---
title: "What does a share of voice spike tell a founder?"
description: "A share of voice spike tells you the measure moved. Name the population, the denominator, and the window, then read the answers behind the spike."
canonical: "https://jaxonparrott.com/answers/when-to-ignore-a-share-of-voice-spike"
last-updated: "2026-11-14"
---

# What does a share of voice spike tell a founder?
A share of voice spike tells you the measure moved. Name the population, the denominator, and the window, then read the answers that produced it. Fund the next test when that reading shows a source a buyer would meet.
Share of voice needs a definition before it needs a celebration. A media report counts the company's share of coverage against a named competitor set. An AI report counts brand mentions in a specified set of answers. Each is a share of its own population. Share of citation is the share of answers that attached a source. Ask which measure rose, against which denominator, over which window.
Check whether the spike survives its own comparison: the same competitors, questions, engines, collection routes, scoring definition, and windows. A new prompt mix or a pile of syndicated copies can move the reported share. The reading that holds is the one on the questions you kept. Failed answers stay in the record as missing. A branded prompt describes a known brand. An unbranded prompt is the discovery test.
Then judge the outcome the work was meant to produce. A correct unprompted recommendation stands on its own. A false claim gets its own correction. A citation matters when the attached source supports the sentence. Keep those observations in separate columns, and let each one justify its own next step.
For a coverage campaign, read who originated the articles. Syndicated announcements do the distribution job, and copies across domains are still that announcement. For an AI-answer claim, read the retained answers and their sources. When the tracked answers cite other pages, you have a map of the sources the sample trusted. When they cite the campaign's pages, you have the placement inside the answer. Either reading tells you the next source to earn.
The note I want before a renewal is specific enough to rebuild the result. Correct the report when the measure is mislabeled. Keep the result when it answers the question you asked. Investigate when the cause is still open, and commit more money when the cause is named. A credible improvement justifies one bounded test. Revenue stays a commercial record. A rise in coverage and citations in the same window is alignment, and a causal claim waits for a test designed to show cause. A reputational error gets its correction on its own schedule.
The Index counts below are domain-class context. AuthorityTech's share of citation guide is the operating metric. You decide what the spike is allowed to change, and you decide it from the reading.

## The Index number

Editorial publications received 20,871 domain-run citations in this release and wire and press-release distribution 744; a run is counted once per domain it cites, so these are not counts of distinct runs. That is 29.1 percent against 1.0 percent of role-classified citations. These are domain-class observations, not article-independence labels or comparative placement returns.

Published November 5, 2026, updated November 14, 2026. Counts from Machine Relations Index release November 5, 2026 (`mri_score_v2.1+2026-11-05+66c560c07279`), runs collected May 10 to November 5, 2026. Dataset: [doi:10.5281/zenodo.23652423](https://doi.org/10.5281/zenodo.23652423).

- Definition: [Machine Relations](https://machinerelations.ai/glossary/machine-relations)
- In practice: [AuthorityTech's share of citation measurement guide](https://authoritytech.io/blog/share-of-citation)

## Related answers

- [How should a founder read what an answer engine says about the company?](https://jaxonparrott.com/answers/what-founders-get-wrong-about-answer-engines)
- [How should a founder read a Machine Relations Index release?](https://jaxonparrott.com/answers/how-a-founder-reads-an-index-release)

## FAQ

### Should I keep measuring share of voice?

Keep measuring it once you have named the population, the competitors, the denominator, and the window. Coverage share, answer mentions, and citation share each describe their own outcome. Use the one that matches the decision.

### Can an AI result matter without an attached citation?

A recommendation, an accurate mention, or a false statement can matter on its own. Keep each outcome in its own column. Revenue comes from the commercial record. Cause comes from a test you designed to show it.

[All answers](https://jaxonparrott.com/answers)