Skip to main content
Track how a person, company or entity is being talked about on podcasts. It answers “which shows mentioned Nike this week versus last?” and “is it coming up more often?” in one call each, from Particle’s own history, so an agent keeps only the slug between sessions and asks again for the window it wants. The tool has two modes:
  • compare (default): which podcasts mentioned the subject in a window and how each compares with the equal-length window just before it. Each show has its episode and mention line counts in both windows, the change in mention lines, and whether it is new to the subject, continuing, or dropped it. Totals for both windows count every podcast, whatever page or change filter you ask for.
  • timeseries: mention counts per day, week or month over a range, each bucket with its episodes, mention lines and the distinct podcasts carrying them. Omit published_after to chart all time.
Mentions inside ad reads are left out unless include_ads is set, so a brand that advertises heavily is measured by what shows say about it rather than by its own ads. That is the difference from particle_podcast_get_episode_timeseries, which counts an episode for a company named only in its ad reads, so a heavy advertiser’s count there can be mostly its own ads. An episode’s mentions are counted once it is processed, usually within a day of publication, so a window ending now reads slightly low, and in timeseries mode the bucket a window cuts through is marked (partial). Compare full buckets, or end a comparison a day earlier for a settled answer.

​
Inputs

Passing no subject returns missing_parameter, and passing two returns invalid_input. A comparison window longer than 90 days, podcast_slug in compare mode, and change, sort, limit or cursor in timeseries mode each return invalid_parameter. A subject slug that matches nothing returns unresolved_reference, so an unknown name is never read as “nobody mentions it”.

​
Output

compare: a markdown document headed ## Mention trends — <Name> (<slug>), then:
  • **Window:**: the current and previous windows, to the second, ends excluded.
  • **Episodes:**, **Mention lines:**: the current count, the previous one, and the relative change.
  • **Podcasts:**: the distinct podcasts in each window and how many are new, continuing and dropped.
A ### Podcasts table follows with each show’s title, slug, mention lines now and before, the change, its episodes, and its change label. With interval, a line per show lists bucket-start count pairs. When more podcasts follow, a paragraph gives the cursor with the window to pass alongside it, and a closing paragraph says how to read a show’s lines with particle_podcast_find_mentions. Sample (company_slug="nike", published_after="2026-10-05", published_before="2026-10-11", limit=5):
timeseries: a markdown document headed ## Mention trends — <Name> (<slug>) by <interval>, with **Range:** and **Totals:** rows and a ### Buckets table of each bucket’s episodes, mention lines and podcasts, the partial ones marked. The range runs from the first bucket’s start to now when published_before is omitted, or to the excluded bound it sets: a bare date includes that whole day, so it shows as the next day. Sample (company_slug="nike", mode="timeseries", published_after="2026-09-01", published_before="2026-10-15"):
A person with no linked knowledge-graph entity gets a note instead of counts.

​
Example

Then read what the top new show said: