Evidence-based calculator
AI share of voice calculator.
Turn prompt and model results you already collected into a position-weighted competitor score. You can inspect every row and change the formula before using the number.
Your data stays on this device
Files are read in your browser and are not sent to Trakkr. Refreshing or resetting clears the analysis. Analytics record only the tool name, input method, file-size band, and row count, never file names or contents.
Add results
Formula
Visible math
The canonical mention share counts one present brand per answer. Recommendation and citation shares use their own classified event totals. Position weighting is an optional prominence view.
position-weighted share = brand position score ÷ all brand position scores × 100
Weights affect prominence, not whether a mention happened. Mention rate always uses present rows divided by prompt × platform observations.
First ≥ second ≥ third ≥ fourth+. Results update as you edit.
Need to choose the denominator or explain the score to a client? Read the complete AI share of voice guide for the difference between mention share, citation share, recommendation share, and absolute visibility.
Method and limits
Comparable set
Use the same prompts, models, date, geography, and tracked brands. Mixing samples makes a precise-looking number that is not comparable.
Observed answers
The tool measures only the answers you supply. It does not estimate prompts you did not run or models you did not include.
No quality claim
Position and citations show prominence and evidence. They do not tell you whether the wording was positive, correct, or likely to convert.
Keep measuring
Turn one calculation into a trend
Trakkr runs real buyer prompts across eight AI models, keeps the answer evidence, and shows how your position changes against competitors over time.
Common questions
How is AI share of voice calculated?
Mention share divides one brand’s mention units by mention units for every tracked brand. Recommendation share and citation share use their own event totals. The optional position-weighted share scores first, second, third, and later mentions separately, then divides each brand’s score by the combined score.
Why do absent brands need a row?
A zero row proves that the brand was checked and absent. Without it, missing data and a genuine non-mention look identical, which makes mention rate and share of voice easy to overstate.
Does this tool call ChatGPT or another AI model?
No. It calculates from results you already collected. That keeps the method auditable and the tool free. Trakkr’s product runs real prompts on a schedule when you need fresh monitoring.
Is my data uploaded?
No. The file is parsed in your browser. Trakkr does not receive the prompts, brands, positions, citations, or file name.