Skip to content
Primores AI wiki

Ad Transparency Surfaces 2026: Every Platform's Ad Library as a CI Data Source

Ad Transparency Surfaces 2026

TL;DR: Every major ad platform now runs a public ad repository — but what you can see depends less on the platform than on where the ad was targeted. The EU’s Digital Services Act (Article 39) forces large platforms to publish targeting parameters and reach estimates for EU-targeted ads, so EU ad data is structurally richer than global data on every surface: TikTok’s library covers only EEA/Switzerland/UK ads, LinkedIn shows targeting and impressions only for EU-targeted ads, Microsoft’s library contains only EEA-served ads. Two universals: no platform exposes spend data on any public surface, and retention converges on ~1 year after the ad’s last view. The X repository is formally deficient (€120M DSA fine, Dec 2025 — the first ever). And since July 9, 2026, Google labels AI-generated ads — but in the consumer-facing My Ad Center, not the Ads Transparency Center, and reliably only for ads made with Google’s own AI.

This page is the data layer under two layers of the competitor-analysis/overview stack: continuous monitoring (Layer 2) and creative reverse engineering (Layer 5). The methodology pages tell you what to do with competitor ads; this page maps where the raw ad data actually lives, what each surface exposes, and what it structurally cannot tell you.

Meta’s Ad Library — the oldest and deepest of these surfaces — has its own cluster in this wiki and is only summarized here: see seo/meta-ad-library-api for programmatic access and the questions/how-to-access-meta-ad-library Q&A cluster for the web UI.

The DSA Article 39 layer — why EU ad data is richer everywhere

The single most useful thing to understand about ad transparency in 2026: most of these surfaces exist because EU law requires them, and their data boundaries follow the law, not the product.

The EU Digital Services Act’s Article 39 obliges very large online platforms (VLOPs) to maintain public ad repositories that include the ad’s content, the advertiser, who paid, run dates, targeting parameters, and aggregate reach numbers — for ads presented in the EU, retained at least one year after last presentation. That produces a consistent cross-platform pattern (synthesis of unanimous per-platform findings, all Tier 1 primary sources; verified in a July 2026 adversarial sweep):

PatternInstances
Coverage tracks the EEA, not the platform’s footprintTikTok’s Commercial Content Library: EEA + Switzerland + UK only. Microsoft Ad Library: EEA-served ads only.
Targeting + reach disclosed only where the law forces itLinkedIn: detailed targeting parameters and impression breakdowns for EU-targeted ads; none of it for non-EU ads. Meta: same split (richer EU commercial data — see seo/meta-ad-library-api).
Retention ≈ the Article 39 minimumTikTok and LinkedIn both keep ads visible ~1 year after last view/impression, then they’re gone.
Enforcement is realX’s €120M fine (Dec 2025) named its deficient ad repository specifically.

And the universal negative: no platform in this set exposes spend data on any public surface. Reach and impression figures, where shown, are estimated ranges — not measured metrics. Treat any tool or article quoting a competitor’s “ad spend” from a public library as inference, not data.

The practical CI consequence: if a competitor runs ads in the EU, look at the EU views of these libraries — even if your market is the US. The targeting parameters, reach estimates, and demographic breakdowns disclosed for their EU campaigns are usually the only public window into how they think about audiences anywhere.

Google — Ads Transparency Center + the new AI labels (July 2026)

Google’s ad transparency story split in two in July 2026, and the split matters:

What shipped July 9, 2026: AI-generation disclosure — in My Ad Center, not the Transparency Center. Google added a “How this ad was made” section to the consumer-facing My Ad Center panel, reachable globally via the three-dot menu / info icon on ads across Search, YouTube, and Discover. Rolling out as of mid-July with no stated completion date — timed 24 days before the EU AI Act Article 50 transparency deadline (Aug 2, 2026; see glossary/content-provenance for the disclosure-law map). (Tier 1 — Google’s announcement, verified 3-0; corroborated by TechCrunch, Search Engine Land, MediaPost.)

Also from July 2026: advertisers may place text or visual AI labels directly inside image and video creatives (exempted from watermark/text-overlay policy violations), with an “AI label setting” rolling out through July across Google Ads, DV360, CM360, Merchant Center, and Ads Editor; ads targeted to the EU, India, and New York that are AI-modified additionally get visual overlays on the media itself. (Tier 1 policy doc, verified 3-0, two merged claims.)

Two framing disciplines for CI use:

  1. This is not a Transparency Center feature. The phrase “Ads Transparency Center” appears nowhere in Google’s announcement; you cannot (as shipped) filter or view AI-generation status in the Transparency Center. Coverage claiming “Google added AI labeling to the Ads Transparency Center” overstates what shipped — that framing was refuted (0-3) in verification. (3-0 on the negative.)
  2. The label is an unreliable signal. Auto-labeling applies only to ads made with Google’s own generative-AI tools (those labels can’t be removed); for ads made with third-party AI, disclosure is a voluntary advertiser control that Google does not verify or enforce — though EU/India/NY law independently requires disclosure for certain AI ads, with the obligation on the advertiser. Absence of a label does not mean an ad isn’t AI-generated. (2-1 split vote — phrase as “not verified or enforced by Google,” not “not required.”)

