Skip to main content
Tracked coverage is where you save specific published pages—such as articles, interviews, podcasts, broadcasts or PR placements—and measure whether those exact pages appear in answers collected by Chatobserver. It is different from source-domain tracking, which classifies an entire cited domain, and group coverage, which describes how prompts are organised.

Add one published URL

Only the complete published URL is required. You can also provide:
  • title;
  • publisher;
  • publication date;
  • format: article, interview, podcast, broadcast or other;
  • attribution: your team, external or not yet classified; and
  • internal notes.
Unknown details can remain empty. The assistant should not open the URL or invent metadata merely to fill the form. Example:
In Clean Digital, check whether https://example.com/research/story is already in Tracked coverage. If it is new, add it as an externally earned article. Leave the title, publisher and publication date empty.
The assistant checks the exact page first, then asks for approval before saving it. Saving a URL does not create historical monitoring data. Evidence appears only when that page is observed in collected answers.

Review a list or CSV before adding it

Give the assistant a list or attach a CSV and ask it to review the rows first:
Review these published URLs against active and archived Tracked coverage. Flag exact duplicates, invalid URLs and missing optional details. Return a corrected table, but do not add anything until I approve it.
The connector can check up to 50 URLs in one duplicate search. For a larger file, the assistant can review it in batches. After approval it adds each intended record with its own safe retry identity and reports the result for every row.

Edit publication details

Ask the assistant to read the record before editing it. Edits use the current record version, which prevents a stale request from silently overwriting somebody else’s change. Example:
Find the Tracked coverage record for this URL and show its current details. Change the publisher to Example News and the publication date to 2026-09-18, preserving every other field.
If the record changed after it was read, the assistant must retrieve the latest version and show you the conflict rather than force the edit through.

Archive or restore a record

Archiving removes a record from the active list without deleting its history or evidence. Restoring makes it active again. Example:
Show me the Tracked coverage record and its evidence first. If this is the duplicate copy, archive it but keep its history.
The assistant should ask before archiving. Active and archived records can both be searched so an old URL is not accidentally added again.

Measure coverage and supporting evidence

The assistant can analyse a period of up to 366 days and compare it with the immediately preceding period of the same length. Results can include:
  • answers assessed and answers citing saved coverage;
  • citation rate and brand visibility rate;
  • daily or weekly trends;
  • matched coverage records;
  • comparison with the preceding period; and
  • the supporting answers and exact source URLs.
You can filter by attribution, one coverage record, publisher or URL text, root domain, prompt, group, platform, region, or active/archived state. Example:
For externally earned Tracked coverage in the last 30 days, compare citations and brand visibility with the preceding 30 days. Show the matched articles and supporting answers, and keep the reporting filters and sample sizes visible.
Tracked coverage reports observed associations. A page appearing in an answer does not prove that the publication caused a later visibility, ranking or sentiment change.

Permissions and availability

Reading Tracked coverage and changing it are separate connection permissions. Reconnect Chatobserver if an older connection does not offer them. Your current workspace role and the availability of Tracked coverage in the Chatobserver Sources area still apply. Continue with sources and sentiment, usage and safe changes, or example requests.