AI Share of Voice: Energy B2B's New Visibility Metric

The metric for a market that has moved into the answer

Share of Voice has always been a proxy for attention: your slice of the advertising, the search rankings or the press coverage in a category. AI Share of Voice applies the same idea to a new surface. It is the percentage of answers from AI engines, ChatGPT, Gemini, Perplexity, Microsoft Copilot and Google's AI Overviews, across a defined set of buyer questions, in which your brand is mentioned, cited or recommended, measured against a named set of competitors.

The reason it matters now is that the answer engine has become the first stop, not the last. Forrester reported that 89 percent of B2B buyers had adopted generative AI as a top source of self-guided research, roughly three times the consumer rate, according to coverage of its data by Demand Gen Report. In 2026 G2 research found that 51 percent of B2B software buyers now begin their purchase research inside an AI chatbot rather than a search engine, with ChatGPT the dominant tool at 63 percent, as reported in G2's research release.

For energy firms this is not an abstract shift. When a procurement lead, an asset manager or an engineer asks an assistant to name credible suppliers for a scope of work, the model returns a shortlist. If your firm is on it, you are in consideration before a salesperson has done anything. If you are absent, you are out, and, unlike a lost search click, nothing in your analytics will ever tell you. AI Share of Voice exists to make that invisible loss visible, so it can be managed rather than guessed at.

Project 54An energy complex lit against the dark. In the answer engine era, being visible when a buyer asks who to trust is the new front of the funnel.

The research journey now starts with a question, not a query

Two forces are compounding. The first is behaviour: buyers prefer to research alone, and generative AI is the most efficient self-service tool ever handed to them. Long before an answer engine existed, analysts noted that B2B buyers spend the majority of the journey away from sales; the assistant simply accelerates that, compressing the early shortlist into a single conversation. Google research cited across the industry in late 2025 found that around 60 percent of B2B buyers now use tools like ChatGPT or Gemini to augment vendor lists, summarise content and surface competitors.

The second force is the search page itself changing under buyers' feet. Google's AI Overviews now sit above the classic links on a large and contested share of queries. Trackers disagree on the exact figure because each samples a different keyword mix, but a defensible 2026 reading is that AI Overviews appear on roughly a fifth to a half of queries depending on the vertical, with BrightEdge data cited near 48 percent of tracked queries and Semrush measuring a rise from about 6.5 percent of queries in January 2025 to a peak near 25 percent before Google recalibrated. Treat any single number as an estimate; the trend is not in doubt.

Gartner sharpened the point in 2024 when it predicted that traditional search engine volume would fall by around 25 percent by 2026 as chatbots and virtual agents absorb queries, advising that companies would need to focus on producing unique, useful content that demonstrates expertise, experience, authoritativeness and trustworthiness. Some 2026 reviews argue that headline drop has not fully landed. The precise magnitude matters less than the direction: attention is migrating from the ten blue links to the single synthesised answer, and energy marketers who only measure the former are now blind to a growing part of the market.

Three numbers on a fixed set of questions

AI Share of Voice is not one number but a small family of them, all computed on the same fixed prompt set so the trend is comparable week to week. Vendors differ on the exact formula, HubSpot, Semrush and Profound each publish materially different methods, so the discipline is to choose one, document it, and hold it steady. Three measures carry most of the signal.

A four step method any energy marketing team can run

The method is simple to state and demanding to do well. First, build the instrument: assemble 50 to 200 questions that real buyers in your category ask, in their language, not yours. Second, fix your competitor set, the five to ten firms you expect to be compared against. Third, run the prompt set across each engine on a regular schedule, capturing the full answer and its cited sources, because answers vary by day, by phrasing and by engine. Fourth, score each answer for presence, position and citation share, aggregate to a share against your competitor set, and watch the trend rather than any single reading.

Two cautions keep the number honest. Answers are probabilistic, so a one-off snapshot is noise; only a repeated, scheduled measurement reveals a trend. And there is no agreed standard, so a figure from one tool cannot be compared with a figure from another. The value is internal and directional: is our presence rising, on which engines, for which questions, against whom. That is enough to steer content and to defend a budget.

The table below sets out the core measures, what each answers, and an energy example, so a team can turn the idea into a repeatable weekly report.

Committee buying, long cycles and thin, technical categories

Energy buying is not a single click. It is a committee decision, often long and high value, made by procurement, engineering and finance functions that each research independently. That is precisely the environment in which an answer engine is most useful to the buyer and most consequential for the seller, because the assistant helps assemble the shortlist the committee then debates. We examine how that committee actually forms and decides in selling into the new energy buying committee, and the AI layer now sits on top of it.

Energy categories are also thin and technical, and that changes the measurement. When a buyer asks who leads subsea inspection, who provides methane monitoring at scale, or who can assure CSRD data for an oil and gas supply chain, the model has fewer credible sources to draw on than it would in a mass consumer category. That cuts both ways: each citation carries more weight, so a single authoritative page can lift your presence sharply, but a single factual error, an assistant claiming you do not operate in a region, does disproportionate damage. This is why sentiment and accuracy belong in the measurement, not just raw mentions.

The prompt set, then, has to be engineered from genuine procurement language, drawn from tender questions, buyer interviews and the intent signals a firm already collects. It is the same discipline we describe in reading energy B2B intent data and buying signals: measure what buyers actually ask, not what you wish they asked. A prompt set built from marketing language measures the wrong market.

Measure first, then optimise, and make it a standing KPI

The first move is to measure before optimising. Optimising for answer engines, the discipline of generative engine optimisation, only pays off if you can see whether it is working, and most teams currently cannot. We set out the optimisation playbook in generative engine optimisation for energy B2B; AI Share of Voice is the scoreboard that tells you whether that work is landing. Doing the optimisation without the measurement is spending without a feedback loop.

The second move is to treat AI Share of Voice as a leading indicator, not a vanity metric. Because buyers shortlist inside the answer before they ever contact sales, a rising presence rate today is a rising pipeline tomorrow, in the same way that intent and attribution feed a long cycle forecast. It belongs on the same board report as the metrics we describe in attribution across the long energy sales cycle, and against the account targets in account based marketing for energy B2B. Reported quarterly against a fixed competitor set, it turns a diffuse anxiety about AI into a number a leadership team can act on.

The third move is posture. An energy brand that is measured, cited and correctly described by the models is a brand that shows up at the exact moment a committee is forming its view, and one that can prove it. That is the growth partner stance we take across our work: visibility, and the revenue it protects, engineered, not assumed.