Field guide
What ad-library data can and cannot tell you
Field by field, what came back on 1,600 ads across ten routes and four platforms: creatives, dates, placements, status, landing pages, and the fields people assume exist and do not, such as spend, targeting and performance.
Written for people asking: Can I get competitor ad spend from the Meta ad library API?What data does the Google Ads Transparency Center actually show?Do ad library APIs show impressions or targeting?How are these ads still running after 12 months in the Meta ad library?How do you spot a winning ad in the Meta Ads Library by how long it has run?
Across 1,600 ads returned by ten ad-library routes on 15 September 2026, every route returned the creative and the platform, and most returned start and end dates and a landing page. No route on any platform returned spend for ordinary commercial ads: 0 of 200 on Apify Meta, 0 of 594 on ScrapeCreators Meta, 0 of 754 on Google, 0 of 240 on LinkedIn, 0 of 176 on TikTok. Impressions exist only as ranges, on 20% of LinkedIn ads (EU delivery) and on TikTok as an audience-size bucket. Google returns first and last shown dates and format but no active status; LinkedIn returns dates on 20% of ads. Anything a vendor sells as competitor spend, targeting or performance from a public library is inferred, not read.
How the fields were measured
Each ad returned in the cohort run was checked for nine canonical fields, mapped from each provider’s own field names: the creative asset (image, video or card), start and end dates, the platform or placement, an active status, spend, reach or impressions, a landing page, and the ad text. A field counts as present when it is non-empty for that ad. The table shows, per route, the share of returned ads carrying each field.
| Route | Ads | Creative | Start date | End date | Placement | Status | Spend | Reach / impressions | Landing page | Ad text |
|---|---|---|---|---|---|---|---|---|---|---|
| ScrapeCreators · Google ads by domain | 369 | 100% | 100% | 100% | 100% | 0% | 0% | 0% | 100% | 0% |
| ScrapeCreators · Google advertiser lookup | 62 | 0% | 0% | 0% | 0% | 0% | 0% | 100% | 0% | 100% |
| SerpApi · Google Ads Transparency | 369 | 100% | 100% | 100% | 100% | 0% | 0% | 0% | 100% | 0% |
| ScrapeCreators · LinkedIn ad library | 240 | 87% | 13% | 13% | 100% | 0% | 0% | 13% | 3% | 99% |
| Apify · Meta Ads Scraper | 180 | 100% | 100% | 100% | 100% | 100% | 0% | 0% | 98% | 99% |
| ScrapeCreators · Meta, by page name | 298 | 100% | 100% | 100% | 100% | 100% | 0% | 0% | 99% | 100% |
| ScrapeCreators · Meta, keyword search | 237 | 100% | 100% | 100% | 100% | 100% | 0% | 0% | 98% | 100% |
| Apify · TikTok Ads Library | 68 | 97% | 0% | 0% | 100% | 100% | 0% | 100% | 0% | 0% |
| ScrapeCreators · TikTok ad library | 68 | 97% | 100% | 100% | 100% | 100% | 0% | 100% | 100% | 100% |
| TikHub · TikTok Creative Center top ads | 0 | — | — | — | — | — | — | — | — | — |
What you can rely on
- The creative itself. Every route returned it, as an image or video URL on the platform’s CDN. Those URLs expire: on Meta the link carries its own expiry, about 106 hours after the fetch for ordinary ad images and most videos (measured on 10,355 links), so a monitoring job should store the asset, not the link.
- When the ad ran. Meta and Google return first and last seen dates on every ad. TikTok returns them through ScrapeCreators but not through Apify’s actor. LinkedIn returned dates on only a fifth of ads.
- Where it ran. Meta returns publisher platforms (Facebook, Instagram, Audience Network, Messenger) per ad. Google returns the format (text, image, video) and, on request, the surface (Search, YouTube, Display, Maps, Play). TikTok and LinkedIn are single-platform.
- Whether it is live. Meta and TikTok return an active flag. Google does not; you infer it from the last shown date. LinkedIn does not.
