Last updated November 9, 2026. Written by the AdAdvisor media buying team: more than 8 years in media buying, over $60M in managed ad spend, and an ex-Meta engineer on the team.
TL;DR
Facebook ad competitor analysis is not collecting screenshots. It is reading three things from a public ad in Meta Ad Library: how long it has been live, what it promises, and what it asks the viewer to do. You then turn each pattern into a hypothesis and test it on your own account. Meta never shows a competitor's budget or ROAS, and only a few targeting fields for EU and UK ads, so every read is evidence to test.
Quick answer: what is competitor ad analysis?
Competitor ad analysis means using public ad data, mainly Meta Ad Library, to learn what another advertiser is putting into market. It shows which angles and offers a competitor keeps live, not whether they make money, so the useful output is a test plan rather than a copy.
What you can and cannot see in a competitor's ads
Meta Ad Library shows creative and selected transparency data; commercial performance stays private. Anyone can use it with or without a Facebook account, and what you see depends on the country and the type of ad (Meta Ad Library). The table reflects Meta's documentation as of November 2026.
| Data point | Where you can see it | What to keep in mind |
|---|---|---|
| Ads running right now | All countries | Ads appear within 24 hours of their first impression |
| Ads that have stopped | EU: one-year archive. UK: ads run after July 1, 2025, kept one year. Elsewhere: not for commercial ads | Political and issue ads keep a seven-year archive |
| Displayed start date and active status | Running ads everywhere; stopped ads only where archived | The date the advertiser set delivery to start, not proof the ad ran without a break |
| Creative, copy, CTA, destination link | All countries | The current landing page may differ from the one the ad first used |
| Platforms (Facebook, Instagram and other Meta products) | All countries | Shows where the ad appeared, not impressions or spend by platform |
| Selected age, gender and location targeting | EU and UK ads (plus political ads delivered to Brazil) | A few disclosed parameters, not the full audience setup, custom audiences or optimisation |
| Estimated impressions and reached-audience demographics | EU and UK ads; a total reach figure for EU ads only | Delivery estimates, not what the advertiser selected |
| Reported beneficiary and payer | EU ads only | As declared by the advertiser |
| Spend | Political and issue ads only, as ranges | Never shown for commercial ads, including in the EU and UK |
| Budget, bids, CTR, CPA, ROAS, conversions | Not available anywhere | Any tool claiming these is estimating |
| Custom audiences, lookalikes, customer lists | Not available anywhere | Private to the advertiser |
Sources: Meta Ad Library API, About the Meta Ad Library.
The EU and UK archives exist for different reasons. Meta ties the EU archive to the Digital Services Act (Meta Newsroom), while the UK archive covers ads run after July 1, 2025 (Meta Ad Library). If a competitor advertises in the EU or UK, selecting that country usually gives you far more to read, including ads they have switched off. Country selection is an easy step to miss in Meta Ad Library competitor research.
The Ad Library Report is a different product: downloadable political and issue ad spend by advertiser and region, so it will not help with commercial competitors (Meta Ad Library Report).
An empty result is not proof a competitor stopped advertising. Meta notes that a Page may simply have no ads running in the country you selected (Meta Help Centre).
"This ad has multiple versions" means the ad runs in more than one version, each with its own text or creative (Meta Ad Library API). The label does not show whether they formed a controlled A/B test or which one performed best.
Drowning in Meta Ads?
Put your campaign on autopilot with Nova.
Read moreThe three signals worth reading in a Facebook ad competitor analysis
We call this the Three-Signal Read: longevity, promise and ask. Together they are usually enough to form a testable hypothesis.
Longevity: how long since the displayed start date? It shows how long an advertiser has kept an ad in market, at least on and off. It is the most useful free proxy for performance in the Library, and the next section covers how far to trust it.
Promise: what does the ad claim? Read the first frame, headline and main copy line, and name the problem it solves and the proof it uses. A promise repeated across many fresh ads is likely a message the competitor is investing in.
