Ad analysis for performance marketers

The problem: you can see what's working, but not why

You're running three creative concepts, two are flattening out, and the third is scaling. You know which video to pour budget into. You don't know which element is doing the work.

That gap is expensive. You can pull the thumbstop ratio and hook rate from Ads Manager and confirm the first three seconds are winning — but is it the visual pattern interrupt, the opening line, the creator's delivery, or the product in the first frame? Guessing costs you a week of production on a creative that answers the wrong question.

The same problem hits in reverse when a competitor's ad has been running on Meta for 90 days. Something in that ad is working. You see the surface — the visual language, the UGC format, the offer callout in the caption. The mechanism is invisible until you run a structured teardown on it.

Most performance teams fix this with manual teardown sessions: someone screen-records a competitor's ad, pastes it into a doc, and writes up observations. The output is notes. Notes don't score. Notes don't slot into a testing queue. Notes from last Tuesday are stale by Thursday when the brief is due.

How AdRevila fits your workflow

The v5 pipeline is one step: paste a Meta Ad Library or TikTok Creative Center URL. From there, a single Gemini call watches the video and extracts the full structure — no human transcription, no back-and-forth, no prompting required.

What comes back is a structured report in under 60 seconds. Every report covers the same four-layer diagnostic that runs in the same order: audience stage (Schwartz), hook archetype, persuasion body, offer math. Consistent vocabulary means any report from any ad on any date speaks the same language as the one before it.

The report is available as JSON if it's going into a dashboard, Markdown if it's going into Notion or Linear, PDF if it's going to a client, and a public link if it's going into Slack without a login wall.

First analysis is free. After that, credit packs — no subscription required if your teardown volume is episodic.

What you get back

Every report scores four distinct layers against named frameworks:

Hook archetype and score — one of five archetypes (Pattern Interrupt, Bold Statement, Question, Direct Address, Proof-First), plus an honest rating of whether the hook earns the next five seconds given the audience stage it's targeting. A hook that self-selects hard is often the right call; a hook that's just loud is not.

Script and visual diagnosis — the persuasion framework running in the body (PAS, AIDA, BAB, or a narrative hybrid), the Cialdini levers firing and the specific moment each triggers, and whether the visual surface is consistent with the copy's job or working against it.

Offer math — the Hormozi value equation scored per lever: Dream Outcome resolution, Perceived Likelihood proof, Time Delay, Effort and Sacrifice. This is the layer most operators hand-wave past. A high-resolution teardown shows you exactly which lever is underwater before you decide whether to restructure the offer or just the CTA.

A/B hypotheses — three concrete test variants ranked by expected lift, written in brief-ready language. Not "try a different hook" — "swap the Pattern Interrupt cold open for a Direct Address qualifier that names your buyer segment explicitly, then hold the script flat."

The grade at the end (A through F, per the AdRevila rubric) names what to copy, what to adapt, and what to leave in the source brand's lane.

From teardown to testing queue in one pass

The creative diagnostic ladder maps the funnel layer each report metric targets: Hook → Hold → CTR → CVR. When a report flags the weakest layer, it's telling you exactly where the creative is leaking — and therefore which fix belongs in the next sprint.

Hook leaking? The test is an archetype swap. Hold dropping? The script structure or pacing is losing the viewer before the offer lands — usually a PAS agitate problem or a body that's describing instead of persuading. CTR flat despite strong hold? The CTA commitment level is mismatched to the traffic temperature. CVR lagging everything? The offer math is fair but uncompelling, or the landing page breaks the ad's promise on the first viewport.

Running a competitor teardown with this ladder in mind means you're not just noting that their ad "feels confident." You're reading which layer their mechanism is winning at — and building a brief that extracts that mechanism without copying the surface. That's the difference between an A/B test that compounds and one that just generates data.

Read the five-minute Meta walkthrough in How to read a winning Meta ad in five minutes if you want the manual version of the same flow before you let the analyzer run it for you.