Advertisng Public Ad Corpus: Q4 2026 Research Snapshot

Direct answer

Advertisng's October 1 snapshot contains 257,588 stored public-ad records: 244,043 Meta and 13,545 Google, across 2,550 advertiser records. Under the current stricter quality model, 204,465 records pass exact public-readiness checks. These counts measure our captured corpus—not market share, spend, reach, conversions or campaign effectiveness—and should guide research coverage rather than performance claims.

What does the Q4 opening snapshot contain?

At October 1, 2026, 12:47 PM IST, the production corpus contained:

• 257,588 stored ad records;
• 244,043 Meta records and 13,545 Google records;
• 2,550 stored advertiser records;
• 94,692 rows marked active in the stored provider field; and
• 204,465 records that passed the current exact public-readiness gate.

The snapshot is a database count, not a census of all advertising on either platform. It reflects the advertisers and searches collected by Advertisng, the public records returned by the source platforms, normalization decisions, recrawls, retention and quality rules. Meta's official library describes public search for currently running ads, while the scope of its API varies by ad category and region. Google's official documentation describes the Ads Transparency Center as a searchable repository of verified advertisers and the ads they have shown on Google surfaces. Neither source promises that an independent collection represents the full ad market.

Original artifact: like-for-like corpus comparison

The following fields match the stored-field definitions reported in the August 9 Q3 baseline. Growth means that Advertisng captured and retained more rows under a broader ingestion programme. It does not establish that the platforms ran the same percentage more ads.

• All ad rows: 6,957 on August 9 and 257,588 on October 1; an increase of 250,631 records, or 3,602.6% / 37.0x.
• Meta ad rows: 6,082 on August 9 and 244,043 on October 1; an increase of 237,961 records, or 3,912.5% / 40.1x.
• Google ad rows: 875 on August 9 and 13,545 on October 1; an increase of 12,670 records, or 1,448.0% / 15.5x.
• Advertiser rows: 541 on August 9 and 2,550 on October 1; an increase of 2,009 records, or 371.3% / 4.7x.
• Rows marked active: 6,356 on August 9 and 94,692 on October 1; an increase of 88,336 records, or 1,389.8% / 14.9x.

Do not compare the Meta and Google row totals as market share. The sources expose different public surfaces, identities, ad histories and retrieval constraints, and the collection programme did not sample them symmetrically.

What changed in the quality and identity model?

The August baseline counted stored rows and the provider-derived active flag. Since then, the operational model has added stricter, separately measurable gates:

• frontendReady separates records that can render on the product surface;
• exact public readiness verifies a non-empty platform ad identity, a linked canonical advertiser, matching verified provider identity and a durable advertiser avatar;
• canonical advertiser records can merge duplicates or quarantine unsafe identities;
• ingestion rejections retain reasons for records that fail admission instead of silently mixing them into the public corpus; and
• quality statuses distinguish ready, partial, source-only, expired and recrawl-needed records.

These changes improve the integrity of what users see, but they make a strict-ready count non-comparable with the August stored-row baseline. The safe comparison is to retain both views.

Original artifact: October 1 quality funnel

• Stored corpus — 257,588 records (100.0%): all retained ad rows in the source-of-truth corpus.
• `frontendReady=true` — 241,402 records (93.7%): records marked renderable before the final identity and durable-avatar assertion.
• Exact public-ready — 204,465 records (79.4%): records passing the current strict display gate.
• Exact public-ready: Meta — 201,860 records (78.4% of all stored rows): a collection and readiness result, not Meta market share.
• Exact public-ready: Google — 2,605 records (1.0% of all stored rows): a collection and readiness result, not Google market share.
• Rows marked currently active — 94,692 records (36.8%): the latest pipeline state, not a count of every live ad on either platform.