Ask: what does the ad want the viewer to do? The CTA, offer and destination tell you the funnel step. "Shop now" with a bundle price, "Take the quiz" or a lead form each point to a different conversion path.
Why ad longevity is a useful free proxy, not proof of performance
An ad that stays live for months often signals continued investment, but it cannot prove performance or profit. Advertisers often switch off ads that lose money, so survival carries some signal. Ad Library data includes no revenue, conversion or profit outcomes, and one vendor teardown that tracked ad age notes it did not test profitability (Adlicio).
A long-running ad can have other explanations:
- A small retargeting or evergreen budget where it barely spends.
- A profitable offer carrying a creative nobody refreshes.
- An awareness objective rather than direct sales.
- Pauses or relaunches hidden behind one start date.
- Nobody has looked at it in months.
Read longevity relative to the advertiser, not in absolute days. If a brand rotates most ads every two to three weeks, one ad live for 90 days stands out far more than in an account that never changes anything. This relative-longevity check is our working heuristic, not a Meta rule.
Longevity plus repetition is stronger than longevity alone. A concept that stays live and keeps reappearing in new executions is likely getting investment, which is still an inference until your test confirms it.
How to turn a competitor read into a test plan
Every observation becomes one of three labels: Observed, Inferred or Unknown. Observed is what the Library shows, Inferred is your best explanation, and Unknown is what public data cannot tell you. Naming the Unknown keeps you from testing a guess.
Thresholds in this table are working heuristics, not Meta rules.
| What you observe | What it likely means (inferred) | Hypothesis for your account | Test to run |
|---|---|---|---|
| The same delivery promise ("arrives in 2 days") across five or more fresh ads | Speed is a message they are scaling | Reducing delivery uncertainty could lift purchase conversion | A delivery-confidence headline against your current control |
| One demo ad live for 90+ days while their other ads rotate monthly | The demo concept likely holds up for them | A problem-first demo opening could improve hook rate for your product | One original demo concept against your control |
| Their ask shifts from "Shop now" to a quiz or lead form | They may be testing a lower-friction first step | A lower-commitment ask could reduce cost per acquisition on cold traffic | Only if your funnel can follow up; quiz entry against direct purchase |
| Offers framed as bundles, never as percentage discounts | They are probably protecting margin or order value | Bundle framing could beat a discount on contribution margin | Bundle offer against your current discount |
| A concept they ran for months disappears | Unknown: fatigue, stock, season or budget | None | Watch and re-check in a month |
Turn the map into a plan in five steps:
- Capture the evidence. Record the ad link, Page, country, observation date, displayed start date, active status, creative, offer and destination.
- Label each line. Separate "three ads promise fast delivery" (Observed) from "fast delivery drives their ROAS" (Inferred).
- Group concepts, not variations. A new hook, proof or offer is a concept; a new headline on the same video is not.
- Pick one hypothesis per test. Write it as a sentence your own data can prove wrong.
- Build original executions. Rebuild the angle with your own product and claims, and set the metric, spend limit and decision rule before launch.
A copyable observation log, one row per ad in a spreadsheet:
Ad link:
Page:
Country:
Observed on:
Displayed start date:
Active status:
Promise:
Ask (CTA + offer):
Label (Observed / Inferred / Unknown):
Hypothesis:
Test status:This is where competitor analysis ends and testing begins. For sample sizes and scoring winners, use our guide to AI creative testing for Meta ads and its Test Ledger, or the step-by-step Facebook ad creative testing guide.
Where AI helps and where it just adds volume
In competitor ad research, AI is fast at sorting public ads into patterns and slow to admit what it does not know. Keep a human on every conclusion that turns into spend.
| Where AI helps | Where it tends to go wrong |
|---|---|
| Extracting hooks, offers, CTAs and formats from many ads | Misreading screenshots, or missing what happens in the video and audio |
| Clustering repeated concepts and near-duplicates | Counting variations of one ad as separate strategies |
| Drafting test briefs from the patterns | Inventing audience, spend or profit explanations the data cannot support |
| Summarising linked landing pages | Treating today's page as the one the ad used months ago |
Require every AI conclusion to carry the ad link, the observation date and an Observed, Inferred or Unknown label. A copyable prompt for ChatGPT or Claude:
Here are competitor ads I collected from
Meta Ad Library, with links and dates.
