OpenAdLibraryOpenAdLibrary
Ad Transparency & Supply Chain

Native Ad Data API: Programmatic Access to Captured Native Creatives

Most native ad intelligence is locked behind a dashboard. Here is how analysts and data teams pull captured creatives, resolved advertisers, supply-chain hops and landing pages as structured records over an API.

Diagram of a native ad data API returning creative, advertiser, supply-chain and landing-page records as JSON

If you have ever tried to do real analysis on native advertising, you know the wall. The intelligence exists, but it is trapped behind a dashboard. You can browse one creative at a time, eyeball a handful of advertisers, maybe export a CSV that is already stale by the time it finishes downloading. What you cannot do is treat native ad data the way you treat every other dataset in your stack: query it, join it, schedule it, pipe it into a model.

This guide is for the people who want the data, not the dashboard. Analysts building competitive trackers. Data teams enriching a CRM. Quants backtesting creative angles. And, increasingly, the AI agents doing that work without a human in the loop. The question is plain. How do you get captured native creatives, the supply chain behind them, and the landing pages they point to as clean, structured records over an API?

For context, a June 2026 snapshot of OpenAdLibrary's collection held 589,036 creatives, 5.4 million ad observations, 25,933 distinct advertisers and 926,259 captured landing pages across 42 networks. That is a historical measure of what we had collected, not the size of the native market and not a current count. The job of an API is to hand you the records you asked for.

What a native ad data API actually returns#

A native ad data API is an endpoint that returns native advertising intelligence as machine-readable records instead of a web page. A record can bundle the creative (image and headline), the publisher and network that served it, the supply-chain hops behind the placement, the landing page the click resolves to, and timing signals like first-seen and last-seen dates. You filter, join, and aggregate it like any database table.

That tidy definition hides a lot of engineering. Unlike a search ad or a Meta ad sitting in an official library, a native placement is rendered by a native ad widget, the "Recommended for you" or "Around the web" strip at the bottom of articles. Those widgets are dynamic, geo-targeted, and personalized. That is why a single static scrape can miss ads that other visitors see. A real dataset has to capture across geos and over time, then deduplicate down to a stable creative.

Taboola finance ad captured in June 2026, with a headline about an IRS tax deadline
Caption: A Taboola finance creative from Fresh Start Information, captured by OpenAdLibrary in June 2026 and observed over 13 days.

The ad above is a typical example of the genre in our captures: a deadline, a vague authority and a dollar figure. We had observed it for 13 days when we last saw it. That tells you how long we saw it, not whether it was profitable or whether its claim is accurate.

The four data layers of a native ad dataset#

Think of native ad data as four stacked layers. Some tools surface only the top layer. The value compounds as you go down.

Layer What it captures Why it matters
Creative Full-resolution image, headline, call to action The actual ad, not a thumbnail. Feeds Copy DNA and visual clustering.
Network and publisher Which native network served it, on which site Lets you answer "who is buying ads on this site?" and segment by source
Supply chain The ad-tech hops between publisher and advertiser Reveals resold inventory, intermediaries, and routing
Landing page The pre-lander and final destination of the click Surfaces the real advertiser and the offer being sold

The jump from layer two to layer four is the whole game. A placement is served by a network (Taboola, Outbrain, MGID, Revcontent, Teads, Yahoo, MSN) but the network is not the advertiser. The advertiser is whoever sits at the end of the click. Counting creatives by network does not tell you who is actually buying. Following the redirect chain, server-side and without firing a real click on a live campaign, is what turns a pile of anonymous widget creatives into competitive intelligence you can act on.

For the mechanics of the routing in between, see the native ad supply chain. For working out the likely buyer behind the serving network, read how to identify the likely network behind a native ad.

Duration is a useful column, and easy to over-read#

A timestamp is among the most useful fields in a native ad dataset, because it lets you measure how long an ad was observed.

You cannot see anyone's spend in native, and duration does not stand in for it. A creative that stays up across many sites for weeks may be profitable, but it may also be a brand campaign, a test with budget left, or an ad nobody has reviewed. In the June 2026 captures cited here, the longest continuously observed creatives had run about 28 days, which was the limit of our observation window at the time, so longer runs could not be seen. The headlines that reached that limit are worth a look as examples.

Outbrain finance native ad from SmartAsset about avoiding taxes on IRA withdrawals
Caption: SmartAsset's IRA-withdrawal angle on Outbrain, observed running 28 consecutive days, OpenAdLibrary, June 2026.

