How Do You Increase AI Share of Voice?

How to increase AI share of voice: find the gap prompts where a rival is named and you aren't, read the sources those answers cite, match the work to the source, then re-count on the same basket before claiming a rise.

Angelique Swain · · 5 min read

Cover reading How to Increase AI Share of Voice, Prompt by Prompt, in cream type on a dark Firemark ground
On this page

Asked to increase AI share of voice, most teams publish more posts and hope the number follows.

The part of the number you can take credit for moves only when answers that used to skip your brand start naming it, on the same questions you counted last time.

So the useful answer to how to increase share of voice in AI answers is a work order: find the prompts where a rival is named and you aren't, read what those answers cite, work on that, then count again.

What makes share of voice rise?

AI share of voice rises for real when more of the answers you collect name your brand, on the same prompts and engines as before.

In Firemark's version, your share is the answers that name you divided by the answers collected, and every rival you track gets the same count. Firemark's guide to AI share of voice covers the formula and the other ways tools divide it.

That leaves three ways a share can move:

  1. You're named in answers that used to skip you. This is the one you earned.
  2. The basket or the rival list changed. Drop prompts you were losing and your share rises with no new answers.
  3. The engine changed how it answers. Engines update on their own schedule, and they don't send a memo.

A rival doesn't have to lose an answer for you to gain one. Two brands can be named in the same answer, so under this divisor the shares across brands can add up to more than 100%.

Where do you start?

Start with gap prompts: the questions where an engine names a rival you track and doesn't name you.

Four steps to increase AI share of voice: find gap prompts, read what the answers cite, match the work to the source, then re-count the same prompts over three runs
Firemark calls a rise only after three comparable runs and a change of 5 points or more

On a gap prompt, the engine has already decided the question deserves a brand. It picked someone else. Firemark's gap tools, added on October 25, 2026, list these prompts through your AI assistant and as an export.

Which gaps come first?

Close missing gaps on bottom-of-funnel prompts first, since those questions sit closest to a buying decision.

  • Missing: a rival is named and you aren't, which is the gap that moves your share.
  • Engine split: some engines name you and others don't, so the work is engine-specific.
  • Unlinked: you're named without a link while a rival in the same answer is linked, so closing it raises your linked count, not your share.

Firemark tags every prompt as top, middle or bottom of the funnel, so you can filter gaps by stage. One more filter before you start: leave out the prompts whose wording contains your brand name.

A prompt that asks about you by name tends to mention you, and it pads the share without telling you anything about the category. Firemark has a switch for this on every screen, in the CSV and in the MCP.

What do the answers cite?

The sources cited on a gap prompt show where the engine looked when it named your rival, and they're your clearest clue to what you can change.

For each cited site, Firemark counts the answers that cite it at least once, split into answers where you're named and answers where a rival is named and you aren't. Its source-gap view lists the sites that cite others but not you.

Read the cited pages on your top gap prompts and sort them into five types:

  • Your rival's own pages: comparison pages, use-case pages and docs that answer the question directly.
  • Review and listing sites: product pages on G2, Capterra, Trustpilot and similar sites.
  • Roundups and trade publications: "best X for Y" lists and industry articles that name several vendors.
  • Forums and communities: threads where buyers ask each other for recommendations.
  • Your own pages: cited, but not enough to get you named in the answer text.

Look for repeats across prompts. One roundup cited behind six gap answers is one page to work on, not six.

What will increase AI share of voice?

Match the work to the source the engine leans on: a page you own, a third-party page you can get onto, or a listing you can fix.

What will increase AI share of voice?
Gap you seeWhat the answers citeWork to tryHow you'll know it closed
MissingA roundup or trade article listing your rivalsPitch the author with what their list leaves out, plus a fact they can checkAnswers on those prompts name you, and the article is cited with you in it
MissingReview or listing pagesComplete and correct your profile, and ask customers for reviewsAnswers name you, and your profile is cited
MissingA rival's comparison or use-case pagePublish your own page answering that exact question, with facts a reader can verifyYour page is cited on that prompt and the answer names you
Engine splitDifferent sources on each engineWork on the sources the skipping engine citesThe prompt names you on the engines that skipped you
UnlinkedOther sites, while the text names youGive the question a page of yours worth citingYour linked count rises, while your share stays put

For an engine split, check access before anything else. Sites that block OpenAI's search crawler won't be shown in ChatGPT search answers, per OpenAI's crawler documentation. On Google Search, a page must be indexed and eligible to show with a snippet to appear in AI features, per Google's AI features guide.

What does the research say about your own pages?

Published research on your own pages is narrow: one benchmark study found adding citations, quotes and statistics made a page's text more visible in AI answers.

