How to track your AI share of voice every week
The short answer
- AI share of voice is the share of AI answers to your category’s questions that name or cite your brand.
- Use a fixed set of 30–100 buyer questions, run it on the same day each week across the engines you care about, and record brand mentions and cited URLs.
- The number only matters if it leads somewhere: tie citations to the pages behind them and to signups.
What is AI share of voice?
AI share of voice is a simple ratio: out of the answers an AI engine gives to a defined set of category questions, how many name or cite your brand? Track it next to competitors in the same answers and you get a picture of who the assistants recommend in your space.
It is worth measuring because AI answers are now a first stop for many buyers, and because they behave differently from search rankings. A brand can rank well in Google and still be absent from ChatGPT’s answer to the same question.
Think of it as a sample, not a census. You are checking a fixed set of questions, not every question anyone asks, so the absolute number matters less than how it moves and how you compare with the brands that appear next to you.
How do you build the prompt set?
Start from real buyer language: sales call notes, support tickets, search queries from Search Console and the questions people post in communities. Write each as a person would type it to an assistant. Group them by intent so you can see where you are strong and weak.
Aim for balance across the groups. If most of your prompts are brand questions, the share will look flattering and tell you little. A set weighted towards category and problem questions shows whether assistants recommend you to people who have not heard of you yet, which is where new demand comes from.
- Category questions: “best tools for …”, “how do I choose a …”.
- Problem questions: “how do I fix …”, “why does … happen”.
- Comparison questions: “X vs Y for …”.
- Brand questions: “is X good for …”, “X pricing”.
- Keep the set fixed for a quarter so trends mean something.
How do you run and score it?
Run every prompt through each engine on the same day of the week, in a clean session. For each answer record four things: was the brand named, was a URL from your site cited, which competitors appeared, and which domains were cited. Answers vary between runs, so look at weekly averages and trends rather than any single response.
A short Python script can do the bookkeeping: read the prompt list, call each engine’s API where one exists or paste in exported answers, search each answer for your brand terms and domain, and append a row per prompt to a CSV. A second step groups the rows by week and intent and computes the share.
| Column | Example |
|---|---|
| week | 2026-W41 |
| engine | Perplexity |
| prompt_id | cat-07 |
| brand_named | yes |
| our_url_cited | /guides/pricing |
| competitors | Brand A; Brand B |
What do you do with the gaps?
Each prompt where a competitor is cited and you are not is a work item. Read the pages the engine cited instead of yours: what question do they answer that your page does not, and how directly? Then fix the page that should win, make sure it is readable in raw HTML, and look for places other sites could mention you.
That is the loop I ran at Imagera AI. The brand appeared in about 8% of tracked AI answers and no line ran from a citation to a signup. A weekly citation check, structured data and fixes to the pages that earn the mention, with every citation tracked through to signups, took AI-answer citation share from ~8% to ~19%.
How do you report AI share of voice to a team?
Keep the weekly report to one screen. Show share of voice per engine as a trend line, the prompts you gained and lost since last week, the competitors that appear most often, and the three page fixes planned in response. Leave the raw table in a linked sheet for anyone who wants to check a specific answer.
Separate the leading indicator from the result. Citations and share of voice go in one block; AI referral sessions and signups from cited pages go in another. When the two move together, the work is paying off; when citations rise and signups do not, look at what the cited pages say and where they send people.
- Share of voice by engine, last 8 weeks.
- Prompts won and lost this week, with the cited URL.
- Top competitors named in your category answers.
- AI referral sessions and signups from cited pages.
- Next three page fixes and who owns them.
What mistakes make AI share of voice misleading?
Changing the prompt set every week is the most common one: the number moves because the questions moved. Counting only brand-name questions is the second, because assistants usually name you when asked about you; the useful signal is in category and problem questions. Mixing engines into one figure is the third, since each engine cites differently.
Finally, avoid reading a single answer as a verdict. Run each prompt more than once if you can, look at weekly averages, and treat sudden jumps with suspicion until they hold for a few weeks.
How do you connect citations to revenue?
Citations are a leading indicator, not the result. Tag the landing pages that AI engines cite, watch referral traffic from AI sources in your analytics, and follow those sessions to signups and purchases. When a citation gain shows up as more qualified signups on the cited page, you have the number a budget owner cares about.
The free AI share-of-voice calculator on this site turns your weekly tallies into share of voice by engine and intent, so the trend is ready for a weekly report.
Questions
How many prompts do I need to track AI share of voice?
Thirty to one hundred questions is enough for most brands, grouped by intent. Fewer and the share swings too much week to week; more and the review work outgrows the value. Keep the set fixed for a quarter.
Why do AI answers change between runs?
AI engines generate answers each time and retrieve sources dynamically, so wording and citations vary from run to run. That is why share of voice should be read as a weekly average and a trend, not from a single answer.
Which AI engines should I track?
Track the ones your buyers use. For most B2B and consumer brands that means ChatGPT, Perplexity, Google AI Overviews or AI Mode, and Gemini, and Claude where your audience is technical. Engines cite different sources, so track them separately.
Sources
