Quick answer
AI Share of Voice measures how much of the AI answer space your brand owns compared with competitors. The simplest version is:
AI Share of Voice = your brand mentions / total tracked brand mentions across the competitor set x 100
That formula is only useful when the setup is clean. The same brand can look strong or weak depending on aliases, competitor list, prompt intent, country, language, platform mix, date range, and whether the score counts only mentions or also weights rank, impressions, or source exposure.
For ReachLLM teams, measure AI Share of Voice this way:
- Freeze the prompt set before reporting.
- Clean brand aliases and competitor classifications.
- Separate branded, unbranded, comparison, and implementation prompts.
- Report platform-level SoV before blending engines.
- Pair SoV with Average Rank, citation rate, sentiment, and raw answers.
- Turn every SoV movement into one owned-page, source, PR, or prompt-hygiene decision.
- Re-measure the same prompt group after the work ships.
Do not report AI SoV as a market-share number. It is a measured answer-space signal from a defined sample.
What AI Share of Voice actually measures
Semrush's scheduled source for this article defines AI Share of Voice as a brand's visibility compared with competitors in AI search. It also gives the basic mentions-over-category formula and shows that tool-specific versions can factor in position, platform, and search-volume context.
That is the right starting point. It is not the whole measurement discipline.
| Layer | What it answers | What can distort it |
|---|---|---|
| Mention share | How often the brand appears versus competitors. | Alias errors, competitor setup, prompt mix. |
| Position-aware share | Whether the brand appears early enough to matter. | Ranking formulas that differ by tool. |
| Platform share | Which answer engines favor or skip the brand. | Engines using different retrieval and source behavior. |
| Topic share | Which themes the brand owns or loses. | Broad topics hiding high-intent buyer prompts. |
| Source-backed share | Whether mentions are supported by citations. | Citations counting links, not recommendations. |
| Trend share | Whether the answer space is moving over time. | Prompt-set, market, or formula changes mid-period. |
The first rule is to label the sample. "Our AI SoV is 18%" is incomplete. "Our unbranded vendor-evaluation AI SoV across 40 US English prompts in ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews is 18% for August" is a measurement.
Start with the competitor set
AI SoV is comparative, so competitor hygiene comes before analysis.
Before you report the number, review these fields:
| Setup check | Why it matters |
|---|---|
| Brand aliases | A model may say "IBM", "International Business Machines", or a product name; decide what counts. |
| Direct competitors | These should affect the denominator because buyers compare them against you. |
| Indirect competitors | These may belong in analysis, but not always in the main SoV score. |
| Other companies | AI answers often name adjacent tools; track them before promoting them into the competitor set. |
| Misspellings and unrelated entities | A shared acronym can inflate or deflate the wrong company. |
| Acquired or renamed products | Decide whether legacy names roll into the parent brand. |
ReachLLM's competitor workflow is designed around that distinction. The platform can show named competitors and other detected companies, but the operating metric is only meaningful after the team decides which entities belong in the comparison set.
If the competitor list changes during the month, start a new baseline. Do not show the new number as a continuation of the old trend.
Freeze the prompt set before measuring
Prompt scope is the second source of distortion. A brand can dominate branded prompts and still lose the buying conversation.
Split the prompt set like this:
| Prompt group | Example question | How to read SoV |
|---|---|---|
| Branded | "What is ReachLLM?" | Entity accuracy and brand fact coverage. |
| Category discovery | "Best AI visibility platforms for B2B SaaS" | Whether the brand is present in the market conversation. |
| Comparison | "ReachLLM vs Semrush AI Visibility" | Whether buyers see a fair, current alternative view. |
| Feature evaluation | "Which tool tracks AI citations and sentiment?" | Whether product capabilities are understood. |
| Implementation | "How do I improve AI citations for my website?" | Whether educational authority supports future recommendations. |
| Local or vertical | "AI visibility agency for UAE real estate brands" | Whether market proof and local sources are visible. |
Report SoV by group first. A blended number can hide the problem. For example, branded prompts may lift the overall score while unbranded buyer prompts still list three competitors and omit the brand.
ReachLLM runs tracked prompts against enabled platforms on a schedule and analyzes each answer for brand mentions, competitor mentions, sentiment, citations, and source evidence. That fixed prompt set is what makes SoV useful as an operating metric rather than a one-off manual test.
Report platform-level SoV before blending engines
Semrush's article notes that AI SoV can vary by platform because each system builds answers and retrieves sources differently. Treat that as a reporting requirement, not a footnote.
At minimum, show:
| Platform view | Decision it supports |
|---|---|
| ChatGPT | Are conversational recommendations naming the brand? |
| Gemini | Is the brand visible when Google-adjacent answers synthesize category sources? |
| Perplexity | Which citations and source pages appear in answer-style search? |
| Claude | Are longer-form recommendation answers representing the category accurately? |
| Google AI Overviews | Are Google-visible pages and snippets supporting the answer? |
If one platform falls behind, do not call it a "ChatGPT problem" or a "Gemini problem" yet. Open the raw answers and cited sources. The cause may be missing content, weak source authority, country mismatch, crawl/indexing limits, old third-party data, or a prompt that is interpreted differently on that engine.
Pair SoV with rank, citations, and sentiment
AI SoV tells you whether the brand appears relative to competitors. It does not tell you whether the appearance is good.