- Where it points. Landing page on every Meta and Google ad, on every TikTok ad through ScrapeCreators, and on only 7% of LinkedIn ads, which return a destination URL only for some formats.
- The copy. Meta, TikTok and LinkedIn return body text or headline. Google returns none: the Transparency Center shows a rendered creative, not the text.
What you cannot get
- Spend. Zero ads out of 1,600 carried a spend value. Meta publishes spend and impressions for political and issue ads and, under EU rules, reach for ads delivered in the EU; commercial ads elsewhere carry neither. Google shows none for commercial ads. TikTok’s
spentfield came back empty on every ad. LinkedIn returns no spend. Any “competitor spend” figure built on a public library is a model, and it should be labelled as one. - Impressions. LinkedIn returned a range such as “1k-5k” on 20% of ads, those with EU delivery, with a per-country breakdown. TikTok returned an audience-size bucket such as “0-1K”, which is an estimate of audience, not impressions, and an
impressioncount of 0 on every ad. Meta and Google returned nothing. - Targeting. Not returned by any route on any platform for commercial ads. Meta exposes EU targeting criteria only through per-ad detail calls on EU-delivered ads; none of the cohort’s first-page ads had it.
- Performance. No clicks, conversions or results anywhere. TikHub’s Creative Center route ranks ads by TikTok’s own performance tiers, but it is a curated showcase you search by keyword, and it returned nothing for any of the ten brand names.
- A complete list. Every route paginates. First pages are 20 to 40 ads; monday.com had 667 active Meta ads and Shopify 4,986 on 15 September 2026, and Google estimated 80,000 for Shopify. “All of a competitor’s ads” is a paginated job, not a call.
Field names by provider
For anyone mapping these into a schema, the source fields behind the canonical ones above:
| Canonical | ScrapeCreators Meta | Apify Meta | ScrapeCreators Google | SerpApi Google | ScrapeCreators TikTok | ScrapeCreators LinkedIn |
|---|---|---|---|---|---|---|
| Creative | snapshot.images / videos / cards |
snapshot.images / videos |
imageUrl, adUrl |
image, details_link |
videos, image_urls |
image, video, carouselImages |
| Dates | start_date, end_date |
startDate, endDate |
firstShown, lastShown |
first_shown, last_shown |
first_shown_date, last_shown_date |
startDate, endDate |
| Placement | publisher_platform |
publisherPlatform |
format |
format |
— | — |
| Status | is_active |
isActive |
— | — | audit_status |
— |
| Spend | spend (null) |
spend (null) |
— | — | spent (empty) |
— |
| Reach | reach_estimate (null) |
impressionsWithIndex.impressionsText (null) |
— | — | estimated_audience |
totalImpressions, impressionsByCountry |
| Landing | snapshot.link_url |
snapshot.link_url |
domain |
target_domain |
url |
destinationUrl, landingPage |
| Text | snapshot.body |
snapshot.body |
— | — | title |
headline, description |
Meta copy comes back in parts, and one message runs under many ids
Both Meta routes return the copy as separate fields rather than one block, so a table with one column each for primary text, headline, call-to-action label and landing URL needs no parsing. Counted on the distinct ads in the 21 September 2026 receipts, on the ad or on its cards (carousel cards, catalogue cards, or the versions of a multi-version ad):
| Field | ScrapeCreators by page name (298 ads) | ScrapeCreators by keyword (289 ads) | Apify (195 ads) |
|---|---|---|---|
Primary text, snapshot.body.text |
288 | 285 | 194 |
Headline, snapshot.title |
287 | 263 | 174 |
Call-to-action label, snapshot.cta_text / ctaText |
297 | 287 | 193 |
Landing URL, snapshot.link_url / linkUrl |
297 | 286 | 192 |
Link description, snapshot.link_description / linkDescription |
170 | 178 | 155 |
Display caption, snapshot.caption |
297 | 285 | 192 |
Two things to plan for. On Apify, 71 of the 195 ads carried an unfilled placeholder such as {{product.name}} in the primary text or headline: these are catalogue ads whose copy is filled per viewer, and the library returns the template. And the ad id does not identify a unit of creative. Leaving the placeholder ads aside, the 298 ids from ScrapeCreators by page name held 149 distinct combinations of primary text, headline and landing path; 198 of the ads shared their combination with at least one other id, and the largest set was 22 ids. On Apify it was 41 of 195 ads in 13 sets, the largest 15. This compares the message at the ad level only. Two ids with the same message can still carry different images or cards, so a dedupe key needs the asset as well.