For each ad, record: displayed start date,
active status, core promise, proof type,
CTA and offer.
Do not assume the ad delivered continuously.
Group the ads into concepts, not variations.
Label every conclusion as Observed,
Inferred or Unknown.
Do not estimate spend, ROAS or targeting.
Then propose up to three hypotheses
I could test on my own account.What about Facebook ads spy tools?
Ad libraries and swipe tools such as Foreplay, BigSpy or Minea add search, saving and history across networks, and some add estimated metrics; creative analytics tools such as Motion mainly read your own account. Many comparisons are written by vendors, and a tool being available does not show its data was collected with Meta's permission (Meta Help Centre). Read any competitor spend or ROAS figure as an estimate.
What not to do
Do not create fake or clone accounts to see more ads. The Ad Library is public, and fake accounts break Meta's authentic identity rules (Meta Transparency Center). They can also put your real ad account at risk, as our guide to a Facebook ad account banned explains.
Do not scrape the Library in ways that break Meta's terms. Meta defines scraping as automated collection from interfaces built for people and separates authorised from unauthorised collection (Meta Help Centre). The official Ad Library API covers political and issue ads worldwide and ads of any type delivered to the UK or EU in the past year; for active commercial ads elsewhere, use the public Library.
Do not copy a competitor's ad as is. Take the customer insight and the structure, never the video, a creator's likeness, their reviews or their claims.
Do not draw conclusions from a single ad. Look for a pattern across several ads and weeks.
Do not read the number of active ads as spend. Fifty live ads can share one small budget.
Where Nova fits
Nova, AdAdvisor's AI media buyer, reads the ads your competitors are running alongside your Meta account, pixel, Shopify store and site, then proposes the resulting test with its reasoning. In Copilot mode the test waits in your approval queue before it spends anything, so you still make the decision.
Frequently asked questions
Summary
Facebook ad competitor analysis is a reading exercise with a testing output, and it is the part of Facebook ad competitive intelligence you can do for free. Meta Ad Library shows what competitors are running, since when and, in the EU and UK, who it targeted and roughly who saw it, but never whether it made money. The Three-Signal Read of longevity, promise and ask, with every conclusion labelled Observed, Inferred or Unknown, turns that public evidence into hypotheses your own account can confirm or reject.
Sources
- Meta Ad Library
- Meta Ad Library API
- Meta Ad Library Report
- Meta Help Centre: See a Facebook Page's ads
- Meta Business Help Centre: About the Meta Ad Library
- Meta Help Centre: Data scraping and what can you do to protect your information on Facebook
- Meta Transparency Center: Authentic Identity Representation
- Meta Newsroom: New Features and Additional Transparency Measures as the Digital Services Act Comes Into Effect (August 2023)
- Adlicio: Competitor Analysis Examples: 3 Real Teardowns With Data (vendor blog, used as an example)
Related reading
Creative & Content
AI Creative Testing for Meta Ads: How to Build the System (2026)
A creative testing system for Meta ads is a loop, not a launch: one treatment per test, predefined evidence requirements, a rubric fixed before launch, and a ledger that feeds the next batch. AI handles the volume and the reading. You approve.
Read more
Creative & Content
AI Creative Analysis for Meta Ads: Find Your Winners
A product-independent method for analyzing Meta ad creative with AI: a seven-dimension creative taxonomy, a format-aware scoring rubric you run against your own account baseline, winning patterns, failure modes, and the loop from analysis into new creative.
Read more
Creative & Content
Dropshipping Facebook Ad Examples (+ Video Ads That Work)
What a winning dropshipping Facebook ad looks like in 2026: the 6-beat anatomy, the Proof Burden rule, six annotated video and static examples, winners vs flops, Meta policy limits, and a production loop for constant new creative.
Read more
