Ad ObserverBy NYU Cybersecurity for Democracy

Ad Observer public database

What is Ad Observer?

Ad Observer, a project of NYU Cybersecurity for Democracy, is a tool you add to your web browser. It copies the ads you see on Meta and YouTube, so anyone can see them in our public database. If you want, you can enter basic demographic information about yourself in the tool to help improve our understanding of why advertisers targeted you. However, we’ll never ask for information that could identify you. (For more information, see Ad Observer.)

What data am I seeing in the public database?

Daily Summary of Political Ads

Data files are generated daily, and include all political ads that Ad Observer collected that day. We publish files on a weekly basis, usually on Wednesdays. These files include all ads that include a political ad disclaimer, signifying that the advertiser has declared ad as a “social issue, elections, or politics” ad. Ads will appear in multiple files as users are able to see the same ad on multiple days.

Monolith file of all political ads

This file includes all political ads collected by Ad Observer. All ads included in daily summary files are in this file.  

What questions can I ask these data?

Ad Observer data comes from volunteers, and therefore do not comprise a representative sample of Meta users. Therefore, these data cannot be used to extrapolate trends or distribution of ads to all Meta users, such as how much was spent on an ad campaign overall, or by a particular ad sponsor. However, these data can be used to answer questions specific to a particular ad such as:

  • What is the creative content of the ad?
  • How was this specific ad targeted?
  • Who is the advertiser paying for this ad?

What information is available?

Most of the fields in these databases were generated by Meta, similar to data fields available in the company’s ad library API, but some are enhancements created by NYU Cybersecurity for Democracy (C4D). In the field descriptions below, bold marks fields created by C4D.

Questions? Contact us at [email protected].

Field descriptions

Daily Summary of Political Ads

  • Ad_id. Unique ID generated by Meta for this ad. (There may be multiple duplicate ads with the same content, but different ids.)
  • Page_name. Page name of the entity which published the ad.
  • Political_value. A number ranging from 0-1, populated by C4D’s political classification model. Ads that Meta indicates are political (because they have disclosures) have a value of -1, as C4D’s political classifier did not run on these ads.
  • Paid_for_by. Name of the entity listed in the ad disclosure. This field will be empty if the ad does not have a disclosure (that is, it’s not a declared political ad).
  • Ad_text. Text content of the ad.
  • Observed_at. Date/time when ad was observed.
  • Call_to_action. Text on button in ad, if there is one. (This might be donate, subscribe, etc.)
  • Targetings. Jsn object containing all of the targeting information collected for this ad. A description of the specific fields in this json object are shown below.

    Monolith file of all political ads All fields are the same as Daily Summary of Ads, with addition of two fields:
    undefinedundefined

    Targeting fields: These include information collected by Ad Observer on how advertisers targeted a particular ad. Many of these terms are generated by Meta, such as those related to “custom audience” and “collaborative ads.” Targeting information may include:
  • Age_gender.
  • Interests.
  • Connection. (Page/event/app).
  • Edu_schools. Ids for schools targeted in the ad.
  •  Locale. Country and language combination that advertiser chose for ad targeting, such as en-US, en-UK, etc.
  • Custom_audiences_engagement_ig.
  • Custom_audiences_website.
  • Collaborative_ads_store_visits.
  • Relationship_status.
  • Friends_of_connection.
  • Work_job_titles.
  • Custom_audiences_datafile.
  • Custom_audiences_lookalike.
  • Work_employers.
  • Location. This can be state, region, city, country.
  • Custom_audiences_engagement_page.
  • Collaborative_ads_store_sales.
  • Custom_audiences_offline.
  • Custom_audiences_engagement_video.
  • Bct. "Broad Category Targeting." Examples include: "Frequent international travelers," "Likely engagement with US political content (liberal)," "Engaged shoppers," "Household income: top 10%-25% of ZIP codes (US)."
  • Actionable_insights.
  • Custom_audiences_mobile_app.
  • custom_audiences_engagement_canvas.
  • Custom_audiences_engagement_lead_gen.
  • Dynamic Role.
  • Collaborative_ad.
  • Ed_status. Examples: EDU_DOCTORATE_DEGREE, EDU_COLLEGE_ALUMNUS, EDU_MASTER_DEGREE.
  • custom_audiences_engagement_event