Introducing AdRivela: ad teardowns that don't sound like LinkedIn

We tried every ad-analysis tool we could find in 2025. They all produced the same output:

"This ad has a strong hook and leverages social proof effectively to drive conversions. The visual is engaging and the copy is compelling."

That sentence could apply to almost any winning ad. Which means it applied to none of them usefully. It's a horoscope.

We built AdRivela around a different bet: that operators don't want generic horoscopes. They want the diagnosis — what mechanism is doing the work — not the description of what's on screen. The difference shapes everything about how the product is designed.

The diagnostic move

A junior strategist looking at a winning ad says "the hook is strong." A senior strategist looks at the same ad and says "this ad is selling to a Solution-Aware buyer using a Proof-First hook — that's why hold rate is fine but CTR underperforms on Problem-Aware traffic."

The first sentence is unfalsifiable mush. The second is a claim a reader could disagree with, which is what makes it useful.

The whole product is built around forcing diagnostic sentences and refusing descriptive ones. Our LLM prompt has a banned-words list. The model is not allowed to write "engaging," "compelling," "leverages," "effectively," "robust." If it reaches for one of those words, it has to rewrite. The output is harder to generate and dramatically more useful to read.

What we ship

You paste a Meta Ad Library or TikTok Ad Library URL. We do the rest. Within 1–2 minutes you get back a structured report:

• A score — A through F, with two sentences of reasoning. One sentence names the one mechanism that lifted the score. One names the one that capped it. Both falsifiable.
• A summary — 60-100 words opening with the one weird thing the ad is doing. Never a metadata recap.
• Four sections — Hook, Script, Visual, CTA. Each with an archetype/framework label, What works + What doesn't work bullets, and reference chips that link to glossary entries.
• Steal / Adapt / Skip — three buckets of action items the operator can ship this week. Each item: title + 2-4 sentence body. Specific enough to brief an editor.
• References — a list of frameworks the analysis used, with sources. Every reference links to a primer page.

Every "what works" / "doesn't work" bullet ends with the metric it most likely affects, in parens. "(hook rate)" "(3-second hold)" "(CTR)" "(CVR)" "(AOV)". This isn't decoration — it forces the bullet to be tied to a number that moves, which forces the writer to think mechanistically.

The frameworks we lean on

A senior teardown isn't a pile of vibes. It's specific frameworks applied to specific evidence. The product is grounded in a curated set:

• Schwartz's 5 Stages of Awareness — the diagnostic axis ("who is this for")
• Hook / Body / CTA — the execution axis ("how does it work")
• Five hook archetypes — Pattern Interrupt, Bold Statement, Question, Direct Address, Proof-First
• Hormozi's Value Equation — for scoring the offer
• Cialdini's 7 levers — as a checklist for the body
• Hott's ugly-native vs. produced axis — for visual diagnosis
• Steal the strategy, not the ad — for the action-item layer

Every framework links to a glossary primer explaining what it is, when it applies, when it backfires, and DTC examples. The point isn't to make the reader memorize them. It's to let the reader trust the report — because every claim is grounded in a named, citable framework, not vibes.

What we explicitly don't do

Three things every other tool seems to do that we deliberately avoid:

We don't predict ROAS. Anyone claiming to predict ROAS from a 30-second ad alone is selling you snake oil. Ad performance is a function of creative + audience + offer + competitive context + seasonality + a dozen other things. The most we can do is diagnose creative mechanisms — and tell you which ones travel and which don't.

We don't write hedged copy. Most LLM tools are tuned to be inoffensive. They never commit to a take because committed takes can be wrong. We tune the opposite direction — every section opens with a falsifiable claim. The model can be wrong, and you can disagree with it, and that's what makes it useful.

We don't try to be everything. No image-only ads in v1. No multi-pass parallel analysis. No "AI brief generator" that fabricates a strategy doc. The product does one thing: pasted URL → diagnostic report. That's it. We earn the right to do more after we're sure this one thing is good.

Pricing

One free analysis when you sign up. After that, refill packs — three for $9 or ten for $25. No subscription. Credits don't expire. The reports themselves are cached by URL: if someone else has already analyzed the same ad, you get their report instantly and they share the work.

We're not trying to charge per-seat or build a moat by complexity. We're trying to make the best possible single-ad teardown and let the value sell itself.

Who this is for

You'll get the most out of AdRivela if you're:

• A DTC operator running Meta + TikTok ads at $5k-$50k/mo
• A creative strategist building swipe files
• An agency lead briefing editors
• A founder doing competitive research before launching a new product

You'll get less value if you're trying to predict winners before launching them, looking for a tool that writes your ads for you, or running B2B SaaS ads — those play by different rules and our frameworks are tuned to consumer DTC.

What's next

We're shipping the product as a free-tier-only launch today. One free analysis on signup, no payment required. We'll add refills (the $9 / $25 packs) once we know the free tier is delivering value at scale.

A few things we know we want to ship in the next 60 days:

• A "compare two ads" view (we have the data; the UI isn't built yet)
• Better TikTok support (the parser is best-effort today; hydration-mode scraping is on the roadmap)
• Briefs against your own product URL (paste a competitor's winning ad + your store; get a teardown specifically translated to your category)
• A Slack/Notion push for swipe-file workflows

If you build creative for DTC and want to try the first one: paste an ad URL and tell us what we got wrong. The fastest way to make the product sharper is to hear which diagnostic claims you'd push back on. The roast cycle is how we got here, and it's how we get better.

— The AdRivela team