OpenAdLibraryOpenAdLibrary
Competitor Ad Research

How to Do Competitor Ad Analysis: A Step-by-Step Framework (2026)

A practitioner's four-stage framework for turning a competitor's captured native ads into dated observations, grouped angles and a backlog of test hypotheses for next week, with the unknowns recorded.

Diagram of a competitor ad analysis framework showing watchlist, live native ad captures, creative scoring rubric, and spend signals from longevity and volume

Most "competitor ad analysis" is just opening a rival's site, glancing at whatever ad happens to load, and saving a screenshot to a folder nobody reopens. That is reconnaissance. It is not analysis. Analysis means turning a stream of live ads into decisions: which angle to copy, which offer to undercut, where to put your next test dollar.

This framework has four stages: build a watchlist, collect dated observations, group and describe creatives, and prioritize hypotheses for your own tests. The captured examples illustrate the method; they do not provide access to competitors' spend or conversion results.

For scale, here's what one slice of the open web looks like right now: we've captured 589,036 creatives from 25,933 advertisers across 42 networks (OpenAdLibrary index, June 2026). Taboola alone accounts for 157,727 of those creatives, Outbrain for 84,252. That is the haystack. The framework below is how you pull signal out of it.

What competitor ad analysis actually delivers#

Competitor ad analysis studies messages, offers, formats and observed placements to understand visible approaches. Its output is a documented set of observations and test hypotheses, with uncertainty recorded wherever spend, targeting or performance is unavailable.

A useful analysis connects each proposed test to a specific observed pattern and explains what remains unknown. It does not need to claim that a competitor's angle outperforms others or that their budget is known.

Stage 1: Build a watchlist worth watching#

Everything downstream depends on monitoring the right set of advertisers. Too broad and you drown in noise. Too narrow and you miss the rival quietly stealing your best customers.

Build three tiers:

Include direct competitors, relevant category examples and advertisers using comparable acquisition paths. Record why each belongs on the list. Do not assume a large creative set identifies the largest spender or that an affiliate runs only profitable campaigns.

An adjacent advertiser may reveal a useful message or funnel to inspect. Its visible activity does not show which tests failed or whether the remaining ads are profitable.

For each entry, record the real advertiser entity, the landing domains they use, and the networks you've seen them on. The "real advertiser" part is harder than it sounds in native. Ads route through tracking redirects, and the brand displayed on the creative often isn't the entity paying for it. A platform that resolves the actual advertiser behind each placement saves you hours of manual redirect-chasing. If watchlist construction is where you want to go deep, the watchlist-to-action workflow covers tiering and maintenance in more detail.

Stage 2: Pull live ads, the evidence layer#

A watchlist is useless without a steady feed of what those advertisers are actually running. This is where most DIY processes break. Manually refreshing publisher pages to catch a competitor's ad is slow, geo-dependent, and misses everything served outside your own browsing session.

You need a source that captures live, public native ads continuously, across publishers and geographies, and stores the real creative, not a thumbnail. The non-negotiables:

  • Live capture, not a stale archive. A library that last refreshed weeks ago tells you what ran, not what's running. Recency is the whole point.
  • Full-quality creative. You're going to study the image, the headline framing, the thumbnail composition. A blurry crop is evidence you can't read.
  • The supply chain, classified. Knowing an ad ran on Taboola vs. Outbrain (now operating as Teads after the two merged in early 2025) vs. MGID tells you which auctions your competitor is buying into.
  • The click traced to the landing page. The creative is the hook. The landing page is where the money is made. Analysis that stops at the ad image misses half the funnel.

A worked example: pull every live native ad for a competitor in health. You'll see the same playbook over and over. Curiosity gaps, fake-news framing, "doctors hate this" authority plays. Here's a live one from our index:

Taboola native health ad with a medication-warning curiosity hook
Caption: A live Taboola health ad from Vital Guardian, captured by OpenAdLibrary, June 2026.

The "MDs Identify 10 Medications Now Attached to Memory Problems (See the List)" angle is a textbook curiosity-gap listicle hook. When you pull the full set you might find twelve active Taboola variations on that same hook, plus four Outbrain creatives running a cleaner before/after angle. That split is already a finding: they test curiosity hooks on Taboola traffic and proof angles on Outbrain. You did not learn that from one screenshot.