In the GEO paper (Aggarwal et al., KDD 2024), adding citations, quotations from relevant sources and statistics made a source's text more visible in generated answers. The top methods beat the unedited baseline by 41% on one of the authors' measures.

That measure counts how much of a page's text appears in an answer, weighted by where it appears, on the authors' own benchmark. It doesn't count brand names, so it's a reason to make pages citable, not a forecast for your share. Firemark's guide to getting cited by ChatGPT sorts the ChatGPT-side levers by the evidence behind each one.

How do you see whether it moved?

Re-count the same prompts on the same engines, and call a rise only after three comparable runs and a change of 5 points or more.

Three runs, because a single answer is a poor witness. In a SparkToro study published January 28, 2026, 600 volunteers ran 12 prompts on ChatGPT, Claude and Google's AI a combined 2,961 times. There was less than a 1 in 100 chance that ChatGPT or Google's AI gave exactly the same list of brands in any two responses, though the most-named brands kept coming back.

Then check three things, in this order:

  1. The gap list shrank on the prompts you worked on. A total that rose while your target prompts stayed gaps is credit you can't claim.
  2. The basket, the rival list and the engines didn't change between runs. Firemark's piece on what makes a good AI share of voice lists six conditions that move a share with no change in the brand.
  3. The range. Firemark shows a share only on 20 or more answers and draws a 95% range on each engine's own base. A rise that sits inside the last run's range could be sampling alone.

For PR work, count the placement itself. Firemark counts a listed placement only when an answer cites that exact article, so you can see whether the piece you placed is the page the engine uses.

What can't you control?

You can't control which sources an engine picks, when it changes how it answers, or what a signed-in user with history sees.

Firemark collects answers from a logged-out session, from a US location in English by default, so a person signed in with history can get a different answer. Its method page says the ranges cover sampling alone, not an engine changing how it answers.

A count isn't a cause, either. Your rival may publish the same week you do. Firemark records what the answers say, and the work that changes them stays with you.

Frequently asked questions

How long does it take to increase AI share of voice?

Plan on at least three comparable runs after a change before you call a rise, since Firemark calls a movement only after three runs and 5 points. On a weekly schedule that's roughly three weeks of counting, plus however long the engine takes to pick the change up. Firemark's guide to your first AI visibility optimization covers testing one change at a time.

Does publishing more content increase share of voice?

Publishing more content increases share of voice only if the new pages get you named in answers that skipped you, on prompts you track. Pages on topics your basket doesn't cover can't move the count. Check whether a new page is cited on a gap prompt before you write the next one.

Can you increase AI share of voice without a tool?

You can increase AI share of voice without a tool, since the tool only counts whether it rose. By hand, you'd ask each gap prompt on each engine and log who's named and what's cited. Firemark shows a share only on 20 answers or more per base, which is a lot of copying and pasting by hand.

Does Firemark increase your AI share of voice?

Firemark doesn't increase AI share of voice; it counts how often AI answers name and link your brand and your rivals. It shows the gap prompts, the cited sources and whether a listed PR placement is cited. The work that changes the answers is yours.

Should you work on every engine at once?

Work on one engine's gaps at a time when the cited sources differ, so you can tell which work moved which engine. An engine split means the engines disagree about you, so start with what the skipping engine cites.

Work the gaps, then count

Publishing more and hoping is how a share of voice stays flat. Find the answers that named a rival instead of you, work on the pages those answers cite, and count again on the same prompts. The demo report shows share of answers, engines, competitors and cited sources on demo data.

Angelique Swain AI-first SEO Strategist

Written by

Angelique Swain · AI-First SEO Strategist

Angelique Swain is an AI-first SEO strategist with over a decade in organic search across retail, medical and B2B software. Most of her work now is on how AI engines choose which brands to name. She builds fixed prompt sets, re-runs them to see how much the answers vary, and traces what the cited pages have in common. At Firemark she writes about measuring AI visibility and what to do once you can see it.

  • Three stats and evidence grades for getting cited by ChatGPT: a 1 in 100 chance of the same list twice, 60 to 100 asks per prompt and 0.74 for YouTube mentions

    Get cited

    How to Get Cited by ChatGPT, Based on the Evidence

    Which tactics for getting cited by ChatGPT are documented, tested on other engines, only correlated or unsupported, and how to test whether a change moved your numbers.

    · 7 min read

  • Cover reading How Often to Run AI Visibility Checks, in cream type on a dark Firemark ground

    Method and proof

    How Often Should You Run AI Visibility Checks?

    How often to run AI visibility checks, worked out from what a check measures, how many runs a movement needs and what research says about run counts, with a plain rule for daily, weekly or monthly.

    · 6 min read

Find out if AI recommends you

Start tracking your brand. New accounts get one free run.