Skip to content
TargetDemon

Competitor ad teardowns for Meta

What competitors run. Where their funnels leak.

Every live competitor ad, read and tagged with its confidence. Every landing page, checked for where the promise breaks.

Start with StarterSee a sample report
One ad, as the product sees itsample data
Thistlemere146138065699608
Thistlemere

Tired of "gentle" that isn't?

7 ingredients. Zero fragrance.

A woman in a plain bathroom applies a white cream to her cheek in close-up, then the shot cuts to the product tube standing on a shelf, then back to her smiling lightly at the mirror.

video · sample, drawn from the caption

Finally, gentle that actually means gentle

Tired of "gentle" products that still leave you red and stinging? Thistlemere Barrier Repair Cream was built for skin that reacts to everything else. Fragrance-free, 7 ingredients, dermatologist-developed. Over 40,000 five-star reviews from people who thought their skin was just broken. Get 20% off your first order today.

Shop Nowlanding page matches
how it opens
problem callout84%
main angle
safety or risk83%
format
product hero82%
offer type
percent off96%
days running
373
variants
34
status
live

Days running is a survival proxy, measured against the longest-running ad in this niche. It is not a performance number.

Across the whole niche

label reader lina × quality or craft

brands
6
concepts
10
ads
14

Another ad, and the page it lands on

product mismatch

Ad promotes a $9 Barrier Repair Duo bundle; the landing page sells only the single cream, with no bundle, duo pricing or second product offered anywhere on the page.

https://www.thistlemere.example/lp/thistlemere-barrier-cream-pdp?utm_source=facebook&utm_medium=paid_social&utm_campaign=barrier_cream_hero&utm_content=th-03&fbclid=EXAMPLE5VH1X3

Fictional brands from our demo niche. Confidences are sample values: no accuracy has been measured yet, and the scorecard page says so plainly.

Not a mock-up. One real pull, 2026-09-21:

live ads pulled
112live ads pulled
real, named brands
36real, named brands
answers per ad, on average
267answers per ad, on averageof up to 303 questions at the time
to tag all 112 ads
126sto tag all 112 ads$0.23 in model costs, total

Starter is $49 a month, and the first payment is refundable for 14 days. Or try the $9 seven-day pass first.

330 narrow questions, asked of every ad.

Hook, format, offer, awareness stage, angle, claim strength. One snap judgment each, with written criteria, and its own confidence.

How it works

Three steps. Nothing to install.

  1. Start with Starter

    $49 a month, or the $9 pass for a first look. Either way, the first payment is refundable for 14 days.

  2. Add the brands you want watched

    Yours and your competitors. Every live ad gets read: video, copy and the page it links to, checked against written criteria.

  3. Get the teardown

    Every answer carries its own confidence. Unclear stays unclear instead of being rounded into a guess.

Five outputs

A teardown, a leak list, a traction map, an audience read, and a test plan.

Each one is built from what the public ad libraries actually publish. Where a number does not exist, the product says so instead of estimating it.

Creative teardown

It watches the video, transcript and first three seconds included, and reads the copy, then answers up to 330 narrow questions per ad. A low confidence is shown as unclear instead of guessed.

how it opens
problem callout84%
main angle
safety or risk83%
format
product hero82%
offer type
percent off96%
persona match
lifelong reactive86%

Landing-page leak report

Same product? Promise continued? Offer repeated? Read with a real browser when the plain fetch is blocked. The one thing a brand can fix this week without shooting anything new.

9leaks across 70 ads in the sample niche

Traction map

What the advertiser keeps funding: days running, variant count, concept-cluster size. Ranked once a brand has enough ads tracked.

days running
373
variants
34
status
live

Days running is a survival proxy, measured against the longest-running ad in this niche. It is not a performance number.

Audience, both ways

For ads delivered in the EU or UK, the targeting set and who it actually reached, straight from Meta’s official API. Everywhere else, an estimate, labelled so.

18of 70 sample ads carry real EU targeting

Test plan and briefs

Ranked test cells with the evidence behind each one. The classifier never writes copy, and the writer never invents a number.

The rules it writes under

Numbers above come from the demo niche shipped with the product: 70 fictional ads across 6 fictional brands. Sample data, not a customer.

Most ad tools stop at the ad. The money leaks after the click.

Every ad is checked against the page it actually lands on: same product, promise continued, offer repeated, page type. Until a check has been measured against hand-labelled pairs, its flag is marked unverified.

The ad

6 weeks, one routine, real difference

My skin didn't just get drier in my late 40s, it got angrier. Redness, stinging, the works. Six weeks on the Barrier Repair Duo (cream + cleanser) and the difference speaks for itself. Bundle the two and save $9 this week only. Individual results vary.

