Choose the questions
- 20–50 questions your buyers really ask, from sales calls and forms.
- A mix of stages: problem, shortlist, comparison, switching.
- Phrased the way buyers phrase them, including company size and stack.
Choose the engines
ChatGPT, Google AI Overviews, Perplexity and Copilot cover most B2B research. The engines cite different sources, so track them separately.
Choose the countries
Answers change by location. Track each question in each market you sell to; many tools count each country as a separate question.
Record
| Metric | Meaning |
|---|---|
| Share of answers | Questions that name you, out of all tracked |
| Position | First, second or later in the answer |
| Competitors named | Who appears instead of you |
| Cited pages | The sources shaping the answer |
| Accuracy | Whether facts about you are right |
Report monthly
Show the trend, the top three cited pages you're missing from, and the three changes you'll make next.
A tracking sheet (fictional)
| Question | ChatGPT | Perplexity | AI Overviews | Copilot | Top cited page |
|---|---|---|---|---|---|
| Best close software for NetSuite | Named, 2nd | Not named | Named, 1st | Not named | A comparison article |
| Close tools for logistics | Not named | Not named | – | Named | A directory |
Questions
Daily or weekly tracking?
Daily tracking smooths out the natural variation in answers. Weekly is fine if you report trends, not single days.
Why do answers change from day to day?
AI answers vary with wording, location and the sources found at that moment. Look at trends over weeks.
How many questions are enough?
20 well-chosen questions are enough to see trends; 50 gives room for stages and countries.
Why trends beat snapshots
A single AI answer is a sample. Ask the same question twice and you may get different vendors. Tracking daily or weekly over many questions smooths that variation and shows the direction you're moving in, which is what decisions need.
What to do with the data
- Missing from answers: find the pages AI cites and work to get into them.
- Named but described wrongly: fix your facts page and the sources repeating old facts.
- Named but low in the answer: strengthen comparison and proof content.
Each pattern points to a different fix, which is why recording position and accuracy matters, not just presence.
Start here: AI visibility for B2B. More articles in AI visibility.