The corpus also contained 145,638 `READY`, 80,499 `SOURCE_ONLY`, 29,706 `PARTIAL`, 1,271 `SOURCE_EXPIRED`, 244 `TEXT_ONLY_OK` and 230 `NEEDS_RECRAWL` rows. The retained ingestion-rejection ledger contained 902 rows: 718 Meta and 184 Google. Rejection totals are an operational quality measure, not a failure rate, because the denominator and retention period differ across ingestion jobs.

How should a boutique agency use this research?

Use the corpus to narrow questions before opening client accounts or launching tests:

1. define a specific competitor, category, market and observation window;
2. collect observable creative, copy, offer, destination and served-date evidence;
3. preserve the public source and capture date for each observation;
4. group repeated patterns without converting frequency into a performance claim; and
5. test the resulting hypothesis with client-owned spend, conversion, margin and experiment evidence.

The Meta Ad Library guide and Google Ads Transparency Center guide explain source-specific research. Use the creative-analysis rubric to code observations consistently. The corpus snapshot answers how much evidence was available in this collection on one date; it does not decide which creative a client should run.

Why does exact public readiness matter for competitor research?

A large unfiltered database can create false confidence. Missing creative, mismatched advertiser identities, expiring source media and duplicate brands weaken the evidence trail even when a row technically exists. The exact gate makes a smaller promise: a displayed record has the minimum identity and media conditions required for a public research surface.

That still does not make every record equally informative. Agencies should record whether a finding is based on one creative, a repeated pattern, a recent change or a long-running observable. They should also revisit source links because public-platform availability can change.

What can public ad data not prove?

Public ad data cannot prove a competitor's spend, bids, budgets, targeting settings, audience composition, placement allocation, attribution model, experiments, conversions, revenue, profit, incrementality, lead quality, inventory, account structure or why an ad was started, stopped or changed. Corpus size, active labels, repetition and longevity are research signals—not proof of business performance or causal effectiveness.

Frequently asked questions

Does 257,588 mean Advertisng contains every Meta and Google ad?

No. It is the number of retained rows in Advertisng's production corpus at the snapshot time. Coverage depends on collection scope, source availability, advertiser identities, searches, recrawls and quality rules.

Does Meta's larger row count mean Meta has more advertising than Google?

No. The platforms expose different public records and the collection programme is not a symmetric market sample. The split describes this corpus only.

Are all stored ads visible in the product?

No. At the snapshot time, 204,465 of 257,588 rows passed the exact current public-readiness gate. Stored, frontend-ready and exact public-ready are separate states.

Is an active label proof that an ad is profitable?

No. It is a source or pipeline status. It does not reveal spend, results, margin, incrementality or the advertiser's reason for continuing the ad.

Can an agency cite these numbers in a client deck?

Yes, if the slide states the October 1 snapshot date, the Advertisng corpus scope, the collection and quality caveats, and the difference between stored and exact public-ready records. Do not relabel the counts as platform totals or market share.

A restrained next step

Choose one client category and build a dated evidence set from exact public-ready records. Record the source, advertiser identity, observable pattern and uncertainty before proposing a test. Advertisng's public ad-intelligence workspace can help preserve that research trail; client-owned measurement must decide whether the test worked.

Primary sources and research record

1. Advertisng production corpus snapshot, generated October 1, 2026 at 12:47 PM IST with the repository's read-only quarterly audit. Counts were grouped from raw stored platform fields and evaluated separately through the current strict public-readiness predicate.
2. Advertisng: Research methodology, collection, normalization, quality, deduplication and evidence limits; reviewed October 1, 2026.
3. Advertisng: Q3 2026 public-ad-library research baseline, August 9 comparison counts and prior methodology disclosure; reviewed October 1, 2026.
4. Meta: Meta Ad Library, public search scope and source context; reviewed October 1, 2026.
5. Meta: Ad Library API, available fields and category or regional scope; reviewed October 1, 2026.
6. Google Ads Help: Ads transparency, searchable advertiser and ad repository across Google surfaces; reviewed October 1, 2026.
7. Google Ads Help: Advertiser verification, disclosure and Ads Transparency Center context; reviewed October 1, 2026.