Share of voice is broken. Here’s what’s replacing it.
Share of voice counts volume in a defined set of outlets. Done incorrectly, it cannot tell you whether a mention was consequential, whether it reached a buyer, or whether it fed the AI answers your clients now read.
Rostrum uses four measures to improve effectiveness in measuring Share of Voice: consequential coverage, buyer reach, citation share and narrative accuracy.
What was share of voice designed to do?
Share of voice was built to compare you against your rivals. It takes a fixed set of publications, counts how often each brand appears, and states your count as a percentage of the whole. The idea came from media buying in the 1960s, and public relations adapted it to quantify earned coverage. It answered one question well: of all the noise in your category, how much of it carries your name?
For a long time it was the best number available. It is objective, comparable across competitors and cheap for a monitoring service to produce. That is why it stuck, and why it’s still central to most communications reports. The problem is what it quietly assumes about who reads your coverage.
What are the three things share of voice cannot see?
Share of voice cannot see whether a mention had any effect. It scores a passing name-check in a roundup, exactly like a long profile a buyer would read before taking a meeting. This creates three key issues for comms teams.
The first is consequence. A mention that moves a prospect towards you and a mention that simply fills a page score identically.
The second is the buyer. General reach rewards a hit seen by a mass audience who might never buy from you over a hit seen by the few dozen people who will.
The third is LLMs. When a client asks an AI assistant to name firms in your category, share of voice alone cannot establish whether you were listed. Instead, it should be augmented by a set of four measures: consequential coverage, buyer reach, citation share and narrative accuracy. We call the set the four-measure standard, and each earns its own section below.
What is consequential coverage and how do you score it?
Consequential coverage is coverage that achieves defined impact. You score each piece on four questions, from zero to three, and add the scores up. The rubric is deliberately blunt, so two people scoring the same article can land close together.
Score every piece of coverage on these four questions:
- Placement: is it in an outlet your buyers read?
- Depth: does it explain your argument, or only mention your name?
- Message: does it carry the specific point you needed to land?
- Prompt: does it give a reader a concrete reason to act?
A piece scoring ten or more out of twelve is consequential and belongs in the board pack. Anything lower is activity you logged.
Why measure buyer reach instead of audience reach?
Buyer reach counts the people with real potential to become clients. Audience reach counts everyone a publication touches, most of whom likely sit outside your market. Buyer reach counts only the slice of that audience inside your addressable set of accounts.
This is the same logic that sits behind account-based PR, which our PR team already runs against named target lists rather than broad awareness. A niche trade title whose readers each control a relevant budget is worth more to you than a national newspaper piece seen by a mass audience that won’t buy from you.
What is citation share in AI answers?
Citation share is how often an AI assistant names you when asked about your category. It is the version of share of voice that survives the shift to AI-mediated search. Writing in Forbes on 8 September 2026, Parry Headrick argued that citation share is the new share of voice, because language models are now reading the articles humans used to.
There is a real gap here, and it is largely going unmeasured. In the Muck Rack 2026 State of PR, published on 9 July 2026, 39 per cent of PR teams reported that they were not measuring AI search success at all. The best methodology is to test the main AI assistants with questions buyers ask when choosing a firm in your category, then record how often they name you and whether these models recommend you. Our AI visibility audit does exactly this for a firm and the competitors it names.
Volume and consequence can diverge here. In October 2026 we audited a lender whose press coverage we track, and what the assistants said about it did not match that coverage. Our AI visibility audit checklist describes what we found and how to run the same check.
What is narrative accuracy, and are you achieving the right answers?
Citation share tells you that you were named in an AI search. Narrative accuracy tells you whether what was said next was correct.
You measure it the same way you measure citation share, with one extra step. Ask the assistants the questions your buyers ask, then read the whole answer and mark each factual claim as right, outdated or wrong. A firm that is cited often but described with an out-of-date leadership team, the wrong specialism or a five-year-old controversy has a narrative accuracy problem that a volume metric won’t surface.
This is an important factor that’s undermeasured and undervalued. It is also the one that can shape a buyer’s decision before they ever reach your site.
How do you run all four without a new budget line?
You already pay for most of this. Consequential coverage and buyer reach are re-scores of the monitoring you buy today, but done with a properly thought out strategic approach to Share of Voice rather than a mention count.
| Measure | What it asks |
|---|---|
| Consequential coverage | Did the mention do a job you can name? |
| Buyer reach | Did it reach people who could buy? |
| Citation share | Do AI assistants name you in your category? |
| Narrative accuracy | Is what the AI says about you correct? |
The reallocation is straightforward. Stop paying for a weekly share-of-voice percentage nobody acts on, and spend that same analyst time on these four measures. Lead your next board report with coverage quality and visibility in AI answers, using volume as supporting context. Build these measures into your existing reporting so the same budget produces more useful insight.
What do we still use share of voice for?
Rostrum still uses share of voice as a trend line. Over quarters, a rising or falling share tells you whether your presence in a category is growing, and that is a useful early signal. The mistake is treating it as an outcome when it is really only an output.
The AMEC Barcelona Principles have held since 2010 that advertising value equivalents do not measure the value of communications. The four-measure standard extends that same discipline to share of voice: keep it as a directional indicator, and judge success on consequence, buyer reach, citation share and narrative accuracy. We are moving our own client reporting onto these four measures from January 2027.
Frequently asked questions
What is share of voice in PR?
Share of voice in PR is your brand’s share of all coverage across a defined set of outlets over a set period, usually shown as a percentage. It measures how much of the category conversation carries your name. It is a useful trend line and a poor outcome measure on its own.
How do you measure PR effectiveness?
You measure PR effectiveness by scoring whether coverage did a job you can name, reached real buyers, earned citations in AI answers and described you accurately. Volume metrics tell you how busy you were. The four-measure standard tells you what the work changed.
What should replace AVE and share of voice?
Advertising value equivalents should be dropped entirely, a position the AMEC Barcelona Principles have held since 2010. Share of voice should be demoted to a directional indicator. In its place, run consequential coverage, buyer reach, citation share and narrative accuracy.
How do you measure whether AI describes your brand accurately?
You prompt the main AI assistants with the questions your buyers ask, then read each answer and mark every factual claim as right, stale or wrong. We call this narrative accuracy. A brand can be cited often and still be described with outdated facts, which is the problem this measure exists to catch.
