Why Do Two GEO Tools Show Different Share-of-Voice Numbers?
Two GEO tools report different share-of-voice numbers because share of voice is not a universal measurement — it is the output of a specific prompt set, sampling schedule, and scoring formula, and every vendor makes different choices for all three. There is no canonical "true" number to converge on, so divergence is expected, not a bug.
The three levers that move the number
Prompt set. If Tool A tracks 40 prompts and Tool B tracks 80, they are literally measuring different questions. A brand strong on "best CRM for startups" but weak on "enterprise CRM" will score higher wherever startup prompts dominate the set. Because each tool's default prompts differ, the denominators differ.
Sampling. LLMs are non-deterministic, so each tool captures a different slice of possible answers depending on how many times it runs each prompt and on which days. One tool sampling once and another sampling ten times over a week will see different mention frequencies purely from variance.
Scoring. This is the biggest hidden source of disagreement. Vendors weight mentions differently:
| Scoring choice | Effect on your number |
|---|---|
| Any mention = 1 point | Rewards breadth; inflates frequently-named brands |
| First mention weighted higher | Rewards position; favors the brand engines lead with |
| Cited (linked) mentions only | Ignores name-drops without a source link |
| Sentiment-adjusted | Discounts negative or hedged mentions |
| Deduped per answer vs per run | Changes whether repeated names count once or many times |
Two tools can watch the identical ChatGPT answer and score it 100% vs 25% depending on these rules.
How to compare tools fairly
Equalize what you can. Load the same prompt list into both platforms, align the date window, and — critically — read the raw captured answers side by side. Understanding the difference between a mention and a citation usually resolves half the gap: a name-drop and a linked source are counted differently by different scoring engines. Whatever discrepancy survives after the prompts match comes down to formula, and any credible vendor will document their weighting openly.
What to actually do about it
Stop trying to reconcile the absolute figures. Pick one tool, freeze its methodology, and track your trend and your gap to competitors measured the same way. Relative movement inside a consistent system is trustworthy; cross-tool absolute comparisons are not. If you must present a single headline number to stakeholders, footnote the prompt count and scoring rule that produced it so the figure stays honest.
Frequently asked questions
- Which tool's share-of-voice number is correct?
- Neither is objectively correct — each measures a different prompt set with a different scoring rule. Share of voice is only meaningful relative to a fixed methodology. Pick one tool, keep its methodology constant, and track the trend rather than chasing an absolute figure.
- Does counting mentions vs weighted position explain the gap?
- Often, yes. One tool may count any brand mention as a full point; another weights first-position or cited mentions higher. A brand mentioned frequently but rarely first will score high on the first method and low on the second, producing very different percentages from the same answers.
- How do I compare two tools fairly?
- Load the identical prompt set into both, run them over the same date window, and inspect the raw answers each captured. Differences that remain after equalizing the prompts come down to scoring formulas, which vendors should document explicitly.
Keep exploring
See how AI engines talk about your brand — track mentions across ChatGPT, Perplexity, Claude, Gemini and 5 more. Start with Menra