# Changelog of Microsoft Ads Library Scraper: Bing Ads, EU DSA Repository (`fetchwright/microsoft-ads-library-scraper`) Actor

- **URL**: https://apify.com/fetchwright/microsoft-ads-library-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/fetchwright/microsoft-ads-library-scraper.md

## Changelog

### 2026-11-16

- README: a short note and Apify's video on running it on a schedule.
- README: related actors updated (same-topic links first, current names).

### 0.1.38 - 2026-11-15

- Price comparison updated with the rivals' price cards of 15 November 2026: the three most-used actors charge $0.50 to $3.00 per 1,000 ads on the Free plan (this Actor $0.40); one less-used actor charges less per ad and is named in the comparison.
- A run that reads ad details says so up front, with a measured example (one run, Booking.com B.V., Germany, 100 ads, 15 November 2026: about 37 minutes with details, about 2 minutes without); advertiser summary runs, which read no details, do not show it. The README answers "How fast is it?" with the same measurement.

### 0.1.35 - 2026-11-15

- Default memory is now 128 MB (was 256 MB): measured peaks stay under 70 MB in the last 25 runs, including runs with details, so the same run costs about half the platform usage. You can still choose more memory.&#x20;
- Microsoft runs with details are slower: allow about 20 seconds per Microsoft request before source response time and any slowdown. After an eligible ordinary rate-limit pause, subsequent reads use longer minimum intervals; refusal handling is unchanged. This pacing does not guarantee a quota or completion.&#x20;
- A run that sets a maximum charge no longer stops one ad early after a dataset write whose answer was lost: only charges the platform made but the run did not count are held back. A run never exceeds its maximum charge.&#x20;

### 0.1.34 - 2026-11-14

- New fields on every ad row: `displayDomain` and `destinationDomain` (the websites shown and linked to, without www.), `advertiserUrl` (the advertiser's page in Microsoft's Ad Library) and `hasImages` (whether Microsoft lists any image for the ad; Microsoft states no ad format, so none is invented).
- Look up one ad: an ad's Ad Library link or `ad:<id>` under Advertisers saves that ad with all its details for one ad's price. An unknown ID is listed in the run summary and not charged.
- No change in price.

### 0.1.33 - 2026-11-14

- The listing title now names Microsoft's EU DSA ad repository, backed by the README's sourced sentence (0.1.32). The listing and README opening state the Free-plan price and a paid-plan price and name the only-new watch; the README opening also names the advertiser summary (one row per advertiser with Microsoft's ad count). No change in behaviour or price.

### 0.1.32 - 2026-11-13

- README: one sentence, with Microsoft's own sources, that its Ad Library is Microsoft's EU advertising repository under the Digital Services Act and holds only ads served within the EEA. The Output tab's Overview table no longer shows first and last shown dates and the impression range, which stay empty unless details are on; the Details table shows them. No change to the Actor, its input or the data it saves.

### 0.1.31 - 2026-11-13

- README: a "Use it in a pipeline" section with one step-by-step recipe for Zapier (new Bing ads into Google Sheets), using Apify's own integration and your own Apify account. It says what it was checked against and that it was not run in a live account. New FAQ answer on downloading the results as CSV, Excel or Google Sheets. No change to the Actor, its input or its output.

### 0.1.30 - 2026-11-13

- Reliability: the impressions range of an ad is read in one pass, so a malformed range in Microsoft's answer can no longer slow a run down. Results are unchanged.

### 0.1.27 - 2026-11-13

- Default memory 256 MB (was 512). Measured on a private copy: small runs gave the same rows at both sizes for about half the compute; a 1,600-ad run with the country split peaked at 67 MB at 512 MB, and a 256 MB run of the same input peaked at 70 MB before a Microsoft refusal ended it. The README's start-fee sentence names the new default.

### 0.1.26 - 2026-11-12

- Field descriptions: `url` is the ad's page on Microsoft's Ad Library website, and with the country split `searchRegion` carries the country whose list a row came from.

### 0.1.25 - 2026-11-12

- README: `url` is the ad's page on Microsoft's Ad Library website (as rows have given since 0.1.10); an older sentence said no such page was given.

### 0.1.24 - 2026-11-12

- README: reading country by country is likely to stop early, because Microsoft answers some requests with a page saying the request is blocked, which this Actor treats as final; figures from our own runs added.

### 0.1.23 - 2026-11-12

- A page saying the request is blocked now ends the run's Microsoft requests at once under every HTTP status, 429 included (before, under 429 it was paused out and asked again).
- Advertisers and keyword searches can read past Microsoft's limit of about 1,000 ads: with **Read past the 1,000-ad limit country by country** on, a list that reaches the limit is read again for each EEA country, largest markets first, up to your maximum per advertiser or keyword, the run-wide maximum, your maximum charge and the run time; while reading country by country any HTTP 429 or refusal ends the run's Microsoft requests at once, keeping what was read and listing the rest as unfinished. It needs a maximum per advertiser or keyword above 1,000 (now allowed up to 31,000). An ad found in several countries is saved and charged once; each row says which country's list it came from, and the run summary says what each country added. Off by default.

### 0.1.22 - 2026-11-12

- Maintenance build: shared code updated. No change in behaviour for this Actor.

### 0.1.21 - 2026-11-12

- Maintenance build: shared code updated. No change in behaviour for this Actor.

### 0.1.20 - 2026-11-12

- New result type, Advertiser summary: one row per advertiser you list, with its country, verified badge and Microsoft's own count of its ads under your country, date and keyword filters, capped at 1,000 (`adCountAtLeast` marks 1,000 or more). One lookup per advertiser and no ad rows; each advertiser row is charged like one ad, also for an advertiser with 0 ads. First and last shown dates are not on these rows.