Use paired diagnostics:
| Pattern | Interpretation | First action |
|---|---|---|
| High SoV, weak Average Rank | The brand appears, but competitors lead. | Read ranked lists and strengthen differentiation. |
| High SoV, low citation rate | AI systems know the brand but do not expose owned evidence. | Improve cited pages, source clarity, schema, and internal links. |
| High SoV, neutral sentiment | The brand is visible without a strong reason to choose it. | Add proof, use cases, comparisons, and current claims. |
| Low SoV, strong own-domain citations | Owned pages are useful but not tied clearly to the brand. | Rewrite answer-first sections and product context. |
| Low SoV, competitor source dominance | Third-party sources repeatedly support rivals. | Pursue legitimate PR, directory, partner, or review updates. |
| Volatile SoV | The answer sample is unstable. | Extend the measurement window before declaring a trend. |
This is why raw answer review still matters. A chart can say the brand is gaining share while the underlying answer describes it as a secondary option, cites an old page, or repeats a stale limitation.
Turn SoV movement into a work queue
The useful output of an AI SoV review is not the number. It is the work that should happen next.
Use this review sequence:
- Pick the highest-intent prompt group where SoV is weak or falling.
- Open the raw answers for the affected platforms.
- List the competitors that appear ahead of the brand.
- Inspect the sources and citations attached to those answers.
- Decide whether the gap is owned content, third-party source, technical access, sentiment, or prompt hygiene.
- Assign one fix.
- Record the metric expected to move.
- Re-run the same prompt group after the fix is live.
The fix should match the evidence:
| Evidence | Better fix |
|---|---|
| Competitors win a missing feature prompt | Add or improve the relevant product/use-case section. |
| AI cites an old owned page | Refresh the page and make the current answer easier to extract. |
| AI cites third-party lists that omit the brand | Pursue legitimate inclusion or create a source-backed comparison asset. |
| AI names the brand with vague positioning | Update brand facts across homepage, docs, llms.txt, and profiles. |
| AI skips the brand in local prompts | Add market-specific proof and sources instead of generic copy. |
| SoV moved because competitors changed | Reset the baseline and document the setup change. |
Google's people-first content guidance is a useful guardrail here: do not create pages only because a topic gap appeared in a dashboard. First ask whether an existing page can absorb the answer. Publish a new article only when it adds original, useful information for a real buyer or operator.
A 30-minute weekly AI SoV review
Run this once a week if the team is actively shipping GEO work.
| Minute | Step |
|---|---|
| 0-5 | Confirm prompt set, country, language, platform mix, date range, aliases, and competitors. |
| 5-10 | Review AI SoV by prompt group and platform. |
| 10-15 | Pair SoV with Average Rank, citation rate, sentiment, and source movement. |
| 15-20 | Open raw answers for the largest gain, largest loss, and highest-intent zero-share prompt. |
| 20-25 | Classify gaps as content, source, technical, sentiment, competitor setup, or prompt hygiene. |
| 25-30 | Assign one owned fix and one source or PR action; set the re-measurement date. |
The review should end with a sentence like:
"For unbranded vendor-evaluation prompts in the US, our AI SoV is 14% this week, down from 18%, mainly because Gemini and Perplexity cite two competitor comparison sources. We will update the agency comparison page and pitch one legitimate source before the next Friday run."
That is more useful than a dashboard screenshot.
Where ReachLLM fits
ReachLLM is built for teams that need AI visibility measurement connected to execution. It tracks prompts across enabled AI platforms, analyzes brand and competitor mentions, calculates Visibility Score, Share of Voice, Average Rank, sentiment, citation rate, and source data, and keeps raw responses available for review.
The execution layer matters. If SoV is weak because a competitor owns a cited source, the next action may be PR outreach. If SoV is weak because an owned page is found but not cited, the fix may be page structure, answer-first copy, schema, llms.txt, or internal links. If SoV is weak because the prompt set is wrong, the fix is measurement hygiene.
Do not optimize the graph. Optimize the answer evidence that creates the graph.
FAQ
What is AI Share of Voice?
AI Share of Voice is the percentage of AI answer visibility your brand owns compared with competitors for a defined prompt set, platform mix, market, and date range. The simplest formula is your brand mentions divided by total tracked brand mentions across the competitor set.
Is AI Share of Voice the same as market share?
No. AI Share of Voice is a measured answer-space signal, not a market-share estimate. It depends on the prompts, competitors, platforms, geography, language, date range, and formula used.
Why can AI Share of Voice change when nothing shipped?
It can move because AI systems refresh sources, competitors publish new content, prompts are interpreted differently, the competitor set changes, aliases are cleaned up, or the tool formula weights position or demand differently. Always check setup changes before treating movement as performance.
Which metrics should be reported with AI Share of Voice?
Report AI Share of Voice with Average Rank, citation rate, cited sources, sentiment, raw answer examples, prompt group, platform split, and the shipped work tied to the period. That combination shows whether the brand is visible, trusted, preferred, and improving for the right reasons.
How does ReachLLM help teams improve AI Share of Voice?
ReachLLM connects Share of Voice to prompt-level evidence, competitor mentions, cited sources, sentiment, raw responses, and execution workflows. Teams can use that evidence to ship content, page updates, schema, llms.txt, GEO audit fixes, PR outreach, and follow-up measurement.
Sources reviewed
- Semrush, "How to measure AI share of voice using Semrush," July 17, 2026: https://www.semrush.com/blog/how-to-measure-ai-share-of-voice/
- ReachLLM Docs, "AI Visibility Tracking": https://docs.reachllm.com/guides/ai-visibility-tracking/
- ReachLLM Docs, "Understanding the Scores": https://docs.reachllm.com/guides/understanding-the-scores/
- ReachLLM Platform page: https://www.reachllm.com/platform
- Google Search Central, "Creating helpful, reliable, people-first content": https://developers.google.com/search/docs/fundamentals/creating-helpful-content
- Google Search Central, "Optimizing your website for generative AI features on Google Search": https://developers.google.com/search/docs/fundamentals/ai-optimization-guide