Campaign Brain by Pulse AI Studios

Campaign Brain research

Historical election research, with citations.

Search 103,743 historical election metadata records by keyword, state, office and year. Export cited HTML or CSV briefs of up to 20 records, with each record's original citations and evidence labels. Missing fields stay visible.

Coverage is uneven. Metadata comes from 78 observed years within 1900–2025, not every year or race. Fields that are missing or inconsistent stay absent.

See a sample source cardRead sources and coverage limits

Search the library View research plans

Plans

A brief is one successful retrieval and export report of up to 20 records. It is not 20 separate charges.

If the service is unavailable, a request is refused, or there are no matching records, you are not charged and the credits for that call are restored. Subscriptions renew monthly, with no overage billing.

One brief

$19

25 credits · one research brief

Research

$49/month

125 credits · five briefs a month

Firm

$149/month

625 credits · 25 briefs a month

Checkout opens only once availability is confirmed. For billing and cancellation, see the research terms and support.

Research workspace

Enter keywords and, if you like, explicit filters. Describe what to find, for example “Michigan governor 2022”, not what to do. Strategy prompts are not supported.

The key issued with your purchase. It is never shown in reports.

Up to 240 characters. Example: Michigan governor 2022

Two-letter code

1900–2025

1–20 per brief

Retrying the same key and search in this tab recovers the same request for up to 24 hours, including after a refresh. “Start a separate brief” deliberately starts a new paid search. Keep this tab open while recovering a receipt.

What a record contains

  • Race, year and office
  • State, where known
  • Winner and margin, where the sources support them
  • The original evidence and confidence labels
  • Citations to the publisher’s own page

What it is not

Campaign Brain is not a certified or complete election-results archive. It is also not a strategy tool or a forecasting service.

It does not reproduce full source articles. It offers no voter targeting, tactical advice or predictions, and it makes no claims about why a race turned out as it did.

How to read the evidence

Search matches your keywords and filters against stored records. It does not ask an AI to write an answer.

Evidence quality varies. Some records are sparse: a missing state, no margin, or little case documentation. We leave those gaps visible rather than fill them in.

Every record keeps its original evidence labels and legacy corpus status, including “beta-only.” These labels remain visible alongside citations; launching the product does not certify every historical record.

Full sources stay on publishers’ websites. See the source notes.