What Is PPCEye? Data Sources, Coverage, and Methodology

PPCEye reorganizes the data in Google's official public ad archive into two views the official interface does not offer: which landing pages an advertiser is driving traffic to, and which advertisers are buying traffic for a given landing domain. As of 2026-08-15 the database holds roughly 853 million creatives, 30.1 million advertisers, and 307,400 distinct landing domains. This post explains where those numbers come from, what happens to them along the way, and where you should not trust them.

Where the data comes from

The only source is the public ad archive Google has operated since 2023 — free, and open to anyone. We do not buy third-party datasets, we do not collect user behavior through browser extensions, and we do not extrapolate from clickstream panels. Every number here can be checked, one record at a time, against Google's own public interface.

That single choice defines the limits of what PPCEye can do: if the official archive does not have it, neither do we. What it does have, we organize in ways the official interface never attempted — and that is where the value is. For a breakdown of exactly what the official tool can and cannot answer, see Google's Official Ad Archive: What You Can and Can't Find.

From one ad to one row of data: five steps

The raw input is an ad image plus a little metadata. Turning that into a searchable statement like "this advertiser is buying traffic for this domain" takes five steps:

Step 3 is the one people underestimate. Search-network image ads do not expose their final landing URL as a structured field — the address is usually just a line of small text printed on the image itself. Without OCR, a creative is only "some advertiser ran some picture": it cannot be attributed to any domain, and domain-level search becomes impossible. We run multi-engine OCR (one multimodal model service as primary, two more as fallbacks), with retries and per-creative status tracking.

Every creative resolves to exactly one root domain, in a fixed priority order: the domain that was searched for > the domain field returned by the API > the URL read by OCR. That rule is enforced at the data layer by database triggers, so the attribution you see on an advertiser page and on a domain page can never disagree.

Coverage

Inside those totals sits a smaller set with enough data to deserve a standalone page: an advertiser needs ≥10 creatives, ≥2 landing domains, and activity within 90 days; a domain needs ≥5 creatives, ≥2 advertisers, and activity within 365 days. That set changes daily, so we don't quote a number here that would be stale by the time you read it — the advertiser directory and the multi-advertiser domain directory show the current count on the page itself.

Industry mix of analyzed domains

Distribution across the 118,000 of 121,600 analyzed domains that carry an industry label (as of 2026-08-15):

One processing detail worth stating plainly: industry labels are model-generated, and the same category comes back spelled several ways (Software / SaaS vs Software/SaaS, Recruitment vs Recruiting, United States vs USA). The table above is after normalization — grouping the raw column directly produces a different ranking. The same applies to geography: merging United States and USA puts the US at 22,100 domains, second only to the 39,900 marked as global.

How many advertisers sit behind one domain

This is the question we get most often, and the one the official interface cannot answer at all. Across every domain above the inclusion bar (snapshot dated 2026-08-14):

More than half of all domains have just 2–4 advertisers — typically the brand itself plus an agency or an affiliate or two. The crowded end is dominated by platform domains: linktr.ee has over 400,000 advertisers pointing at it, and canva.site, squarespace.com and naver.com are all in the hundreds of thousands. Those numbers describe "many small advertisers using the same site-builder or link tool as a landing page," not a brand under siege. Always classify the domain before reading the count.

The advertiser side has the same shape, inverted. Most advertisers push one or two landing domains; the long tail belongs to agencies, resellers and directory businesses — Yellow Book USA, Inc touches over 7,000 landing domains, Dealer Spike more than 1,700, Epsilon Data Management, LLC around a thousand. Set that against Tripadvisor, which runs more creatives than any of them across exactly two domains. A high landing-domain count identifies an intermediary, not a big spender — and the two call for completely different responses.

Known limits, and what we deliberately don't do

Better stated up front than extracted under questioning:

• This is not every ad on the internet. Coverage centers on search-network image creatives; display ads run through a separate path. Video and text-only ads are not a current focus.
• OCR fails sometimes. A meaningful share of creatives yield no usable URL. Those are left unattributed rather than guessed at — undercounting beats misattributing.
• Industry, geo focus and risk level are model outputs. They are research leads, not findings of fact, and certainly not legal conclusions. Risk level especially: of the 121,600 analyzed domains, 6,900 are flagged medium and 1,400 high, which means only that a model thinks a human should look.
• The data is a snapshot, not a live feed. Each page shows its own data cutoff, and free access carries an additional freshness lag.
• We do not sell keyword-level bid data. No public source can tell you what a competitor paid for a specific keyword; anything claiming otherwise is modeled inference. What we provide is verifiable delivery fact: who ran ads, pointing where, for how long, in which countries.

Getting started

No account required: enter a root domain in domain search to see who is buying traffic for it, or search a company in the advertiser directory to see which landing pages it promotes. To browse by activity, start with the rankings.

Anonymous visitors see a sample — totals are real, rows are sampled, and each page states how many records are hidden. Complete enumeration and the freshest data require a subscription. We are explicit about one thing here: everyone, search engine crawlers included, receives byte-identical HTML. We do not serve different content to bots than to people.

FAQ

Is this data legal to use?
The source is an ad archive Google publishes deliberately; anyone can look up the same records through Google's own public interface. We organize and index public data — nothing non-public is involved.

How often is it updated?
Advertiser and domain crawls run continuously, with more frequent refreshes for high-activity entities. Every page displays its own data cutoff — trust that, not the 2026-08-15 date in this post.

Why can't I find a particular advertiser or domain?
Three possibilities: it has no records in the official archive at all; OCR could not read a landing page from its creatives, so nothing is attributed yet; or we simply have not crawled it — in which case a logged-in user can trigger a fetch manually.

Can I export the data?
Subscribers can browse it in full on-site. For bulk export or API access, email [email protected].

How does this relate to SpyFu, SimilarWeb and similar tools?
Different sources, therefore different kinds of claim. Their keyword and traffic figures are modeled from panels and clickstream samples; we show delivery records that can be verified one by one against the official archive. The two are complementary, not interchangeable — see the comparison pages for specifics.

Methodology note: every figure above comes from the PPCEye production database as of 2026-08-15; the advertisers-per-domain distribution is a 2026-08-14 snapshot. Counts move as crawling progresses — the live values on each detail page take precedence.