On the Pro features: competitor comparison is the painkiller. The rest are vitamins. Agencies don't need better copy suggestions — they need to show clients "here's why your page loses to this competitor on these 3 dimensions." That's a deck that gets deals closed.

Nowadays, AI makes our daily lives much easier; now we just have to know how to choose the tools that solve the problems that afflict us. If it's free, I'll test my landing page out of curiosity and see what's wrong and what they advise to improve it 😅 Congratulations and much success on your journey from here on out...

Niche idea, I actually was looking for a simple audit before looking into professional solutions.

Why should I trust your rubric?
You say AI's responses are useless, but this is just AI with an extra long prompt, right? Could I arrive at a similar rubric with a few prompts of my own? Could I replace this with a few hours and some "screw this, I'll do it myself" energy? Did I already do something like that before I see your software?

Love this, CRO linter vibes. Roasts make feedback fun and actionable. If you expand, add severity scoring, A/B-ready rewrites, and easy exports. Would love to beta test!

What tools do you usually use for research when working on AI products?

The client-ready PDF/report is the killer feature. Agencies pay for saving time and delivering professional audits, not just AI scores. Competitor comparison is the next most valuable feature. The biggest risk is generic AI advice—if the insights aren't specific and actionable, users won't trust it. Focus on helping agencies save hours, not just generating scores.

I like the idea of getting brutally honest feedback before investing more time building.
I've been thinking a lot about validation lately. Too many founders build first and learn later. Curious — what's the one feature users seem to care about most so far?

I’d separate the “fun demo” from the actual paid job here.

The free roast feels like the right hook because it gives people an immediate reason to paste a URL and see if the tool is useful. But I don’t think the roast itself is the thing agencies would pay for. The paid value is probably the client-ready report plus the ability to edit it before sending. That turns the tool from “AI gave me feedback” into “I can save an hour and send something credible to a client.”

Competitor comparison also sounds strong, but only if it stays very practical: show the specific headline/CTA/trust gaps, not a vague scorecard. I’d be careful with the 17-point score unless you can explain why each point matters. A simple confidence note or evidence trail would make the AI score easier to trust.

If I were testing this, I’d launch with free roast -> editable report -> PDF/share link, then add monitoring later once users prove they are using reports repeatedly.

Roast as the free entry point is smart.
Low friction, immediate value.

I'm Minchul, 57, former construction manager from Korea.
Built Slash it — an Email Decision OS.
Also validating right now. Good luck!

I think “AI roast” is the hook, not the paid product.

Agencies probably won’t pay for another opinion on a landing page. They’ll pay for something they can show a client to justify changes and move the work forward.

So I’d be careful adding Pro features before separating attention from payment intent.

Happy to put the tighter feature-priority and paid-hook angle in writing if useful.

please comment