Your “local” alert is probably a keyword in disguise
Many desks build “Chicago news” or “Portland coverage” by stuffing the city into a keyword field. That habit creates false confidence: the feed looks busy, the Slack channel never sleeps, and half the hits are the wrong market, a person named Austin, or a national wire that mentioned the city once.
This page is a diagnostic, not a full geography-param cookbook. It shows five failure modes with live volume gaps, then points you to structured filters. For the complete local news API surface (county, radius, Stories, Sources), use Perigon’s Local news API search guide.
- What goes wrong when the city is only a keyword?
- How big is the gap for ambiguous markets (Portland, Austin, Springfield)?
- How do event-city and newsroom-city filters differ (Chicago)?
- What should you change in the alert job tomorrow?
Try it: Articles sandbox · Docs: Local news · Plans: API pricing
Failure mode 1: Ambiguous city names (the Portland problem)
Portland is not one market. A keyword match cannot tell Oregon from Maine - or from cement, travel, and sports noise that shares the string.
Live Articles counts for pubDate 2026-06-25 through 2026-07-25:
| Filter | About how many articles |
|---|---|
| q = Portland | ~67,000 |
| city = Portland + state = OR | ~6,200 |
| city = Portland + state = ME | ~1,300 |
Keyword Portland is roughly 10× Oregon event-scoped volume and 50× Maine. If your “Portland desk” used q, most of the firehose was not your DMA.
Failure mode 2: Person and brand collisions (Austin)
Austin is a capital city and a common given name. Keyword search pulls both.
| Filter | About how many articles |
|---|---|
| q = Austin | ~143,000 |
| city = Austin + state = TX | ~11,000 |
Texas event geography is about 13× quieter than the keyword - still busy, but no longer every headline that mentions someone named Austin.
Failure mode 3: Many towns, one string (Springfield)
Springfield is a classic trap: Illinois capital, plus cities in other states, plus fiction and sports nicknames.
| Filter | About how many articles |
|---|---|
| q = Springfield | ~24,000 |
| city = Springfield + state = IL | ~2,200 |
If the job is Springfield, Illinois government or plant news, the keyword feed is mostly other Springfields and incidental mentions.
Failure mode 4: Event city vs newsroom city (Chicago)
Even when the place string is unique enough, where the story happened and where the publisher sits are different questions. Keyword Chicago mixes both - and everything else.
| Filter | About how many articles |
|---|---|
| q = Chicago | ~321,000 |
| city = Chicago + state = IL (event / story place) | ~60,000 |
| sourceCity = Chicago + sourceState = IL (publisher based there) | ~144,000 |
Event-scoped Chicago is far below the keyword firehose. Newsroom-scoped Chicago is larger than event-scoped - national and metro outlets headquartered in Chicago cover many non-Chicago events. Pick the axis on purpose: city council and weather emergencies → event city / state; hyperlocal reader apps and DMA ad products → sourceCity / sourceState. Combining both is a third product (events in a county from in-state publishers) - details in the local news guide linked above and location & geo docs.
Failure mode 5: Treating volume as quality
A keyword city alert “working” usually means it is loud. Use a one-week bakeoff before you ship:
- Run q = CityName and city + state side by side for the same from/to window.
- Spot-check twenty titles from each - count wrong-market and name-collision hits.
- If you need outlets in the market rather than events there, swap in sourceCity / sourceState (or discover publishers on the Sources API first).
- Only then add topic, company, or radius - geography first, keywords second.
Radius feeds (lat / lon / maxDistance) belong when the user or plant is a point on a map, not a city string - again owned by the full local guide, not this diagnostic.
What to change in the alert job tomorrow
| If your job looks like… | Stop doing | Start doing |
|---|---|---|
| “Portland OR desk” via q | Keyword Portland | city + state = OR |
| “Austin CI” via q | Keyword Austin | city + state = TX (+ company filters) |
| “Springfield IL plant” | Keyword Springfield | city + state = IL (or county) |
| “Chicago hyperlocal app” | Keyword Chicago | sourceCity / sourceState, or Sources then allowlist |
| “Stories in the metro” | Article keyword flood | Stories with city / state (see local guide) |
Key takeaways
- Place-name keywords are not local news monitoring - they are string matches with market collisions.
- Live gaps are large (e.g. Portland q ~67k vs OR ~6.2k vs ME ~1.3k; Austin q ~143k vs TX ~11k in 30 days ending 2026-07-25).
- Chicago shows event city (~60k) and sourceCity (~144k) answer different product questions - neither equals keyword Chicago (~321k).
- Fix the geography axis first; leave the full param and radius cookbook to Perigon’s Local news API search guide.
Get started: Articles sandbox · Local news docs · Local news API search guide · API pricing
Top comments (0)