I roasted 96 landing pages with AI this week. Here is what I'm changing.
Hey IH,
I've been building RoastMyLanding in public this week. I trained an AI to find "Trust Leaks" and UX friction points.
The Stats:
Total Audits: 96
Conversion Rate: Surprisingly high for a "brutal" tool.
Biggest Lesson: Founders are too close to their product. 80% of pages I audited had headlines that did not explain the benefit in the first 5 seconds.
The Pivot:
Tomorrow, I'm moving to a "Value-First" model ($5 for a full unlock). But today, I want to hit my 100-audit milestone.
I have 4 free spots left for the IH community before the paywall logic goes live tomorrow. If you want a cynical AI to tell you why your page isn't converting, try it yourself instantly here: https://roastmylanding.vercel.app/
Trending on Indie Hackers
Launched Lemonvite on Product Hunt today: $5 per event, no ads, no subscription.
96 landing pages is a great dataset. The question I'd ask before changing anything ... what percent of them passed the 5-second test?
I run something I call the Cover-the-Logo Test with every founder I work with. Cover your logo on your homepage. Show it to a stranger. Five seconds. Ask them: 'Who is this for and what problem does it solve?'
If they hesitate ... the problem isn't the design. It isn't the headline length. It isn't even AI writing your copy.
It's that you haven't yet answered the upstream question. Who specifically is this for? What problem are we solving that's untenable? What's our point of view that nobody else is saying?
Generic landing pages are a symptom. The disease is unresolved internal conversations about who you actually are.
AI didn't make landing pages worse. AI made the cost of bad copy zero. So now there's just MORE of it. The fix is the same fix it's always been ... figure out the truth first, then put it on the page.
What percent of the 96 had a real point of view vs. 'AI-powered platform for modern teams'?
The 80% headline failure rate is a great data point, and it matches what I've seen too. The root cause I keep running into: founders write headlines that describe their product, not the outcome for the visitor.
One framework that helps: "For [who], get [outcome] without [pain]." If your headline can't fill in all three blanks, it's probably too vague.
A quick litmus test: cover the logo and product name, then show the page to someone who's never seen it. Can they tell you in 5 seconds who it's for and what problem it solves? If the answer is "it's some kind of AI tool" — the headline is doing product description, not value communication.
The other thing I'd push back on is trust element overload. Badges, partner logos, "as seen in" strips, star ratings — founders stack these because each one individually tests well, but together they create visual noise that actually undermines trust. One real element (a genuine customer result, a before/after screenshot, or a specific workflow demo) beats five generic badges.
Curious whether the RoastMyLanding audits flag this "headline describes product vs. outcome" distinction specifically, or if it shows up under a broader "trust leak" category?
Update: We just officially launched Version 2.0 on Product Hunt for #VercelDay! 🌶️
I've integrated a lot of the feedback from this thread into the new engine (especially the deeper conversion logic and faster UI).
If this thread helped you, I’d love your support on the PH page today:
https://www.producthunt.com/products/roastmylanding-2?launch=roastmylanding-2-0
See you there! 🚀
Just hit our first US sale! 🇺🇸
It’s amazing to see the journey from our first Norway sale last week to the US today. Version 2.0 is officially solving problems across borders.
If you haven't checked out the new conversion logic yet, grab a roast here: https://roastmylanding.vercel.app/
96 audits in a week is genuinely impressive dataset. The 80% finding lines up with what I've seen too — but I'd add a second failure mode that's almost as common:
The headline explains what the product does, but not who it's for. So even when the benefit is technically stated in 5 seconds, visitors who aren't sure "is this for me?" still bounce. E.g. "AI-powered analytics for teams" vs "AI analytics for product managers at Series A SaaS" — same product, very different trust signal.
Quick question — when RoastMyLanding flags a "trust leak," does it rank them by severity? Because in my experience the fixes people actually execute on are the top 1-2, not the full list. Curious whether you surface that prioritization or leave it to the founder.
Also: would love to submit my own landing page for a roast — is the form still open for the 4 free audits? 👀
Great observation on the 'Who is it for' vs. 'What is it' distinction—I’m actually seeing that in about 40% of the failed roasts.
To your question: Yes! Based on exactly what you said, the new version I just pushed today prioritizes the Top 2 'Critical Fixes' for free. I realized giving people 10 things at once was overwhelming.
The free spots for the 'milestone' are gone, but since you gave such solid feedback, I’ve kept the Score + Top 2 Fixes completely free (no email) so you can still get the core value. Check it out here: https://roastmylanding.vercel.app/
Nice, prioritizing Top 2 fixes is a smart call — that alone probably increases the action rate significantly. The "Who is it for" pattern is interesting at 40%, might be worth making that a standalone check category since it's so common yet easy to miss.
Just tried it — clean experience. I'll submit my own landing page through there and see how it scores. Happy to share the result back here if it surfaces anything useful.
Exactly! The 'Who is it for' gap is a silent killer. Let me know what score you get! Also, we're live on Product Hunt today—would love for you to share your score in the PH comments!
https://www.producthunt.com/products/roastmylanding-2?launch=roastmylanding-2-0
Tried it just now — submitted my landing page but the score didn't
come back (page just sat there after submission, no error either).
Might be worth checking on launch day — would be a shame to lose
PH traffic to a silent failure.
Either way, dropped a comment on your PH launch about the Top 2 Fixes
decision, which I genuinely think is the strongest part of the
positioning. Good luck today! 🚀
Hey opps! Thanks for the patience. I dug into the logs and found what happened. 🛠️
It looks like the first attempt hit a scraping timeout (which caused that silent hang), but the second one actually went through! Since the scraper finally 'saw' the content, the AI was able to generate a proper audit for you.
You can see your free grade and the 'Top 2 Fixes' here:
https://roastmylanding.vercel.app/roast/26WqB6V2
The 96-audit dataset is your real asset. Instead of teasing it, publish the top 5 patterns as a standalone post. That kind of data-backed content would drive more signups to the tool than the “4 free spots left” urgency framing. Show the insights, and the product sells itself.
Wonsik, this is a masterclass in content marketing. I actually took your advice to heart—I’m currently bucketting the data from the first 100 roasts to release a 'State of Landing Pages 2026' report.
The 'Top 5 Patterns' you mentioned are exactly what powered the new $5 Conversion Blueprint I launched today. It’s less 'AI guessing' and more 'Data-backed fixing.'
Appreciate the nudge to move from 'Urgency' to 'Insight'—it’s already making the product feel much more premium. 🚀
This is actually a great insight — most pages don’t fail because the product is bad, but because something feels unclear or slightly off in the first few seconds.
Feels like people decide “trust vs doubt” almost instantly, and everything after that just confirms it.
Out of curiosity — did you notice if those trust leaks were more about messaging clarity, or things like naming / positioning that shape first impression before they even read properly?
Exactly, Aryan. The 'Trust vs. Doubt' moment happens in the first 2 seconds.
From the 100 roasts, I found it's usually Messaging Clarity that kills the vibe. If the user has to 'squint' to understand the value, they doubt the product's quality.
I just updated the engine today to focus specifically on 'Cognitive Load.' The new $5 blueprint now breaks down exactly where the messaging is 'blurry' vs. 'sharp.'
Thanks for the insight—it definitely shaped the v2.0 logic! 🌶️
That’s a strong direction — reducing cognitive load is probably the highest leverage fix.
One pattern I kept noticing though is that some pages were already “losing” before the copy even had a chance.
Not because the headline was unclear, but because the first impression felt slightly off — name, positioning, even how the product presents itself at a glance.
So even when the messaging was technically clear, it didn’t fully land because the initial frame didn’t feel solid enough to trust.
It’s subtle, but it creates that instant “this feels polished vs this feels experimental” difference in the first second.
Curious if any of your audits showed that kind of gap — where the copy wasn’t the real bottleneck?
"That 'experimental vs polished' frame is everything. It's often the 'Visual Hierarchy' that fails before the copy even talks. I’m actually discussing this exact gap on our Product Hunt launch today. Would love your expert take in the comments there:
https://www.producthunt.com/products/roastmylanding-2?launch=roastmylanding-2-0
That makes sense — visual hierarchy is usually where the “first second” gets decided.
I’ll take a look on PH.
One thing I’ve seen there as well — even with clean hierarchy, if the underlying positioning feels slightly generic, people still hesitate.
So it ends up being:
clarity + hierarchy + how “distinct” it feels at first glance.
Curious how you’re thinking about that layer alongside the audits.