Your first project

You will need access to a CiteLadder workspace, your website address and a clear idea of the market you serve. Available runs and features depend on your workspace’s access and allowances.

1. Review your company facts

Create a project from your website. Review the business category, buyer type, market, owned domains and suggested competitors before confirming.

Suggestions are a starting point. Remove companies that do not compete for your actual buyers and correct ambiguous names or domains. Review inferred positioning, audience and offerings in your project as well; discovered text is not automatically a confirmed company fact.

You are ready to continue when: the project describes your business and intended audience accurately.

2. Review your prompts

Open Prompts after the initial portfolio finishes generating. Read the questions as if you were a potential customer. Keep a focused set that covers the decisions buyers make, and distinguish general discovery questions from questions that name your brand.

For a fictional scheduling product, a useful discovery question might be “Which scheduling tools work well for a small consultancy?” A question such as “How does Example Scheduler handle reminders?” answers a different, brand-specific question.

See Prompts and topics before expanding the portfolio.

3. Collect the baseline you need

For AI search, start an audit with the prompts, engines and repetitions you intend to measure. Review its estimate and confirm the run. Check Runs for completion and failures.

For website evidence, open Site Health and choose Run new crawl. Creating a project does not automatically start its first crawl.

For search performance, connect a supported provider and map the correct property to this project. Wait for imported coverage before interpreting an empty chart.

An estimate is a preview. It does not start a run, and it is not the final measured cost.

4. Inspect one result

Open an AI answer, an affected website page or a search query behind a summary. Check the observation date and scope. If a result is partial or unavailable, read the reason before treating it as a problem with your brand.

The goal is one finding you can explain with evidence, rather than a long list of numbers.

5. Take it into the Agent

Open a relevant Action and select Work on this, or use Ask agent on a supported evidence screen. Check the attached references in the composer, then describe the outcome you need.

Try: “Explain this finding, show the evidence behind it, and propose a small page edit for review.”

Review the result before using it. Once your team actually makes the change, follow implementation and measurement.

If setup does not finish

Keep the existing project and inspect its displayed status before creating another. Initial prompt generation can finish after the project is created. A provider failure or unavailable dataset needs its own recovery; repeatedly opening the page does not rerun it.

For access and project selection, see Workspaces and access.