KnitKnot

# Getting started with KnitKnot

Set up the benchmark, read the answers behind your results, and choose a supported next step.

## At a glance

  1. 1. **Review the inputs**: Confirm your products, competitors, buyer topics, and prompts in Benchmark setup.
  2. 2. **Run the benchmark**: Choose the engines available to your workspace and start a run.
  3. 3. **Read the evidence**: Open a result, then read the captured answers before choosing a response.

KnitKnot measures how AI systems represent your company when buyers ask about a category or compare vendors. The benchmark keeps the buyer prompt, engine, answer, competitive outcome, claims, and available source evidence together. It supports ChatGPT, Claude, Perplexity, and Gemini; the engines you can run depend on your plan.

## Sign in and review the setup

Sign in at [app.knitknot.ai](https://app.knitknot.ai/signin). Open **Get started** if your workspace is still in Benchmark setup. Review the company and product lines, competitors, capabilities, decision criteria, buyer roles, topics, and prompts. Include the items you want the benchmark to measure before moving to the next step.

After setup, [read your company profile](/docs/read-company-profile/) and approved facts under **Facts**, then inspect the active questions under **Prompt library**. Start with [Run your first benchmark](/docs/run-your-first-benchmark/) for the exact path. Once it completes, use [Read your Overview](/docs/read-overview/) to orient yourself before opening a result.

## Run, read, respond

Open **Benchmarks → Results**, select the engines available to your workspace, and use **Run all**. When scoring finishes, open the result and read the answers behind any surprising number. A low score or a losing comparison tells you where to look; the captured answer tells you what the AI said.

For a disputed statement, follow [Investigate a claim](/docs/investigate-a-claim/). For repeated problems, [work an issue and its linked playbook](/docs/work-an-issue-and-playbook/) to review the proposed response. After your team ships a change, a later comparable benchmark can show whether the measured answers moved. It cannot prove which change caused them to move.

If you need a regular reading, set a schedule from **Benchmarks**. The [run and schedule guide](/docs/run-and-schedule-benchmarks/) explains the controls and how to interpret benchmark periods.

Docs navigation
Docs › Getting started

Getting started with KnitKnot

Set up the benchmark, read the answers behind your results, and choose a supported next step.

Updated

At a glance

  1. Review the inputs

    Confirm your products, competitors, buyer topics, and prompts in Benchmark setup.

  2. Run the benchmark

    Choose the engines available to your workspace and start a run.

  3. Read the evidence

    Open a result, then read the captured answers before choosing a response.

KnitKnot measures how AI systems represent your company when buyers ask about a category or compare vendors. The benchmark keeps the buyer prompt, engine, answer, competitive outcome, claims, and available source evidence together. It supports ChatGPT, Claude, Perplexity, and Gemini; the engines you can run depend on your plan.

Sign in and review the setup

Sign in at app.knitknot.ai. Open Get started if your workspace is still in Benchmark setup. Review the company and product lines, competitors, capabilities, decision criteria, buyer roles, topics, and prompts. Include the items you want the benchmark to measure before moving to the next step.

After setup, read your company profile and approved facts under Facts, then inspect the active questions under Prompt library. Start with Run your first benchmark for the exact path. Once it completes, use Read your Overview to orient yourself before opening a result.

Run, read, respond

Open Benchmarks → Results, select the engines available to your workspace, and use Run all. When scoring finishes, open the result and read the answers behind any surprising number. A low score or a losing comparison tells you where to look; the captured answer tells you what the AI said.

For a disputed statement, follow Investigate a claim. For repeated problems, work an issue and its linked playbook to review the proposed response. After your team ships a change, a later comparable benchmark can show whether the measured answers moved. It cannot prove which change caused them to move.

If you need a regular reading, set a schedule from Benchmarks. The run and schedule guide explains the controls and how to interpret benchmark periods.