Honest gap: the Ads Transparency Center’s own core field set (per-advertiser/per-ad data, EU-specific fields, verification requirements, any API or bulk access) did not survive verification in this research round — no claims are made here about it. Marked open for a follow-up sweep; don’t fill the gap from vendor blogs.

TikTok — Commercial Content Library + Commercial Content API

TikTok runs two different surfaces, and only one was verifiable this round:

Commercial Content Library (CCL) — TikTok’s DSA repository, covering only ads shown to users in the EEA, Switzerland, and the UK. No global coverage: the 2023 launch promised “more countries in the future,” and no expansion has shipped as of mid-2026. (Tier 1, verified 3-0.) Per-ad detail pages expose:

  • advertiser name + who paid for the ad + advertiser’s registered location
  • first/last shown dates
  • estimated unique-users-reached range (bucketed, e.g. “30K” — not exact counts) + per-country reach
  • targeting summary: age, gender, location, interest criteria
  • no spend data, no CTR — anywhere. (Tier 1 support page + API schema, verified 3-0; the API schema contains no cost field at all.)

Commercial Content API — programmatic access to the same data, and notably targeting and reach are programmatically accessible (targeting_info with country/province/city, age/gender maps, interests, number_of_users_targeted; unique_users_seen with regional/age/gender breakdowns) — more than most global ad-library UIs expose. Access is gated behind a TikTok for Developers account plus an application review (~2 business days); applicants can be located anywhere, but the data covers EU-country ads only in the current phase. (Tier 1 developer docs, verified 3-0.)

Timing and retention limits that shape CI workflows (Tier 1, verified 3-0):

  • an ad enters the CCL only after at least one view; appears within up to 24 hours of first view or of a campaign change
  • visible for one year after its last view (including paused/inactive ads); API access matches the same window — the DSA Article 39 minimum. Ads older than that are simply gone from the public record: archive what you want to keep.

Honest gap: TikTok’s Creative Center / Top Ads — the separate, globally-scoped inspiration surface — produced no surviving claims this round. It exists, it’s a different beast (engagement-ranked ad showcases, not a compliance repository), but its 2026 specifics are unverified here.

LinkedIn — Ad Library (the sharpest EU-vs-global split)

LinkedIn’s Ad Library is the cleanest illustration of the two-tier system (all Tier 1 LinkedIn Help pages, live-fetched July 2026):

Global baseline (all ads): ad preview, ad format, advertiser name, payer name, restricted status — that’s the entire field list. No spend or budget data anywhere. Coverage: ads that ran after June 1, 2023; each ad searchable for one year after its last impression; new ads appear within 24–48 hours of first impression. Restricted ads show even less (preview, advertiser, and payer hidden). (Verified 3-0.)

EU-targeted ads (the rich tier): detailed targeting parameters across ~12 categories — language; job function, seniority, title, skills, years of experience; education; company category/size/revenue/name/industry; location (permanent or recent); matched audiences; inferred age; inferred gender; member interests/traits; predictive-audience and audience-expansion use — plus estimated total impressions with a per-country breakdown (rounded to the nearest percent; values under 1% shown as “<1%”), run dates, and EU-only search filters for impression ranges and targeting parameters. (Verified 3-0 ×3 + 2-1 on the rounding detail, confirmed live against the primary source.)

For B2B competitor analysis this is remarkable: LinkedIn’s EU disclosures are the only public window into a competitor’s job-title/seniority/company-size targeting logic — the core of any B2B media strategy. Caveats: individual ads display only up to ~3 targeting parameters (Tier 3 practitioner guides only — treat as soft), and impression figures are estimates.

X — formally deficient (treat as unusable for now)

On December 5, 2025 the European Commission fined X €120 million for DSA transparency breaches — the first non-compliance decision ever issued under the DSA — with the deficient ad repository (Article 39) among the three named breaches. The Commission found the repository lacks the ad’s content and topic and the paying legal entity, and imposes access barriers (excessive processing delays) that hinder independent scrutiny. X had 90 working days (~mid-April 2026) to submit an action plan, followed by a Board opinion and a final Commission decision — a remediation timeline running through 2026, with periodic penalty payments possible. (Tier 1 — EC press release IP/25/2934, verified 3-0, three merged claims.)

CI read: don’t build competitor monitoring on X’s repository until remediation lands; whatever it returns today is officially incomplete. Whether the action plan was submitted/accepted is an open follow-up item.

Microsoft — the quietly useful one

Microsoft’s Ad Library (ads shown on Bing) has a free public REST API — adlibrary.api.bingads.microsoft.com/api/v1/ — that requires no sign-up or login. Unauthenticated requests face stricter (unpublished) rate limits; higher limits need a Microsoft Advertising account + developer token. Fields include TotalImpressionsRange, ImpressionsByCountry, and Targets. The big caveat: the library contains only ads served within the EEA (since June 2023) — it’s a DSA-compliance surface like the others. (Tier 1 — Microsoft Learn docs, updated June 2026; verified 3-0.)

