Run a check you can repeat that shows how often your business is cited or mentioned in AI answers to your priority questions.
If you ask an AI assistant about your business once, you learn very little about how visible you are in AI answers. The answer can be different next time, and one answer doesn't tell you what to fix. A fixed check gives you something you can use. You ask the same questions in the same places and record the answers the same way every month. Over several months, that shows trends a single answer cannot.
Mei Ling runs Mei Ling's Bakes, a home baking business in Punggol, Singapore (an invented example). She asked ChatGPT once whether it knew any home bakers in Punggol, saw her name in the answer and was relieved. A week later she asked again in slightly different words, and her name was missing. She didn't know which answer to believe, and neither answer told her what to do next.
Start with the keyword map from lesson 2.5, "Build your keyword map". Search keywords are short and clipped. Questions to an assistant are longer and closer to speech. Take your priority keyword clusters and rewrite them as ten questions, worded the way a real customer would type or say them.
Use a mix of question types. Some should be business-recommendation questions, where the customer asks who to buy from. Some should be information questions that your pages answer. Add one or two questions that name your business directly, so you can see what assistants say about you when someone asks.
Write each question exactly as you will ask it, and keep that wording every month. If you change the wording, the answers change too, and then you can't compare one month with the next.
Run each question in Google and note whether an AI Overview appears. An AI Overview is the AI-generated answer that sometimes appears in Google results, so you record for each question whether there was one.
Then run each question in at least two AI assistants with web search switched on. Without web search, an assistant answers from its training data and gives no citations, as lesson 8.2, "How other AI assistants find and cite sources", explains. Pick the assistants your customers are most likely to use. If you're unsure which ones those are, ask a few customers.
Use the same setup every time. Sign out or start a fresh chat where possible, so that earlier conversations don't shape the answer. Ask from Singapore, and record the date.
Set up a tracking sheet with one row per question. Give each place you ask its own group of columns: one group for Google and one for each assistant. In each group, record three things:
whether your business is mentioned by name whether any of your pages is cited as a source, and which page which other sources are cited, which shows you who the assistant relies on
Keep mentions and citations in separate columns. A mention means your business is named in the assistant's or AI Overview's answer. A citation means one of your pages is shown as a source, so it was retrieved and used. A mention without a citation means the assistant knows your name, possibly from a directory or review site. Both are useful, and each one points to a different fix.
Mei Ling wrote ten questions and ran her first structured check in Google and two assistants. One question was "Which home bakers in Punggol make custom birthday cakes?" Google showed an AI Overview that cited a food blog's list and two other bakers' sites. The first assistant named her but cited a directory page instead of her own site. The second assistant named three other bakers and cited a forum thread.
Another question was "How far ahead should I order a custom cake in Singapore?" No AI Overview appeared. Neither assistant mentioned or cited her, even though her ordering page answered the question. This was before she rewrote the page in lesson 8.3, and at the time the answer was buried in its fourth paragraph.
Ten questions in three places gave her 30 answers. She was mentioned in 4 of them and cited in 1. She marked three pages to improve. The first was her ordering page. The second was her custom cakes page, which never said "Punggol" in its main text. The third was her flavours page, which showed its menu as an image. Her counts come from one month and will change from run to run, so she will watch whether they rise over the next few months as she makes changes.
When you read your sheet, look for two patterns. If you aren't cited while competitors are, your page usually answers that question less clearly than theirs do. Lesson 8.3, "Write passages an assistant can quote", covers that fix. If you aren't mentioned while other businesses are named, your business may not be described consistently or prominently enough across the web. Modules 6 and 7 cover that fix.
Answers change between runs and between users, as lesson 8.2 explains, so don't react to one month's result. Run the same ten questions on roughly the same day each month. Add each month's results as new columns, and judge the direction over three to six months. Lesson 9.4, "Build your monthly SEO review routine", adds this check to your monthly review.
This check measures what assistants say about your business. It doesn't measure visitors. Counting how many visitors arrive from AI assistants, and seeing what they do on your site, is a job for analytics. The separate course "Marketing analytics: GA4, attribution and testing" covers it.
Write your ten questions now, set up the sheet, and run your first check, as the activity below describes.
Run the ten-question check, enter the citations and mentions in the tracking sheet, and note the three pages most worth improving.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).