SmartAsset's "How Can I Avoid Paying Taxes on IRA Withdrawals?" was on the 28-day list. So were several "What's Your IQ Level?" quiz ads from My IQ on the Microsoft Audience Network, and a hearing-aid ad from Hidden Hearing with two different creatives at the same limit. An advertiser with several variants observed for the full window was still running them. We cannot tell whether the offer was profitable, whether the budget was large, or whether the claims in the ads were accurate.

One caution worth stating plainly. The "90-day winner" you hear quoted in media-buying circles is industry lore, not our measurement. What we could verify in these June 2026 captures was continuous runs of up to roughly 28 days per creative. Treat the 90-day figure as folklore until your own dataset timestamps it.

API vs dashboard: match the tool to the job#

There is nothing wrong with a dashboard for ad-hoc research. The two interfaces are built for different jobs, and conflating them can lead teams to pay for tools they barely use.

  • Use a dashboard when a human is doing exploratory work: eyeballing a competitor's angle, pulling a few creatives for a deck, scanning what is hot this week. This is the classic ad spy tool workflow.
  • Use an API when the consumer is a system: a scheduled job, a dbt model, a feature store, a Slack alert, or an AI agent. You want records, not pixels, and you want them on a cron.

The practical tell is repetition. If you find yourself manually exporting the same view every Monday, you do not have a research problem. You have an integration problem, and an API solves it. This is also where the broader discipline of ad intelligence is heading: less manual browsing, more programmatic feeds wired straight into competitive and creative workflows.

A concrete pipeline#

Here is how a data team actually wires native ad data into something useful. The pattern holds whether you are tracking a whole vertical or one competitor.

  1. Define the watchlist. A set of advertisers, landing-page domains, or publisher sites you care about. The search endpoint documents filters such as advertiserDomain, landingDomain and publisher, so you pull only what you track.
  2. Pull on a schedule. A nightly job asks for ads last seen since yesterday (the search endpoint documents a lastSeenFrom filter, and a separate changes endpoint takes a since watermark) and upserts them by id, so the same creative is not stored twice.
  3. Join to your own data. Match landing-page domains against your CRM or a competitor table. Creatives whose landing domain matches are then attributed to a company you already track.
  4. Compute observed duration and spread. Group by creative ID. Measure days observed (last seen minus first seen) and the count of distinct publishers. Long-observed, widely spread creatives are candidates for a closer look, not confirmed winners.
  5. Act. Push a daily digest of new long-observed creatives to Slack, feed the creatives into a clustering model, or hand the structured plan to an agent.
Taboola health ad captured in June 2026, with a headline about replacing hearing aids with a new device
Caption: A Taboola health creative from Nebroo observed over 26 days, the kind of long-observed creative step four is designed to surface, OpenAdLibrary, June 2026.

Step four is where a timestamped dataset is useful. A creative like the Nebroo hearing-device ad above, observed for 26 days, is the kind of record your pipeline can flag for review. Duration is legible only if your dataset timestamps every observation, which is why, in the June 2026 snapshot, 5.4 million observations sat behind 589,036 creatives. The observations are the time series. The creative is just the key. A flag is a prompt for a person or a model to look closer, not a verdict that the ad works.

What separates a usable dataset from a noisy one#

Volume is the wrong headline metric. Ten million stale, poorly-deduplicated, network-only records are worse than a smaller set that is fresh, attributed, and clean. When you evaluate a native ad dataset or its API, pressure-test it on five things:

  • Freshness. Are placements captured live, with first-seen and last-seen dates? Or is it a periodic dump that is already weeks old?
  • Creative fidelity. Is the actual image stored at full quality, or a downscaled thumbnail that is useless for visual analysis?
  • Attribution depth. Does it stop at the network, or trace the click to the final landing page and the real advertiser?
  • Capture method. Does it read public ads without firing real clicks on live campaigns? Synthetic clicks spend other advertisers' money and corrupt their data, and yours.
  • Coverage breadth. Does it span the major native networks and capture across geos, or is it one publisher in one country?
Taboola home and garden native ad about solar home batteries
Caption: A Taboola home-and-garden creative from Solar Battery Subsidy, observed over 27 days and captured at full resolution by OpenAdLibrary, June 2026.

Notice the detail in that solar-battery ad. At full resolution you can read the offer, the framing, and the visual treatment, which is what feeds visual clustering and copy analysis. A thumbnail throws all of that away. These five tests are also what distinguish genuine ad transparency from a screenshot library with a search box bolted on. Native has no official library comparable to Meta's or Google's, and our guide to what a native ad library is and how it differs from Meta's and Google's tells that story.