For anyone piping ad data into an LLM workflow, this is the lowest-friction programmatic surface of the whole set — no token dance at all for small volumes.

Cross-platform comparison

SurfaceCoverageTargeting dataReach/impressionsSpendAPIRetention
Meta Ad LibraryGlobal (richer EU tier)EU commercial + regulated onlyEU commercial + regulated onlyRegulated categories onlyYes — free, token required (seo/meta-ad-library-api)~7 yrs political; ~1 yr EU commercial
Google Ads Transparency CenterGlobal(unverified this round)(unverified this round)No(unverified this round)(unverified this round)
TikTok CCLEEA + CH + UK onlyYes (age/gender/location/interest)Yes (bucketed ranges + per-country)NoYes — gated application (~2 days)1 yr after last view
LinkedIn Ad LibraryGlobal since Jun 2023EU-targeted ads only (~12 categories)EU-targeted ads only (% by country)NoNo1 yr after last impression
X ad repository(formally deficient)——No——
Microsoft Ad LibraryEEA-served onlyYes (Targets)Yes (ranges + by country)NoYes — free, no sign-up(not stated)

Using these surfaces for CI — four rules

  1. Go where the disclosure is, not where your market is. A US-focused competitor with EU campaigns discloses its targeting logic in the EU views. That’s frequently the only public evidence of how they segment audiences.
  2. Never quote reach as measurement. Every figure is a bucketed estimate (“30K”, “<1%”). Useful for relative weight across a competitor’s ads; meaningless as absolute metrics. Spend figures from public libraries don’t exist — anything claiming otherwise is modeled.
  3. Archive on sight. The one-year-after-last-view windows mean the public record of a competitor’s campaign self-destructs. The weekly-pull-and-diff pattern from seo/meta-ad-library-api generalizes: pull, store, diff — your archive becomes the only historical record.
  4. Don’t trust AI labels as a creative-analysis signal. On Google, auto-labels cover Google-AI-made ads only; third-party-AI disclosure is voluntary and unverified. Label absence ≠ human-made. (Platform labeling mechanics live in glossary/content-provenance.)

What this page deliberately doesn’t claim (open items)

Verification survivorship, honestly labeled — these produced no surviving claims in the July 2026 sweep and are open targets, not findings:

  • Google Ads Transparency Center core field set, verification requirements, API/bulk access
  • TikTok Creative Center / Top Ads 2026 specifics
  • Snapchat, Pinterest, and Amazon transparency surfaces (existence/state unverified — do not read this as “no library exists”)
  • Cross-platform aggregators (whether any legitimately aggregate Meta + TikTok + Google, and what they can legally access given EU-gated APIs)
  • X’s remediation status (action plan due ~April 2026 — accepted? rejected?)

Do-not-cite (refuted in verification): “the July 2026 Google change is in-creative labeling only” (0-3 — the My Ad Center panel disclosure also shipped); “Google added AI labeling to the Ads Transparency Center” (unsupported by the primary source).

Key Takeaways

  • The DSA Article 39 layer is the structural story: EU-targeted ad data is richer than global data on every platform — coverage, targeting, and reach disclosures all track the law, not the product.
  • No public surface exposes spend data. None. Reach figures are bucketed estimates.
  • Retention converges on ~1 year after last view (TikTok, LinkedIn) — archive competitor ads on sight or lose the record.
  • Google’s July 9, 2026 AI labels live in My Ad Center, not the Transparency Center — and are reliable only for Google-AI-made ads. Label absence means nothing.
  • TikTok’s Commercial Content API and LinkedIn’s EU views are the underused gems: programmatic targeting + reach data (TikTok) and B2B targeting-logic disclosure (LinkedIn) that most CI programs never look at.
  • X’s repository is formally deficient (€120M DSA fine — the first ever); Microsoft’s is the lowest-friction API (free, no sign-up, EEA-only).
  • Meta remains the deepest surface — covered in its own cluster: seo/meta-ad-library-api.

Sources

Tier 1 (primary, all verified 3-0 unless noted):

Tier 2 (corroboration): TechCrunch, Search Engine Land, MediaPost, ppc.land (Google labels); eucrim, IAPP, Goodwin, Pinsent Masons (X fine); TheOptimizer, Admiral Media (TikTok CCL guides).

Verification note: built from a July 13, 2026 deep-research sweep — 25 sources, 123 claims extracted, 25 adversarially verified (23 confirmed, 2 refuted, 0 unverified). Verdicts and split votes preserved on-page; the two refuted framings are carried above as do-not-cite. Time-sensitivity: Google’s label rollout has no completion date, the EU AI Act Art. 50 deadline (Aug 2, 2026) may change label prevalence within weeks, and X’s remediation could change its repository state during 2026.

By Andrej Ruckij