It's not random which verticals dominate this feed. Across our whole index, finance leads with 17,232 creatives, then insurance (15,629), health (14,895) and ecommerce (13,872). On Taboola specifically, health (6,048) and finance (5,558) are the two biggest categories. So if you sell in finance, insurance, or health, the bad news is your space is crowded. The good news is the volume means winning angles surface fast and often.

This evidence layer is what a native ad spy tool exists to provide, and it's the single biggest force-multiplier in the framework. The better your feed, the better every downstream stage. For the network-specific mechanics of pulling Taboola, Outbrain and MGID ads, see how to spy on competitor native ads.

Stage 3: Score the creatives against a rubric#

Now you have evidence. The mistake here is treating all of it as equally interesting. You need a scoring rubric so "I like this ad" becomes "this ad scores 8/10 on a framework I can defend."

Score each creative, or each cluster of near-identical creatives, on five dimensions:

Record the angle, visible offer, creative format, first and last observation dates, and observed publisher coverage. Note missing captures and uncertain advertiser identity. Use these fields to compare visible approaches, not to score unobserved profits or scaling.

A creative observed across several dates or publishers deserves contextual review, but its duration and spread do not establish performance or distinguish a test from a scaled campaign.

Take this one, also live in our index:

Taboola native ad selling an alternative hearing device with a trend-framing hook
Caption: A Taboola health ad from Nebroo observed running 26 days, captured by OpenAdLibrary, June 2026.

The Nebroo and medication-list examples have different recorded observation spans. That difference describes collection history, not their conversion rates, campaign ages or status as controls and experiments.

Cluster before you score. Native advertisers run dozens of micro-variations of the same angle with different thumbnail crops. Group them, score the cluster once, and you see the real picture: maybe they have three angles in market, not forty ads. That collapse from forty to three is the actual insight.

Translate relevant patterns into original, substantiated test ideas. Record the observed feature, the proposed change, the business outcome and a test limit. Do not borrow a fabricated authority claim or assume a score predicts performance.

Stage 4: Describe observed activity and its limits#

Public creative observations do not establish a competitor's spend. Without a calibrated model and validation data, even a relative budget ranking can be wrong.

Observation duration indicates when the collection system saw a creative. It does not reveal continuous spend, conversion economics or why the advertiser retained it.

A quick caveat on the numbers, since it's the easiest place to fool yourself. Our index currently spans up to about 28 days of continuous observation per creative, so when I say an ad "ran 28 days" I mean we've seen it live every day for as long as we've been watching it, not that 28 is its lifetime ceiling. The general industry lore about "90-day winners" is real but it's lore, not something we can confirm from our own capture window. Treat our day-counts as a floor, not a finish line.

Here's a finance creative sitting at the top of that observed range:

Outbrain finance native ad framed as an expert Q&A about IRA taxes
Caption: An Outbrain finance ad from SmartAsset observed running 28 days, captured by OpenAdLibrary, June 2026.

The SmartAsset example has a longer recorded observation span than the tax-deadline example. Both can be inspected for messaging, but neither span demonstrates that the ad converts.

Taboola finance native ad with an IRS tax-deadline urgency hook
Caption: A Taboola finance ad from Fresh Start Information observed running 13 days, captured by OpenAdLibrary, June 2026.

"2026 - IRS Forgives Millions By June 30th Tax Deadline" leans on a hard date, so its 13-day run partly reflects a deadline mechanic rather than pure durability. Read longevity in context. A deadline ad and an evergreen Q&A ad earn their day-counts differently.

Count distinct creatives and observed publishers at a consistent scope to describe captured activity. Differences can result from coverage, formats, targeting or creative reuse; they do not establish a spend tier.

Summarize the activity you can observe:

For each advertiser, report the capture dates, distinct observed creatives, observed publishers and known collection gaps. Leave spend and conversion performance unknown unless supported by independent evidence.