The counts and the sets of ids are in summary.json and groups.csv; the script reads the published receipts and makes no provider call.
How long the ads have been running
Every Meta ad carries start_date, so an ad’s running time at the moment you fetch it is one subtraction. On the 19 October 2026 cohort run, 472 active ads sat on the ten fixture advertisers’ own pages once Apify’s page scrape, ScrapeCreators’ company endpoint and the ads from its keyword search that belonged to the same page were deduplicated on ad id. Every one of them was flagged active, so these are live ads at the run date.
| Running for at least | Active ads | Share of 472 |
|---|---|---|
| 30 days | 359 | 76% |
| 60 days | 265 | 56% |
| 90 days | 208 | 44% |
| 180 days | 57 | 12% |
| 365 days | 27 | 6% |
- Over a year is rare but real. Canva had 9 of its 47 active ads past a year, Shopify 6 of 55, ClickUp 4 of 50; HubSpot, monday.com, Notion and Semrush had one or two each; Grammarly and Apollo.io none. The oldest active ad had started 1,614 days earlier (monday.com), and five of the ten advertisers had an ad over 800 days old still live.
- A first page is not a sample. All 27 year-old ads came through Apify’s page scrape or the keyword search; the company endpoint’s first 30 ads per advertiser contained none older than a year for any of the ten. To say how many of an advertiser’s active ads are older than a year, page through that page id’s whole active library and compute the age of each; the routes return dates, not a count.
- Where a threshold starts to mean something. Three in four active ads were already past 30 days, so a 7 or 14 day rule separates nothing for advertisers of this size; 90 days selected under half, 180 days one in eight. Running time is also the only public signal of performance here, since no route returns spend or results (see above).
Searching the advertiser’s name as a keyword through ScrapeCreators returned 29 or 30 ads from 4 to 20 distinct pages, and the named advertiser’s own page accounted for between 0 (ClickUp) and 25 (Grammarly) of them. That is the shape of a “how many advertisers are selling this” count: distinct page ids in a keyword search, with the caveat that it is one page of results. It also means a brand-name keyword search is not an advertiser lookup; fetch by page id for that.
The per-ad ages are in ages.csv and the counts in summary.json; the script reads the published 19 October receipts and makes no provider call.
Which pages and countries the ads point at
Two questions the fields answer only indirectly, both read from the 2 November 2026 cohort receipts with no new call.
Every ad pointing at one landing page. No Meta route checked takes a destination URL as input (Adyntel’s and Crawl4AI’s take a keyword and a country), so this is a filter over linkUrl or link_url after the pull. On nine advertisers’ own pages, 293 of 298 ads carried a landing URL, on 76 distinct pages with query strings removed. Concentration varies: Apollo.io had 23 of 30 ads on one sign-up page, HubSpot 5 of 35 on its most used page out of 18. The tenth advertiser, ClickUp, had no ads on its own page in these responses. Notion’s most common destination was a fb.me short link (26 of 30), so resolve short links before grouping. This counts first-page responses, not a whole library.
Which countries an ad runs in. None of the 735 Meta ads carried a country: targetedOrReachedCountries and countryIsoCode were empty on all 200 from Apify, and country_iso_code on all 535 from ScrapeCreators. The Apify requests carry country=US in the library URL, and the catalog’s Meta routes list country as an input. So a tool’s country view depends on which filters it sent; a per-country answer means one query per country, and these receipts do not hold that comparison. LinkedIn returned impressionsByCountry on 24 of 240 ads.
