Free AI Share of Voice: A Spreadsheet Method You Can Run Today

A share of voice check answers one question: when buyers ask AI engines about your category, how often are you named compared with the brands you compete against? The definition and limits of AI share of voice matter here, because the number is only as good as the prompts and competitors behind it.

A spreadsheet gives you two useful numbers per engine:

• Mention rate: the share of answers that name you at all.
• Share of voice: your mentions divided by the mentions of every brand you track.

It doesn't tell you why an engine picked a brand, and it doesn't give you a rank. No first-party source publishes position data for AI answers, so treat any "AI position" figure with suspicion. For the difference between rate-style and share-style metrics, see citation rate vs citation share.

Plan on about two hours for the first run. Later runs go faster.

1. Write 15 to 25 unbranded prompts. Use questions a buyer asks before they know you: "best payroll software for a 20-person agency", not "is YourBrand good". Give each prompt an ID (P01, P02) and freeze the wording.
2. Pick three to five like-sized competitors. Choose brands with similar size, price point and market. Leave category giants out of the denominator (more on that below).
3. Pick your engines and fix the settings. For example ChatGPT, Perplexity, Gemini, Google AI Mode and Claude. Use a logged-out session or a fresh chat with memory and custom instructions off. OpenAI says ChatGPT may use saved memories when it rewrites a search query, which makes logged-in results hard to repeat.
4. Set up a Log tab. Columns: A Date, B Engine, C Prompt ID, D Run, E Citations shown (Y/N), F Brand, G Mentioned (1 or 0). Add one row per tracked brand per answer.
5. Run each prompt twice per engine. Answers vary between runs. Two runs catch the most obvious noise. Note in column E whether the answer showed source links, since that hints whether it searched the web.
6. Add the formulas on a Summary tab. For your share on ChatGPT: =SUMIFS(Log!G:G,Log!B:B,"ChatGPT",Log!F:F,"YourBrand")/SUMIFS(Log!G:G,Log!B:B,"ChatGPT"). For your mention rate on ChatGPT: =SUMIFS(Log!G:G,Log!B:B,"ChatGPT",Log!F:F,"YourBrand")/COUNTIFS(Log!B:B,"ChatGPT",Log!F:F,"YourBrand"). Copy each row and change the engine name.
7. Repeat monthly with the same prompts, brands and settings. Add new prompts in a separate tab so they don't break the trend.

The step-by-step share of voice playbook covers prompt selection in more depth.

Share is a fraction, and the denominator moves without you. If competitors gain mentions faster than you do, your share drops even as you appear more often.

A hypothetical: say you run a scheduling tool. In March, ChatGPT names you in 6 of 20 answers, and your three competitors collect 14 mentions between them. Your share is 6 of 20, or 30%. In April you appear in 9 answers, a real gain. Competitors rise to 27 mentions. Your share is now 9 of 36, or 25%.

Mention rate went up. Share went down. Both are true. That's why the sheet reports both, and why a falling share alone is not proof that something broke. Check the mention rate first, then look at which competitor grew and on which engine.

AI answers concentrate on a small set of well-known sources. A 2025 study of AI search citations (arXiv:2507.05301) found the top 20 news sources accounted for 67.3% of citations for OpenAI models. Semrush's index of 126 million prompts found 36 brands visible across all four surfaces it tracked every month, including YouTube, Google, Reddit and Amazon.

That finding is brand size, not technique. If you put a category giant in your denominator, its mentions swamp yours and your share barely moves whatever you do. The result mostly measures how famous the giant is. Compare against brands of your scale and the number starts telling you something you can act on. For choosing that set, see this free way to see which competitors AI recommends.

Read each engine on its own. Google says AI Overviews and AI Mode may use different models and techniques, so even two Google surfaces can disagree. A blended score hides the engine where you're absent.

Then work through it in this order:

• Zero mentions on one engine: check access first. Confirm your robots.txt allows that engine's search crawler (for ChatGPT search that's OAI-SearchBot, not GPTBot) and that your key facts appear in the HTML before JavaScript runs.
• Mentioned, but described wrongly: that's an accuracy problem, not a share problem. The fix plan for AI getting your brand wrong covers it.
• Competitors named via third-party pages: when answers cite review sites or lists you're missing from, being named on those sites is associated with appearing in the answer. It's an association, not a guarantee.

None of this promises a higher share next month. You control access and the facts on your pages. You don't control what the models say.

Your sheet is a sample, and an imperfect one.

• You can't see the real queries. OpenAI says ChatGPT search typically rewrites your query into one or more targeted queries. Your prompt is a proxy for an input you never observe.
• Retrieval state is mixed. Engines decide per query whether to search the web. Some answers come from the model's training data alone. Column E helps, but it's a hint, not proof.
• Plans differ. Free and paid tiers can route to different models. See whether ChatGPT answers differently on free and paid plans.
• Small samples swing. With 20 prompts and two runs, one answer can move share by several points. Look at trends over months, not single readings.

A critical survey of 45 GEO studies (arXiv:2607.14035) found no technique with a stable, cross-platform causal effect on discoverability. Be wary of anyone selling a guaranteed share lift.

If you want a quick snapshot before building the sheet, the free checker on this site runs up to three buyer questions (yours, or ones written from your homepage) and scores the answers. It's a one-off snapshot, not continuous tracking, so keep your spreadsheet for the monthly trend.

If you'd rather not run prompts by hand each month, you can start on the 99Visibility Free plan with no card. Either way, keep the same rules: fixed prompts, per-engine numbers, like-sized competitors.