A statement such as 'we observed more distinct creatives this month under the same collection scope' is supportable when the records show it. 'Spend is increasing' is a different claim and requires additional evidence.

Putting it together: a worked end-to-end pass#

Tie the stages into one pass so the framework feels concrete:

Hypothetical example: a watchlist produces 140 captured creatives grouped into 19 angle clusters. Select several relevant clusters for closer inspection, record their claims and destination patterns, and write original tests. Treat the numbers as an illustration, not a measured case study; do not raise bids solely because another advertiser appears in more captures.

That's the difference between watching competitor ads and analyzing them. The screenshot folder produces opinions. The framework produces a roadmap.

The supporting concepts worth knowing#

Two pieces of background sharpen the framework. First, this is one discipline inside the broader practice of competitive intelligence in advertising, the systematic, ethical gathering of public signals about how rivals go to market. Second, native runs on the same plumbing as the rest of the open web. It's programmatic advertising under the hood, which is why the same creative appears across dozens of publishers and why programmatic native winners scale so fast once proven. If your remit also covers banners, the display advertising primer rounds out the picture.

One practical note on the legal and ethical frame, since it always comes up. Studying public competitor ads is legitimate, and regulation is moving toward more transparency, not less. The EU's Digital Services Act now requires very large platforms to maintain public ad repositories, and the Commission has begun enforcing those obligations. The operational line to hold is simple: capture and study public creatives and landing pages, but don't click live ads in ways that cost a competitor money or pollute their data. Tooling that follows the click to the landing page without triggering a live ad click keeps you on the right side of that line. For the full hands-on method of finding what's running, see how to find out what ads your competitor is running.

Where OpenAdLibrary fits#

The framework is tool-agnostic. You can run it manually with enough patience. But Stages 2 and 4 are where a purpose-built source pays for itself: a continuously refreshed library of live native ads, captured at full creative quality, with the supply chain classified and the click traced to the landing page. We do exactly that across Taboola, Outbrain, MGID, Revcontent, Yahoo, MSN and more, and we surface the longevity and spread signals this framework runs on. The whole thing currently spans 589,036 creatives and 5.4 million ad observations (OpenAdLibrary index, June 2026), at $29.99/mo against rivals charging $80 to $400. You can see the 20 newest ads on the free tier with no card to pressure-test the workflow on your own competitors first. Start free and build your watchlist today.

Frequently asked questions

What is competitor ad analysis?
It is the systematic study of visible competitor messages, offers, formats and observed placements to develop hypotheses for your own tests. Public ads alone do not reveal conversion results or establish winning angles.
How do you estimate a competitor's ad spend without their data?
Creative counts, observation duration and publisher spread do not reliably establish spend by themselves. Report observed activity and collection limits instead. A spend estimate needs additional evidence and a validated method.
Which native ad networks should I monitor?
Start with the high-volume open-web platforms: Taboola, Outbrain (now operating as Teads after the 2025 merger), MGID and Revcontent, plus Yahoo, MSN/Microsoft and MediaGo. These carry the bulk of direct-response native spend in finance, insurance, health and ecommerce, which are the four largest verticals in our index (17,232, 15,629, 14,895 and 13,872 creatives respectively as of June 2026) and the spaces where competitor copying is most active.
Is analyzing competitor ads legal?
Public availability is not blanket permission to copy, redistribute or collect content by any method. Use authorized access, respect applicable terms and rights, and avoid live ad clicks that charge advertisers. Legal requirements depend on the activity and jurisdiction.
How often should I run competitor ad analysis?
Run a deep analysis monthly to reset your creative strategy, plus a light weekly scan to catch new launches and surging creatives while they're still fresh. The weekly cadence matters most in fast-moving verticals like finance and health, where an angle that ran for three weeks before you noticed it is already being copied by everyone else.
The OpenAdLibrary Team
Written byThe OpenAdLibrary Team
Ad intelligence & native advertising research

We build OpenAdLibrary, the open ad-transparency platform. Every day our systems capture live native ads across Taboola, Outbrain, MGID, Revcontent, Teads, Yahoo and MSN, identify the real advertiser behind each one, and follow the click to its landing page. These guides distill what we see in that data so you can research the market faster.