Why native is harder, and more rewarding, than it looks#

Meta, Google, Microsoft (for Bing ads in the EU and EEA) and TikTok publish official ad libraries, and some of them exist because the EU's Digital Services Act requires a public ad repository from designated very large online platforms and search engines. The European Commission's list of those services (updated 31 August 2026) does not include Taboola, Outbrain/Teads, MGID or Revcontent, and we are not aware of a public ad library from them, so programmatic native advertising has no official library comparable to Meta's. That gap is why independent datasets exist. The ads are shown in public; organising them is a separate job.

The reward for solving the capture problem is a record of what advertisers show in public. Pair that with programmatic access and you get native ad data you can use as infrastructure, not only as a site you log into. For the wider context on how ad libraries, laws, and tools fit together, start with the pillar on what ad transparency is and how to use it, and open ad libraries explained covers the tooling landscape.

How OpenAdLibrary fits#

OpenAdLibrary publishes this blog and sells the product described here. It captures public native ads from Taboola, Outbrain, MGID, Revcontent, Teads, Yahoo and MSN, API-first with a browser crawler as a fallback, and since October 2026 its library also includes Meta and TikTok ads. It stores the creative, classifies the supply chain and follows redirects to the landing page without clicking live ads. A REST API and a hosted MCP server expose public data from the same library to an analyst, a pipeline or an AI agent. Ad data and research over both are part of the Founders plan: a free account's key reaches only that account's own endpoints, and every other endpoint answers with a payment-required response. According to the API documentation, it returns public creative content, thumbnail URLs, aggregate counts and the redirect chain behind an ad, including the advertiser URL the click ends on. It does not return the Meta- and TikTok-only data shown in the web app, such as the Meta spend estimate and EU reach, and its free-text search is answered from the 3,000 most recently seen matching ads rather than the whole corpus. Each key has a daily request quota (the API reference lists 5,000 requests a day by default). Search results can include Meta and TikTok ads, whose ids begin meta- or tiktok-, and the single-ad endpoint has no record for them.

The Founders plan is $29.99 a month and includes the REST API and the hosted MCP server. A free account needs no card and shows the 20 newest Meta, TikTok and native ads in the app, so you can check whether the dataset looks relevant to your watchlist before paying; a thorough check, and any API use, needs the Founders plan.

Start free to look at the data in the app; API access needs the Founders plan.

Frequently asked questions

What is a native ad dataset?
A native ad dataset is a structured collection of native ad records where every field is queryable, each one capturing the creative image and headline, the publisher and network that served it, the supply-chain hops behind the placement, and where the click resolves. Unlike a screenshot gallery, you can filter, join, and aggregate it like any other table, which is what lets you measure things like how long a creative has run.
How is a native ad data API different from a dashboard ad spy tool?
A native ad data API returns the same intelligence as machine-readable records, while a dashboard is built for a human browsing one ad at a time. With an API you can pull thousands of creatives, join them to your own data, schedule refreshes, and feed downstream models or AI agents without anyone clicking through a UI. The practical tell that you need one: you keep exporting the same view by hand every week.
Does pulling native ad data require clicking on live ads?
No. A well-built native ad dataset captures public ads and follows redirects to record the landing page without clicking live ads. OpenAdLibrary captures native ads API-first, with a browser crawler as a fallback, and traces destinations without live ad clicks. Avoiding clicks keeps you from spending other advertisers' budgets or distorting their analytics, which is why capture method is one of the first things to vet in any provider.
Can I get the real advertiser behind each native ad, not just the network?
Yes, as long as the dataset traces the full click path to the final landing page. The placement is served by a network like Taboola or Outbrain, but the click resolves through redirects to a destination tied to the actual advertiser. Following that chain is what separates useful native ad data from a list of anonymous widget creatives labeled only by serving network.
How fresh does native ad data need to be?
For competitive work, freshness matters more than raw volume, because native creatives rotate constantly and a stale dump misses what is running now. A dataset that captures placements over time and timestamps first-seen and last-seen dates lets you measure how long an ad was observed. That shows activity, not profit: a long observed run may reflect a brand budget, a test or slow optimisation. In the June 2026 captures cited in this article, the longest continuous runs we could observe were about 28 days per creative, the limit of our observation window at that time.
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.