Counts are in summary.json and landing-pages.csv; the script reads the published receipts.
What this run cannot tell you
Ten advertisers, first page each, mostly US-targeted. A cohort of EU advertisers would show Meta’s EU reach and targeting fields populated on more ads, and LinkedIn’s impression ranges on more than a fifth. The absence of spend is structural, not a sample effect: the platforms do not publish it for commercial ads through any route.
Next step
- Raw responses for every ad in the table are on the evidence page, one file per call.
- The choose an API page compares the routes on cost and coverage; the Meta and Google pages cover the official access routes.
Questions this page answers
Can I get competitor ad spend from the Meta ad library API?
No. Across 1,600 ads from ten routes on 15 September 2026, zero carried a spend value; Meta publishes spend only for political and issue ads. Any competitor-spend figure built from a public library is a model and should be labelled one.
What data does the Google Ads Transparency Center actually show?
Per ad: the creative, its format, first and last shown dates and the advertiser's domain; per advertiser: verified name, payer name and a library-size estimate. No ad text, no active status, no spend, no impressions for commercial ads.
Do ad library APIs show impressions or targeting?
Only in narrow cases: LinkedIn returned impression ranges on 20% of ads (those with EU delivery), TikTok returned an audience-size bucket rather than impressions, and Meta exposes targeting only for EU-delivered ads through per-ad detail calls. Nothing returned targeting for the cohort's US-targeted ads.
How should I store Meta ad library copy, and can I dedupe ads by ad id?
The copy arrives already split, so store it that way: on 298 Meta ads returned by ScrapeCreators on 21 September 2026, primary text was its own field on 288, headline on 287, call-to-action label on 297 and landing URL on 297. Ad id is a weak dedupe key: those 298 ids held 149 distinct combinations of primary text, headline and landing path, and one combination ran under 22 ids.
Can you tell how long a Meta ad has been running, and how many of a big advertiser's active ads are older than a year?
Yes to the first: every Meta ad carries start_date, so its running time is the fetch date minus that. On the 19 October 2026 cohort run, 27 of 472 active ads (6%) on the ten advertisers' own pages had started more than a year earlier, the oldest 1,614 days; Canva had 9, Shopify 6, Grammarly and Apollo.io none. The count for a whole library needs every page of that advertiser's active ads: the first 30 from ScrapeCreators' company endpoint held none over a year for any of the ten, while Apify's page scrape and the keyword search between them returned all 27.
Which ad library fields tell you an ad has run long enough to matter, and how do you count the advertisers behind one product?
Running time is start_date against today, and the active flag says whether it is still live; no field gives spend or results. On the 19 October 2026 cohort, 76% of 472 active ads were past 30 days, 44% past 90 and 12% past 180, so a 7 or 14 day threshold separates nothing for advertisers of this size. To count advertisers, run a keyword search and count distinct page ids: a search for each cohort brand name returned 29 or 30 ads from 4 to 20 pages, and the brand's own page accounted for between 0 and 25 of them.
Can you get every ad pointing at one landing page?
Not as a query: the Meta routes checked take a keyword, page or country, not a destination. It is a post-filter on the landing URL field after pulling an advertiser's ads. On the 2 November 2026 cohort, 293 of 298 ads on nine advertisers' own pages carried one, pointing at 76 distinct pages once query strings were removed, and 44% of the ads (128) sat on each advertiser's single most used page.
Which countries is a competitor's Meta ad running in, and why do tools show different ads per country?
The ad record does not say. On the 2 November 2026 cohort, none of 735 Meta ads across three routes carried a populated country field, including Apify's targetedOrReachedCountries. Country is a filter you send with the request, so a per-country view means one query per country, and tools that query different countries return different ads. LinkedIn is the exception: 24 of 240 ads carried a per-country impression breakdown.
Spotted an error or a changed price? Email corrections@adapiguide.com with the page URL and a source. Corrections are logged on the corrections page.