Journeys
Journeys measures what AI assistants and coding agents say and do with your product: you give an agent a question or a task, EkLine runs it as many times as you choose, grades every answer with checks, and shows you the result.
Any member of your organization can create, edit and run journeys and collections, and share their reports.
The parts of a journey
Section titled “The parts of a journey”| Term | What it is |
|---|---|
| Journey | A question or task to test. It holds one to 20 turns (messages sent in order, in one conversation), a persona the agent plays, the checks that grade each answer, and the agents to run it on. |
| Run | One agent working through every turn once, as one conversation. Each run produces one answer to grade. |
| Report | What you get each time you start a journey: every run from every agent you selected. |
| Check | A test each answer must pass, such as you are recommended, not just mentioned. Code measures some checks; an AI model judges others against a written rubric. |
| Collection | A set of journeys that run together and share one report. The free report creates a collection called Buyer questions. |
On a collection’s report, each journey is called a buyer question.
Two kinds of journey
Section titled “Two kinds of journey”| Kind | What it tests | Agents that can run it |
|---|---|---|
| Research | What an agent answers when someone asks about your category, your product, or a problem your product solves. New journeys start as Research. | ChatGPT, Claude, Perplexity, Gemini, Claude Code, Codex |
| Agent Task | Whether a coding agent can do a job with your product, such as following your quickstart. Each run starts in an empty working folder. | Claude Code, Codex |
You choose one to six agents for a journey, and a model for each. See Choose agents and models.
Agents reach your product the way your buyers’ assistants do: through public web search and the pages they open. They do not read anything you have not published.
What you get
Section titled “What you get”- A report for every run of a journey. A scorecard of checks against runs, what went wrong and why, who was named, which sites the agents cited, and totals. See Read a Buyer questions report.
- A report for every run of a collection. Each buyer question is marked You win, Split or You lose, with who was recommended instead of you, where assistants get their answers, and up to three fixes to start first. See How results are measured.
Where Journeys fits
Section titled “Where Journeys fits”Journeys is the measure step in EkLine’s loop:
- Measure. Run a journey or a collection.
- Fix. From a report’s What went wrong, click the button the cause offers: Fix in docs starts a Docs Agent session, and Draft in Marketing starts a Marketing Agent session. From a collection report, click Start the fix. See Turn a gap into a fix.
- Approve. Your team reviews the pull request or draft, as with any other agent session.
- Run again. Run the collection again. Its report compares each buyer question with the collection’s first run. See Track results over time.
A journey runs when someone in your organization starts it.
Journeys pages
Section titled “Journeys pages”- How results are measured: what happens in a run, when an answer passes, and how You win, Split and You lose are decided.
- Read a Buyer questions report: every part of the report, in order.
- Write checks that mean something: choose and write checks so a result means what you think.
- Create a research journey: write a question and run it.
- Agent Task journeys: test whether a coding agent can build with your product.
- Choose agents and models: which agents to run and why.
- Collections and sharing: group journeys, run them together, and share the report.
- Turn a gap into a fix: hand a failed answer to the Docs Agent or Marketing Agent.
- Track results over time: compare a run with earlier ones.
- Journey agents and checks: reference tables for agents, models, checks and limits.
To get your first report without building journeys yourself, see Get your first AI report.
Which AI assistants and coding agents does Journeys test?
Section titled “Which AI assistants and coding agents does Journeys test?”ChatGPT, Claude, Perplexity and Gemini for Research journeys, and the coding agents Claude Code and Codex for both kinds. No other agents are available.
Does a journey read our private or signed-in docs?
Section titled “Does a journey read our private or signed-in docs?”No. Agents find your pages through public web search, as your buyers’ assistants do. The documentation sites in Company Profile are used only to recognize your pages among the sources an answer cites.
What is the difference between a journey and a collection?
Section titled “What is the difference between a journey and a collection?”A journey is one question or task. A collection runs many journeys together and gives one report that marks each question as You win, Split or You lose. A single journey’s report shows its checks run by run, with no win or loss.
Who in my organization can use Journeys?
Section titled “Who in my organization can use Journeys?”Any member. There is no separate role for creating, running or sharing.
Built on Potluck Docs, our open source template.