This repo contains data for our story "How California sent residents’ personal health data to LinkedIn".
The Markup and CalMatters used The Markup's Blacklight tool to scan more than 200 California state and county government websites that offer services for immigrants lacking permanent legal status. We were looking for sites that shared personally identifying information through ad trackers and third-party cookies. The scan revealed that Covered California, the state's health insurance marketplace, had 20 times the average number of ad trackers on the sites we scanned, and 50 times the average number of third-party cookies.
We took a closer look, navigating the site as we watched activity in the browser's network panel. As we filled forms on the site, we noticed that every click on a form element initiated a network request to https://px.ads.linkedin.com/wa/, an address associated with LinkedIn's Insight Tag. Included in the request was a domAttributes parameter, which contained information about the element on the page that we'd clicked, including its full text. This information was sent to LinkedIn each time a form element was clicked, before the form was submitted.
Below is a catalog of the data gathered while investigating this story.
These Blacklight scans demonstrate the state of tracking on the Covered California website when we started investigating it, and how it changed in the weeks afterwards.
- February 25, 2025 - A Blacklight scan showing 63 ad trackers, 121 third-party cookies, and the Meta Pixel.
- April 4, 2025 - A Blacklight scan showing 6 ad trackers, 4 third-party cookies, and the Meta Pixel.
- April 21, 2025 - A Blacklight scan showing 3 ad trackers.
As we investigated the site, we saved HAR files, which document network activity. HAR files can be read in many modern browsers and with the HAR Analyzer tool.
The network traffic captured in these HAR files was generated as a Markup reporter investigated the site, entering made-up data about people that don't exist. Personally identifying information has been redacted from the HAR files. Any unredacted information that appears to be personally identifying (e.g. names, birth dates, marital status, ethnicity, race, gender) is not real.
Each HAR file is accompanied by a directory of screenshots highlighting significant events, many are linked below. Screenshots are listed in the order they occurred for that particular browsing session; bold text indicates types or values of information appearing for the first time.
Shows the following information being sent to LinkedIn:
- That a member of the applying household is pregnant
- That a member of the applying household is blind or disabled
- That the applicant prefers a plan with a lower premium
- How often members of the household plan to visit the doctor for tests and treatment
- How often members of the household plan to use a pharmacy
- The full name, specialty, and office address for the applicant's choice of doctor
Shows the following information being sent to LinkedIn:
- That a member of the applying household is blind or disabled
- That a member of the applying household is pregnant
- That the applicant prefers a plan with a lower premium
- How often members of the household plan to visit the doctor for surgery or outpatient treatment and tests
- How often members of the household plan to use a pharmacy
- The applicant's choice of hospital
Shows the following information being sent to LinkedIn:
- The applicant opting to create an account
- The applicant agreeing to terms and conditions
- The applicant disclosing how they found coveredca.com (internet search)
- The applicant agreeing to have their information verified
- A household member's gender (male)
- A household member's marital status (married)
- A household member's preferred language (English)
- A household member's ethnicity: one, two (Hispanic/Latino/Spanish)
- A household member's gender (transgender)
- A household member's relationship to the head of household (spouse/partner)
Shows the following information being sent to LinkedIn:
- That a member of the applying household is pregnant
- That a member of the applying household is blind or disabled
- That the applicant prefers a plan with a lower premium
- How often members of the household plan to visit the doctor for tests and treatment
- How often members of the household plan to use a pharmacy
- The full name, specialty, and office address for three doctors the applicant chose: first, second, third
- The applicant agreeing to have their information verified
Shows the following information being sent to LinkedIn:
- The applicant disclosing how they found coveredca.com (internet search)
- The applicant agreeing to have their information verified
- A household member's marital status (divorced)
- A household member's preferred language (Spanish)
- A household member's ethnicity: one, two (Hispanic/Latino/Spanish)
- A household member's race (White)
- A household member's gender (female)
- A household member's marital status (single)
- A household member's preferred language (English)
- A household member's relationship to the head of household (sibling)
- A household member's gender (male)
- A household member's marital status (widowed)
- A household member's relationship to the head of household (parent/guardian)
Shows the following information being sent to LinkedIn:
- The applicant indicating that they qualify for special enrollment (victim of domestic abuse or spousal abandonment)
Screenshots of the coveredca.com pages showing form elements that generated the network traffic captured in the HAR files. These screenshots were not taken at the same time that the network logs were captured.