Measure, repeat, decide — with a template

Why measure at all?

Because answers change. Assistants update their models, their search partners and their sources, and the same question can give a different answer next month. Without a baseline you cannot tell whether your work helped, or whether the assistant simply changed.

The rules that make a measurement

• Lock the questions before the first run. Do not drop the ones where you look bad. A changed wording is a new question.
• Same assistants, same settings. Web search on or off, logged out, new chat each time.
• Two or three runs per question per assistant. Count how often you are named, not whether you were named once.
• Store everything. The full answer, the names in order, the sources, the date. You will want to look back.
• Three numbers, kept apart. How often you are named; which sources are cited, yours or independent; how many factual errors remain. A single combined score hides which of them moved.

How often to repeat

Often enough to see change, rarely enough to have done something in between. For most small businesses that is once a month or once a quarter. Between two runs, change one thing where you can — fix the directory entries, rewrite one service page, earn one article — so that a difference can be traced to its cause.

Google notes that recrawling a changed page can take anywhere from several days to several months. Do not expect the next run to show last week's change.

One decision per cycle

After each run, answer three questions:

1. Where did we move — which questions, which assistants?
2. Which sources did the assistants use for the companies named instead of us?
3. What is the one thing we strengthen before the next run?

Write the decision down next to the results. Over a year, that column is your strategy.

The template

The template has one row per answer and the columns this guide uses: date, assistant, model if shown, web search on or off, prompt, run number, named yes or no, position, other companies named, sources cited, own or independent, factual errors, notes.

Download the check template (CSV, opens in Excel, Numbers or Google Sheets)

If you would rather have it run for you, the free check does the first run on your domain. Our method describes how we repeat it for clients.