Get Offercarousel12 variants

The page it lands on

product mismatch

Ad promotes a $9 Barrier Repair Duo bundle; the landing page sells only the single cream, with no bundle, duo pricing or second product offered anywhere on the page.

https://www.thistlemere.example/lp/thistlemere-barrier-cream-pdp?utm_source=facebook&utm_medium=paid_social&utm_campaign=barrier_cream_hero&utm_content=th-03&fbclid=EXAMPLE5VH1X3

Sample data. 9 of the 70 ads in the demo niche carry a planted mismatch, which is what makes it a useful test of the check.

37/ 46

Real data, 2026-09-21

Of 112 live ads from 36 named brands, we could confidently score the ad-to-page match on 46. 37 of those broke the ad’s promise: below 2 of 3 on message match.

One brand’s own sweepstakes promo: 5 different ads, all pointing at the same link. That link now reads “404, Page Not Found.”

What most ad‑swipe tools skip.

Three things this product does that the rest of the category, by and large, does not.

  • The landing page is part of the check

    Seven checks between the ad and the page it links to, and the exact line that breaks named out loud. Most tools in this category stop at the ad itself.

  • Real audience data, not just a guess

    For ads delivered in the EU or UK: the advertiser's own targeting and who it actually reached, read straight from Meta's API. Kept apart from our own estimate, never blended in.

  • Confidence on every tag, and a way to correct it

    A low-confidence answer is shown as unclear, never rounded up. No spend, no CTR, no ROAS, no “winner”, anywhere on this site or in a report.

Mix engine

Which persona and angle combinations the niche is actually betting on.

Counted in distinct brand-and-concept pairs, never in ad ids, so one brand running forty variants of one idea counts once. Cells below two brands and two concepts get no number at all.

PersonaAngleBrandsConceptsAdsConcentration
label reader linaquality or craft6101420% one brand
lifelong reactivesafety or risk681425% one brand
minimalist mayasave time or effort471043% one brand
postpartum priyasafety or risk45740% one brand
midlife maralow evidenceperformance or results23667% one brand
midlife maralow evidencehealth or wellbeing22350% one brand

Sample data from DTC skincare for sensitive and reactive skin, United States, as of 2026-10-17. Descriptive counts only. Before the product says one combination holds up better than another it has to pass a support floor, shrinkage against the niche’s own prior, a brand-level bootstrap interval and a multiple-comparison correction. That ladder is on the methodology page.

The honest part, which is also the whole pitch.

No public ad library exposes spend or performance for commercial ads. Tools that imply otherwise are guessing. This one draws the line on the page and prints it in every report.

Observed or classified

  • What the ad says and shows, read from the creative itself
  • How long it has run, and how many variants sit behind it
  • Whether the landing page keeps the ad's promise
  • Targeting set and audience reached, for ads delivered in the EU or UK

Never claimed

  • Spend, CTR, CPA or ROAS for any competitor, on any platform
  • Share of voice or market share: Meta inventory has no denominator
  • Placement (Feed, Reels, Stories): the libraries publish platform, not placement
  • That a long-running ad is a winner. It is an ad someone keeps paying for
What is knowable, question by question

Pricing

Three plans, a $9 seven‑day pass, and no trial that converts itself.

The pass ends unless you pick a plan. Unused credits roll over for a month, re-reading analysed ads is free, and every plan's first payment is refundable for 14 days.

StarterStart here

$49a month

8 brands, 1,000 credits, mix engine on one niche

Growth

$199a month

30 brands, 4,000 credits, audience module and alerts

Agency

$549a month

100 brands, 10 client workspaces, white-label reports

See the full price bookAnnual is about 17% less.

An honest FAQ.

The words this site uses, explained plainly.

What does confidence mean?
A probability the model attaches to its own answer. High confidence prints as a plain tag. Below the threshold we show unclear instead of rounding up to a guess.
What does unclear mean?
The model did not clear the bar for that question, on that ad. It is a real answer, not a placeholder and not an error: it means we could not tell.
What counts as traction?
What an advertiser keeps paying for: days running, how many variants, how big the idea's cluster is. A lead worth investigating, never a “winner”. We do not use that word.
What is a landing-page match, and what is a leak?
Whether the page an ad links to keeps the ad's promise: same product, same offer, same claim. Scored on 7 checks. Below 2 of 3, we flag it as a likely leak.
What is a credit?
One credit analyses one ad, once, for your organisation. Re-tagging after we change a question, and reading ads you already analysed, cost nothing.
How fresh is the data?
Brands on a plan's priority slots are checked daily. The rest are checked weekly.
Do you publish an accuracy percentage?
Not yet, on purpose. There is no hand-labelled gold set to measure against, so we will not print a number we cannot defend. Every tag shows its own confidence instead.
Is there a guarantee?
Not for you? Tell us within 14 days of your first payment on a plan and we refund that payment in full. Credits from it that you have not spent are removed with the refund; credits you already spent on an analysed ad stay spent. Email [email protected] to ask, no form to fill.
Read the full methodology

Read the methodology before you read the marketing.

If the limits do not convince you, nothing after them should.

Start with StarterMethodology and limits