Skip to content

Data Reference Overview

This section shows the data a REX browser extension collects, so you can judge whether it fits your project before you build anything. Each page covers one type of record and includes sample records with invented values.

  1. One record per observation, all in the same format. Each record carries the participant ID, the record type, and the time it was created in the browser. A server back end also records when it received it. Any record can be joined to survey or panel data on the participant ID.

  2. Each study chooses what to collect. REX is modular. A study turns on only the modules it needs, and each module has settings that control how much detail it records. Common record types include:

    • web browsing history (rex-history);
    • conversations with AI chatbots (rex-spider-chatgpt);
    • search engine results pages, including AI overviews and news results (rex-search-mirror);
    • search autocomplete suggestions (rex-search-mirror);
    • the browser’s default search engine (rex-browser-settings).
  3. Every browsing visit is kept, at one of three levels of detail. The researcher sets which sites are recorded at which level, using lists of sites:

    • full URL and page title;
    • website domain only;
    • a category label (for example health, banking or other) and the time of the visit, with no URL, title or domain.

    Because every visit is counted, researchers always know how much browsing took place. See rex-history.

  4. Free text can be redacted. Chatbot messages and search queries can pass through automated redaction in the browser before they leave it. Redaction replaces items such as email addresses with placeholders. Whether redaction is switched on is set per study.

  5. Search results show what the participant actually saw. Each result records its rank on the page, and news items record their position on screen. A foreground flag separates pages the participant viewed from comparison searches the extension ran out of sight, when that feature is enabled.

  6. Every record can be traced back to how it was collected. Records name the extension version and carry a hash of the study configuration in effect, so a researcher can tell which settings produced any record.

  7. Operational records support data-quality checks. Lifecycle events and upload status show when collection ran and finished and when the participant reached each study step. Researchers use them to spot participants whose data is incomplete.

A REX extension builds each record in the browser in the same format, whatever happens to it next. Each study chooses its back end, for example:

  • A server. The extension uploads records to a server such as Passive Data Kit (PDK), which stores each record with its own fields around it.
  • No server. With rex-local-download, the participant saves their records as a JSON file and hands it to the researcher. The file is a list of records, each with passive-data-metadata. See Server-Free Considerations.

The pages in this section document the record itself, which is the same across back ends. A back end can add fields: local download adds group and encrypted_transmission to passive-data-metadata, and a server adds its own fields around the record.

PDK stores each record as a data point:

{
"source": "participant-0001",
"generator_identifier": "rex-history-visit",
"secondary_identifier": null,
"generator": "rex-history-visit: Research Extension/1.0.0 Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/141.0.0.0 Safari/537.36",
"user_agent": "Research Extension/1.0.0 Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/141.0.0.0 Safari/537.36",
"created": "2026-07-30T14:01:00Z",
"recorded": "2026-07-30T14:16:00Z",
"server_generated": false,
"generated_at": null,
"properties": {
"name": "rex-history-visit",
"date": 1785420060000,
"url": "https://www.example.org/articles/garden-guide",
"domain": "example.org",
"title": "A Guide to Pollinator Gardens",
"passive-data-metadata": {
"source": "participant-0001",
"generator": "rex-history-visit: Research Extension/1.0.0 Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/141.0.0.0 Safari/537.36",
"generator-id": "rex-history-visit",
"timestamp": 1785420060,
"timezone": "America/Chicago",
"configuration-hash": "1f3c0e9a7b2d4c5e6f708192a3b4c5d6e7f8091a2b3c4d5e6f708192a3b4c5d6",
"enqueued-at": 1785420061
}
}
}
  • Top level: fields PDK stores for every record type: the participant (source), the record type (generator_identifier), the browser and extension version (user_agent), when the record was created in the browser (created), and when the server received it (recorded).
  • properties: the record as the extension sent it. Its fields depend on the record type. This is the part the other pages document; the example above is shortened.
  • properties.passive-data-metadata: fields REX attaches to every record. Exploring the Data explains them.

To export records from a PDK server, see Understanding PDK Data.

  • date and other *_time fields are milliseconds since the Unix epoch, in UTC. passive-data-metadata.timestamp is in seconds.
  • A field name ending in *, such as content*, holds free text. REX always runs this text through its redaction step, when redaction is enabled, before it leaves the browser.

One example of each common record type, with invented values. These are records as REX builds them, before any back end adds its own fields. 00 shows a complete record with its passive-data-metadata; to keep the others short, their passive-data-metadata points to the one in 00.

FileRecord typeWhat it records
00-record.jsonrex-history-visitA complete record, including the passive-data-metadata every record carries
01-rex-history-visit.jsonrex-history-visitBrowsing history visits at the three levels of detail
02-rex-conversation.jsonrex-conversationAn AI chatbot conversation, with its messages, citations and the web searches the chatbot ran
03-search-mirror-result.jsonsearch-mirror-resultOne organic result on a search results page, with its rank
04-search-mirror-result-ai.jsonsearch-mirror-result-aiThe AI overview block on a search results page
05-search-mirror-result-news.jsonsearch-mirror-result-newsOne news item on a search results page, with its position on screen
06-search-suggestions-result.jsonsearch-suggestions-resultAutocomplete suggestions for a partly typed query
07-rex-browser-settings-search-engine.jsonrex-browser-settings-search-engineThe browser’s default search engine
08-pdk-app-event.jsonpdk-app-eventExtension lifecycle events
09-pdk-system-status.jsonpdk-system-statusUpload-queue length and basic device information