In AI answers on energy, oil majors draw the scrutiny and NextEra draws the clean-transition credit

This snapshot, measured on 2026-09-28, covers 25 answers: five benchmark questions across five assistants. The answers split by question. Asked about the worst environmental record, all five assistants named ExxonMobil, Chevron, BP and Shell together. Perplexity's answer opened with a hedge: "There is no single official ranking of the “worst” oil and gas companies". It then listed those four plus ConocoPhillips and Saudi Aramco.

Asked who is leading the clean transition, all five assistants named NextEra Energy. Perplexity's answer said NextEra "is repeatedly described as the world’s largest producer of wind and solar power, and a premier clean-energy leader." Most answers filled out the list with companies outside the tracked set, such as Ørsted, Iberdrola, Enel and Adani Green.

The ESG-avoidance question produced more hedging. ChatGPT said "there isn’t a single universally “worst” ESG energy company" but used ExxonMobil as its example. Claude and Gemini named no company. Perplexity and Grok named the four majors. The reliability question produced no tracked brand at all: ChatGPT and Claude asked what "utilities" meant, and Gemini and Grok pointed to metrics and local providers.

Shell appeared in 10 of the 25 answers, BP in 9, ExxonMobil in 8, Chevron in 7 and NextEra Energy in 7. Duke Energy appeared in one, Perplexity's clean-transition answer. The four oil majors were credited almost entirely under Sustainability & Ethics. For ExxonMobil, that label covered all eight of its mentions.

That label does not mean the credit was flattering. Five of Shell's ten mentions and five of BP's nine came in the worst-record answers. On the climate-leadership question, Shell and BP appeared mainly with qualifiers. Gemini grouped them under "Mixed Progress". Grok said Shell and BP "generally lag TotalEnergies in pace and scale of renewable buildout".

NextEra is the only tracked brand whose mentions were concentrated on the positive-transition question. It was also credited once each for Financial Performance and once as Not Substantive. Duke Energy is effectively absent, and no utility in the tracked set appeared in the reliability answers.

In the 30-day PeakMetrics news index, Chevron had the most coverage, with 31,127 mentions and 4,138 with the brand in the headline. BP had 17,148 mentions, Shell 15,709, ExxonMobil 12,332, NextEra Energy 6,327 and Duke Energy 5,269.

The sampled headlines are mostly about deals, stock and operations. Shell's included a convenience-retail acquisition and a completed acquisition of ARC Resources. ExxonMobil's included a debt tender offer and an Azerbaijan operator role. Chevron's were investor-position stories and Venezuela agreements. NextEra's were earnings guidance and peer comparisons. Duke's covered peak-demand appeals and data-center costs. Several BP headlines were about other companies or products with similar names, such as Aker BP and a filler grade called DRAGONITE-BP-D501M.

None of the sampled headlines carried the environmental-record or clean-transition framing that dominated the answers.

Chevron is the most heavily covered brand in the index but was named in fewer answers than Shell, BP or ExxonMobil. Duke Energy has thousands of mentions and a headline footprint comparable to NextEra's, yet appeared in one answer. The reliability question, where a utility could have won a place, named none of the tracked utilities. The one answer that listed specific utilities, Perplexity's, named OUC, Con Edison, ComEd, FPL, Tucson Electric Power and SRP, drawing on rankings and awards.

The oil majors show the opposite pattern. Their news footprint is largely financial, while the answers place them in environmental-record and ESG discussions. In this snapshot, press volume and the framing an assistant gives a brand are separate things.

Look at the sources behind the framing. The most-cited domains in this snapshot were energydigital.com, bioplasticsnews.com, violationtracker.goodjobsfirst.org, greenpeace.org and jdpower.com. The source mix was 55% trade, 30% other and 15% review and affiliate. The worst-record answers leaned on violation databases and watchdog reporting. The clean-transition answers leaned on trade rankings and CDP-style lists. Company-issued material was not a leading source.

The practical read is to check where your own brand sits on each benchmark question, not only how much press it gets. Note which third-party datasets and trade lists feed each answer, and whether your data is present and current there. For utilities, a reliability question can return no company at all, and the answers may instead point to state regulators and metrics such as SAIDI and SAIFI.

These are single runs from one date. Live answers vary between runs, so treat this as a baseline to re-measure, not a ranking.