Step 1. Start from buying decisions, not keywords
For each product line, write down the decisions a buyer makes before choosing: which TV suits a bright living room, and where to buy it with easy returns. Each decision becomes a topic: one subject you want to be found for, with several questions under it.
Step 2. Write each question the way a person asks an assistant
Use whole sentences and the details people add: budget, use, size, place, timing. Longer questions are where AI answers show up most on Google. In Pew Research Center's study of 68,879 Google searches made by 900 US adults in March 2025, an AI summary appeared for 53% of searches of ten words or more, and for 8% of one- or two-word searches.[1]
Keyword65 inch tv dealsQuestionWhat's the best 65-inch TV for a bright living room if I mostly watch sports?Step 3. Cover learning, comparing and buying
Give every topic at least one question for each stage of the decision:
- Learn: "What should I look for in a TV for a bright room?"
- Compare: "Best Buy or Amazon: which is better for buying a TV with easy returns?"
- Buy: "Where can I buy a 65-inch TV with free delivery this week?"
The free audit asks for exactly these three kinds of question, one of each. In aSERP, every question carries its intent: informational, comparative or transactional.
Step 4. Track more than one phrasing of the needs that matter most
People word the same need in very different ways. When SparkToro and Gumshoe asked 142 people to write a prompt for the same need, almost no two were alike: their average semantic similarity was 0.081, on a scale where 1 means identical.[2] For your most important needs, track two or three phrasings taken from your customers' own words, casual and formal, short and detailed.
Step 5. Name the place when the answer depends on it
If the right answer depends on where the buyer is, put the place in the question: "… with free delivery in Chicago". A question that names the place tells every engine which place you mean.
Step 6. Mix questions that name you with questions that don't
Questions about the category ("Where should I buy a TV online?") show whether engines recommend you when nobody mentions you. Questions that name your store, for example whether it is reliable for returns, show how engines describe you. Keep most of your set about the category: those are the answers where you can be chosen over a competitor.
Step 7. Add the questions to aSERP
In your workspace, open Brand Settings›Topics & prompts and choose Add topic. Type the topic's name, then type each question and press Enter after it. Leave commas out of your questions: in this box a comma starts a new question, just as a new line does. When all your questions are listed, choose Confirm. A question can be 3 to 300 characters long, and each one counts toward the tracked questions your plan includes (see Plans). Then, right away, open the topic on the Topics page, choose Edit questions, set each question's intent (informational, comparative or transactional) and choose Save current version. New questions are measured from the next scan.
Step 8. Settle a topic before its first scan
Changing the questions of a tracked topic, by rewording one or adding one, creates a new version of the topic, and some of its figures then count only the current version. Settle a topic's questions before its first scan. To try new wording later, add it as a new question rather than rewording an old one, so the old question goes on being asked exactly as before.
Check it worked
Right away
The topic appears in Brand Settings›Topics & prompts with the number of questions you added, and on the Topics page each question shows its intent. A question already in that topic is not added a second time.
In aSERP
After the next scan, each question on the Topics page has answers from every engine you track; an engine that could not answer that day shows AI unavailable, which is not counted as not mentioned. After several days of daily scans, read Visibility for each topic and each engine. aSERP shows how sure it is of each figure: a 95% interval and, on each metric card, a confidence label (High confidence, Directional or Low confidence; the headline above the cards shows only High or Low). Don't act on a topic while its Topics page shows Needs more data. Compare two periods only when aSERP shows the comparison; it declines one when a period has too few measured days.
With the free audit
It asks your three questions once on six engines, which gives a first look, not a trend. It runs once per visitor, so use it for a first look at three of your questions, not to check a change.
If nothing moved
If a question still shows no answers after the next scan, open Brand Settings›Tracking schedule to see when the next scan is, and check that the question's topic is still listed in Topics & prompts.
What not to do
- Don't track only questions that name your brand.
- Don't paste keywords from an SEO tool as if they were questions.
- Don't reword a tracked question to make its result look better.
- Don't judge a topic on one day of answers.
- Don't write leading questions that contain their answer, such as asking why a named store is the best place to buy a TV.
Sources
- Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results”, 22 July 2025.
- Search Engine Journal, reporting a SparkToro and Gumshoe study, “AI Recommendations Change With Nearly Every Query: SparkToro”, 30 January 2026.