Organisation quickstart
From a new account to a first real answer, then a five-day plan to turn it into a loop you can keep running.
Setup is short. Onboarding finds your business, confirms what you sell and where, and proposes prompts. The useful part is what you do in the five days after that, and this is the plan.
What you'll learn
- What onboarding asks for, and why each answer changes your results
- What to expect while the first checks run
- A five-day plan that ends with one change shipped and measurable
Day 1: Onboarding
Go to onboarding and work through it. It asks for:
- Your business. Search for it, then confirm what comes back. The name, the website and the description are what the judge model matches against, so use the trading name your customers use, not the legal entity.
- What you sell. Your category and your offerings. This shapes the prompts that get proposed.
- Where you sell it. Your markets. A prompt without a place in it gets a different answer from one with your city in it, and for a local business the place is most of the value.
- The Trace tag, optionally. You can install it later. See install Trace.
Then accept the prompts that match how your buyers actually ask, and remove the ones that do not. Fifteen good prompts beat two hundred generic ones. See your first prompts.
Your own name never goes into a prompt. The measurement is whether the engines name you unprompted.
Day 2: Add the rivals you know, then wait
Add the businesses you sell against. You do not need a complete list: the judge records every competitor named in every answer it reads, so the interesting ones tend to arrive on their own.
While the first cycle runs, there will be gaps. A new workspace is not a broken one. See why is there no data yet.
Day 3: Ask your first real question
Open the assistant and ask "What is my citation rate on each engine?" Read the answer, then read the panel beside it. Two things to look for:
- Which surfaces have a rate at all. A surface reported as excluded has no measurement yet, which is not the same as a zero.
- The spread. If one surface is far below the others, that is your first thread to pull.
Then ask "Show me the exact sentences AI wrote about us." This is the moment the product usually clicks: not a number about you, but a sentence about you, written by an engine, dated.
Day 4: Find the one thing to fix
Ask "Is my website ready for AI answers?" The Readiness panel grades four levels in the order they fail, and the first real failure is your first job. Most brands find it at level 1 or 2: a blocked crawler, or content that only exists once JavaScript has run.
Then ask "Where should I focus first, and what is it worth?" That gives you the ranked plan. Treat the order as more reliable than the sizes.
Day 5: Ship one change through the ledger
Take the top row, approve it, ship it. Approving it inside the platform is what captures the control the result will later be scored against, which is the difference between a change you can prove and a change you can only claim. See ship and prove a change.
Then set your alerts so the next finding comes to you: set up alerts.
The week, in one table
| Day | Do this | You end up with |
|---|---|---|
| 1 | Onboarding, prompts | A measured prompt set |
| 2 | Rivals, wait | The first checks running |
| 3 | Ask for rates and sentences | A baseline you believe |
| 4 | Readiness and the plan | One thing to fix, ranked |
| 5 | Approve and ship one change | A change that can be proven |
Try this in Outercite
Open the assistant and ask "Which prompts am I losing, and who is taking them?" If the answer is thin, your prompt set is the thing to work on before anything else.
Related
Your first prompts
How to choose the fifteen that matter.
Reading your first answer
What every part of an answer and its panel means.
Meet the assistant
The screen you will be using every day.
Last updated on
