What is share of voice in AI answers?

Share of voice is the share of the answers to a fixed set of questions that cite your site or name you, compared with the other sites cited in them.

The term is an old one from advertising and public relations, where it meant your brand's share of everything published or broadcast about its market. It has moved into AI answers and become one of the measures of optimising for AI answers (GEO), where the question becomes how many of every ten answers about your field run through your site, not how often you were mentioned in the papers. And because a search may end at the answer itself, with no click, as in what is zero-click search, and what does it mean for your site?, your share of the answers has become a measure that visits alone cannot replace.

It can be counted in two ways, each with its own meaning, so always say which one you mean:

• Share of answers: the number of answers that cite your site, divided by all the answers. Cited in 5 answers out of 20, your share is 25%.
• Share of citations: the number of links to your site, divided by all the links in the same answers. If the twenty answers hold 100 links and 8 of them point to you, your share is 8%.

With a simple table and five steps; the first day takes no more than an hour:

1. Write a fixed set of questions: ten questions your customers ask before they buy, in their words. How to choose them is in how to choose the questions you measure on.
2. Ask each engine, on the same day, ChatGPT and Gemini for example, which gives you twenty answers.
3. Record each answer's sites: the sites cited and the names mentioned without a link, each in its own column.
4. Work out your share and your strongest competitor's, with the method you chose, keeping citations with a link apart from mentions without one.
5. Count again on other days: answers change, and one day's figure is a snapshot, not a trend, as in why ChatGPT's answer changes, and when a change is real.

A rank is your page's place in a list of search results. Share of voice is your place among the sites cited inside the answers. The Tasadr Score is a number out of 100 for your own site, built from every answer to every question on every engine, in which a citation with a link counts for more than a mention without one, and in which a missing answer is never counted against you.

The practical difference is that share of voice is relative: it can rise when a competitor drops out without you changing anything, and fall when a new site enters your field. The score says where your site stands on its own questions. So read them together: the score says where you are, the share says how much is left for others.

In Tasadr you see the competitor ranking for your questions, and your share of citations next to the strongest of them. What the number says and does not say is in what the score out of 100 tells you.

Because each sector's answers are spread differently. On buying questions we measured on Perplexity and Claude between 11 and 15 October 2026, in Saudi Arabia, the UAE, Kuwait and Egypt, the answers cited one global review site on 83 of 101 restaurant questions, so its share of "best restaurant" questions is larger than any single restaurant's.

On 96 hotel questions, the answers clustered around a few sites: a review site and two booking platforms were each cited on at least 69 of them. On 102 e-commerce questions, by contrast, the citations were spread across 1,338 different sites, 1,061 of them cited on a single question only, so most stores' shares are small by the nature of the market.

So choose the questions your customer really asks, not the ones you would like to win, and keep your name out of them: whoever types your name already knows you, and your share of questions that carry your name is a reassuring number that tells you nothing. Compare your share with your competitors' on the same questions, never on different ones.

And do not compare your share with that of a business in another sector: 25% on e-commerce questions may not mean what it means on hotel questions. Examples from these sectors are in how does AI choose which restaurant to recommend?, how AI answers recommend your hotel to guests and how ChatGPT and Google recommend your online store.

You read it to know where to work, not to put it in a report and forget it. Four uses:

• Read it question by question: the overall figure hides the questions you lose entirely.
• Look at who takes the share on your questions, a competing site like yours, a directory or a platform, because the work differs in each case, as in which competitor ChatGPT recommends instead of you, and why.
• Follow it on the same questions month after month, and do not change the questions each time, or you will be measuring something else.
• Set it beside the edits you made and their dates, to read what changed after each one; the method is in how to measure the success of optimising for AI answers.