Docs Sign inGet started
Concepts

Changes

A feed of what moved between runs, filtered for noise.

Ask Claude Ask ChatGPT

After every run Onym compares it with the previous run and lists what changed beyond noise. Changes use prompts that don't name your brand, and only prompts and surfaces measured in both runs.

Change types

kind Reported when
visibility Your overall visibility rose or fell significantly
surface Your visibility on one surface changed significantly
competitor A tracked competitor gained or lost significantly
prompt_lost One model named you for a prompt in each of the 3 runs before, and in neither of the last 2
prompt_won One model missed you for a prompt in each of the 3 runs before, and named you in both of the last 2
new_brand An untracked brand appeared in at least three answers after none in the last run
new_source A site was cited in at least three answers after none in the last run

Each change has a tone: bad, good or info. Bad changes are listed first.

Prompt changes are reported per model, like "Lost on ChatGPT: “Best auth for Rails?”". A single answer can't show a real change, so Onym looks for a consistent pattern across runs instead: named every time, then not at all.

Where you see them

  • The Changes page in the dashboard, newest run first.
  • The email digest and Slack message after each run.
  • The change.detected webhook.
  • List changes in the API, and get_changes in MCP.
A change
{
  "kind": "competitor",
  "tone": "bad",
  "title": "Northwind gained on Perplexity",
  "detail": "41% → 58% of answers (n=74)",
  "link": "/competitors"
}

link is a path inside the domain's dashboard.