### 0.1.19 - 2026-11-12

- Maintenance build: shared code updated. No change in behaviour for this Actor.

### 0.1.18 - 2026-11-12

- New: Only ads not returned in earlier runs. For the advertisers you list, a scheduled run returns and charges only ads it has not returned before, with one run of a watch at a time; the first run can return current ads or only remember them. A watch run stops once it has returned your maximum; new ads it has not reached come in later runs. An ad whose save could not be confirmed is checked against that run's results in the next run, so it is never returned or charged twice and never lost. Another actor's `monitorMode` turns it on.

### 0.1.17 - 2026-11-12

- Maintenance build: shared code updated. No change in behaviour for this Actor.

### 0.1.16 - 2026-11-12

- A row is never saved or charged twice when the answer to saving it is lost: the run checks its own dataset, counts a row it finds once (and against your maximum charge), and stops at a row it cannot confirm, listing what it did not finish. The run summary describes it under `uncertainWrite`.

### 0.1.15 - 2026-11-12

- If the platform restarts a run (for example to move it to another server), the run now asks Microsoft nothing more and saves nothing more, so no ad is saved or charged twice. Its status message and summary say how many ads it saved before the restart; start it again for the rest.
- A verification or CAPTCHA page from Microsoft now ends the run's requests at once whatever status it comes with, a rate limit included, and so does a block page that comes with a timeout status. Microsoft's ordinary rate-limit answer still makes the run pause and continue, and an advertiser or ad whose name or wording mentions such words is read as data, never taken for a refusal.
- README: the list of Microsoft answers that end a run's requests is now complete and exact.

### 0.1.14 - 2026-11-12

- README: the Python example now matches the current Apify Python client.

### 0.1.13 - 2026-11-11

- Keyword searches without advertisers now fill `advertiserCountry` and `advertiserVerified` for the first 10 different advertisers they find, at no extra charge. A new input, "Advertiser country and verification lookups", sets that number (0 to 100; 0 turns it off). Each advertiser adds about eight seconds to the run.
- Advertisers beyond that number, or found after the run ran short of time or Microsoft limited or refused it, keep both fields empty; the run summary counts them under `advertiserIdentity` with the reason. Rows are saved as before and the maximum charge holds as before.

### 0.1.12 - 2026-11-11

- Maintenance build: shared code updated. No change in behaviour for this Actor.

### 0.1.11 - 2026-11-11

- Offers: price guarantees are now found as `guarantee` with value `best price`, `lowest price` or `price` ("Best price guaranteed", "Mejor precio garantizado", "Precio mínimo garantizado", "Bestpreisgarantie", "Preisgarantie", "Meilleur prix garanti", "Laagste prijs gegarandeerd", "Miglior prezzo garantito").
- Offers: a new type, `free_cancellation` ("Free cancellation", "Kostenlose Stornierung", "kostenlos stornierbar", "Cancelación gratis", "Annulation gratuite", "Gratis annuleren", "Cancellazione gratuita").
- Offers: a price cut stated without a figure ("Reduced prices", "Prix cassés", "Precios rebajados", "Reduzierte Preise", "Prijzen verlaagd", "Prezzi ridotti") is a `discount` with value null. "Deals", "Angebote" or "save" on their own are still not offers, and a satisfaction figure such as "98% Kundenzufriedenheit" is not a discount.
- These are found in English, Spanish, German, French and Dutch, and in Italian, by fixed word rules, so an offer worded another way is still missed; a negated one ("Keine Preisgarantie", "No free cancellation") is not listed. Offers found before are unchanged. In a sample of 644 travel ads from 11 November 2026, 181 now state at least one offer, against 5 before.

### 0.1.10 - 2026-11-11

- Each row's `url` now opens the ad's page in Microsoft's Ad Library instead of a data address.

### 0.1.9 - 2026-11-11

- README: what keyword-search rows leave empty, what a country filter selects, and a second measurement of country-filtered lists. Four example tasks with real run times and costs are on the listing.

### 0.1.0 - 2026-11-11

- First version: Bing ads from Microsoft's Ad Library for the European Economic Area, by advertiser name, numeric ID or Ad Library link, or by keyword in ad text. Optional countries, date window, keyword filter, and dates, impression range, share per country, targeting and payer at no extra charge. One row per ad at $0.40 per 1,000 ads on the Free plan, plus Apify's start event per run ($0.00005 up to 1 GB of memory); failed and empty lookups give no ad charge. A hard cap per advertiser or keyword that holds even when an advertiser is entered twice, a coverage verdict for each in the run summary, failures counted by reason, offers stated in each ad's wording (English, German, French, Spanish and Dutch word rules; negated offers are not counted), and alerts to Slack, Discord or a webhook.
- Input written for other Microsoft ad library actors is read as this Actor's own; their run-wide limits (`maxItems`, `maxResults`, `maxSearchResults`) stay run-wide, and the README states which field wins when a field is given both ways.
- When Microsoft limits how often it is asked, the run pauses, slows down and continues. When Microsoft refuses a request, or keeps limiting the run after two pauses, or the run has too little time left for a pause, the run asks Microsoft nothing more; the rest is listed as failed or unfinished in the run summary and what was not read is not charged. A country code that is not a real country stops the run before any request. A country-filtered list that Microsoft cannot answer within 60 seconds is reported with what to do. The run summary is saved before the alert, and requests and the alert end before the run's timeout.
