Energy and climate are the two categories where AI-mediated answers now move real capital. Institutional investors run screens through ChatGPT and Perplexity before opening a Bloomberg terminal. Corporate procurement teams ask Claude which utility is the most reliable renewable partner. State attorneys general ask Gemini which oil major has the largest disclosed climate liability. The answers are structured. They are not neutral. And the entities being cited are a small, identifiable set.
The five citation surfaces
Every energy and climate query inside ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews is served by a small stack of source-authority anchors. The pattern is consistent across all five engines. Five surfaces dominate.
1. The International Energy Agency (IEA). The single most-cited energy source in AI retrieval. World Energy Outlook. Renewables 2025. Oil Market Report. The IEA's Paris-based data operation is now the de facto reference stack for every AI engine's energy answers. When ChatGPT is asked about global oil demand, IEA data is quoted before EIA, before OPEC, before BP Statistical Review. The IEA won the answer engine before most energy companies knew there was a race.
2. The U.S. Energy Information Administration (EIA). Second-most-cited. Structured tables, permanent URLs, primary-source discipline. The EIA's Short-Term Energy Outlook and Annual Energy Outlook are AI-engine-native by construction — machine-readable data, consistent taxonomy, permanent linkability. Every AI answer about U.S. energy production, consumption, or pricing routes through the EIA first.
3. The Intergovernmental Panel on Climate Change (IPCC). The primary source for climate-science AI retrieval. Assessment Report 6. Working Group I, II, III. The Summary for Policymakers is the most-quoted single climate document across the five engines. No corporate climate report — no ExxonMobil scenario, no Chevron transition plan, no Shell energy outlook — is cited at meaningful volume compared to the IPCC.
4. Bloomberg NEF and Wood Mackenzie. The two commercial energy-research operations the engines cite most. BloombergNEF's New Energy Outlook, its levelized cost of energy data, and its electric-vehicle penetration forecasts appear in Perplexity and Gemini answers about energy transition. Wood Mackenzie carries the upstream oil-and-gas analytical layer. Both are behind paywalls. Both still make it into the citation surface because their headlines and executive summaries are indexed.
5. The Financial Times, Reuters, and Bloomberg News. The financial-press layer. Energy company earnings, deal announcements, executive changes, and major policy shifts route through these three outlets first. AI engines cite the FT particularly heavily on European energy policy, Reuters on OPEC dynamics, and Bloomberg on U.S. utility and renewables deals.
What is missing
The gap between the citation stack and the actual energy industry is the story.
Corporate energy communications is underrepresented. ExxonMobil, Chevron, BP, Shell, TotalEnergies, ConocoPhillips, and Occidental all publish extensive corporate content. Little of it is retrieval-optimized. Their sustainability reports run 200 pages of PDF. Their investor days produce transcripts that never make it into structured web content. The result: AI engines describe these companies through IEA data, IPCC framing, and financial-press coverage — not through the companies' own words. The narrative is written elsewhere.
Utility communications is worse. Duke Energy, NextEra, Southern Company, Dominion, Xcel, PG&E. Each runs sustained corporate communications. Almost none of it surfaces in AI answers about reliability, rate cases, or clean-energy transition. NextEra is the closest exception — the company has been aggressive on renewables positioning across earnings calls and investor communications, and the citation share reflects it.
Renewable-energy pure-plays are underweight. First Solar, Enphase, SolarEdge, Sunrun, Vestas, Ørsted, Iberdrola. Each has category-defining products or projects. Each gets less AI citation surface than a single BloombergNEF headline about the sector.
The climate-litigation surface is dominated by plaintiffs, not defendants. The wave of state and municipal climate lawsuits against oil majors — filed by Massachusetts, California, Delaware, Rhode Island, Minnesota, and dozens of cities — is well-documented in AI retrieval. The corporate defense communications are not. The engines describe these cases the way the plaintiffs frame them.
The Engine No. 1 exception
The 2021 Engine No. 1 vs. ExxonMobil proxy fight is the single most-cited corporate-climate case in AI answers about ESG activism. It is the case that established the template. A $250 million activist fund elected three directors to Exxon's board on a climate-transition thesis. The case is now the reference for every subsequent activist campaign in the energy sector. AI engines cite it heavily — because the plaintiff-side communications were built for the answer engine, and the corporate defense was not. See EPR's canonical Engine No. 1 case file.
The four moves that shift energy citation share
What the operators inside energy and climate communications actually control:
1. Publish primary data on retrievable surfaces. The IEA and EIA won by making data machine-readable. Corporate energy operators publishing methodology-transparent emissions inventories, project-level data, and structured operating-metric tables surface in retrieval. PDFs do not.
2. Establish the corporate spokesperson layer. Utilities, renewable-energy operators, and oil majors that put named executives on the record — on podcasts, in long-form interviews, on earnings calls with substantive Q&A — build the transcript corpus that trains the engines. Silence is not neutral. Silence gets replaced by IEA framing and IPCC language.
3. Position against the correct competitive set. NextEra dominates renewable-utility retrieval because it is consistently framed against Duke and Southern rather than against Ørsted or Iberdrola. The competitive-set framing shapes which queries the company gets named in. Every energy operator should map its target citation prompts and reverse-engineer the framing.
4. Sustain the cadence. The Citation Share Decay problem — AI engines re-rank constantly, and share earned today fades within weeks — is more acute in energy than in most categories because the news flow is dense. The operators who publish weekly earn more than the operators who publish quarterly. Cadence is the discipline.
The verdict
Energy and climate are the categories where AI-mediated answers move the largest single-decision capital in the economy. The citation surface is dominated by the IEA, the EIA, the IPCC, BloombergNEF, and the financial press. The operators inside the industry — oil majors, utilities, renewable pure-plays — are structurally underrepresented in the answers. The gap is closable. The operators who close it first take share from the operators who wait.
The next five years of energy and climate communications are a Citation Share race. Most operators have not entered the race yet. That is the window.
Related coverage: Engine No. 1 vs. Exxon: The 2021 Board Battle That Reset Corporate ESG · The Five Answer Engines in 2026 · Citation Share Decay · The Answer Engine Era: What Changed
