Energy AI Visibility Index: Report #1
Report #1 of the Energy AI Visibility Index: 600 recorded ChatGPT and Perplexity answers measuring which energy consulting and research firms get named, which get cited without being named, and how far the two engines disagree.
When an energy buyer asks an AI assistant who to hire, some firms get named and others never come up. This is the first edition of a measured index of that visibility, built from 600 recorded AI answers.
Procurement teams, engineers and executives increasingly start vendor research inside an AI assistant. The assistant names a handful of firms, and that shortlist is formed before anyone speaks to a salesperson. Report #1 measures who appears in those answers, who is used as a source without being named, and how far the two engines disagree.
The finding: the best known firms are recommended without evidence
The Index separates two things that every single-score visibility tool combines. Named means the firm appears in the answer a buyer reads. Cited means the engine used that firm’s own website as a source. Those two numbers come apart, and the direction they come apart in says what is wrong.
One group is named far more often than it is cited. The engine recalls the brand without consulting the company:
| Firm | Named | Cited | Difference |
|---|---|---|---|
| BCG | 257 | 81 | -176 |
| Rystad Energy | 217 | 63 | -154 |
| Wood Mackenzie | 239 | 97 | -142 |
| Bain & Company | 250 | 114 | -136 |
| Deloitte | 160 | 39 | -121 |
| S&P Global | 177 | 59 | -118 |
| Accenture | 129 | 14 | -115 |
The other group is the mirror image. These firms are read and not credited: good enough to source an answer, not established enough to earn the sentence.
| Firm | Named | Cited | Difference |
|---|---|---|---|
| Umbrex | 20 | 72 | +52 |
| ModalPoint † | 219 | 244 | +25 |
| Alexander Group | 19 | 42 | +23 |
| AlphaSense | 15 | 37 | +22 |
| EWR Digital † | 106 | 127 | +21 |
| Stramasa | 27 | 48 | +21 |
These are two different problems. A firm named on reputation has brand equity and thin citable substance. A firm cited without being named has the substance and no recognition. The fix is not the same, which is why one blended score cannot direct it.
The Index
All 25 measured firms, ordered by how often they were named across the 600 recorded answers. Ranges are 95% confidence intervals. A dagger marks self-measurement: ModalPoint administers the Index and is measured by the same pipeline as everyone else, disclosed rather than removed.
| Firm | Named | Rate | 95% range | Cited | Cited rate | Prompts |
|---|---|---|---|---|---|---|
| McKinsey & Company | 264 | 44.0% | 40% to 48% | 210 | 35.0% | 23 |
| BCG | 257 | 42.8% | 39% to 47% | 81 | 13.5% | 24 |
| Bain & Company | 250 | 41.7% | 38% to 46% | 114 | 19.0% | 23 |
| Wood Mackenzie | 239 | 39.8% | 36% to 44% | 97 | 16.2% | 21 |
| ModalPoint † | 219 | 36.5% | 33% to 40% | 244 | 40.7% | 22 |
| Rystad Energy | 217 | 36.2% | 32% to 40% | 63 | 10.5% | 20 |
| S&P Global | 177 | 29.5% | 26% to 33% | 59 | 9.8% | 20 |
| Deloitte | 160 | 26.7% | 23% to 30% | 39 | 6.5% | 20 |
| Accenture | 129 | 21.5% | 18% to 25% | 14 | 2.3% | 20 |
| Enverus | 113 | 18.8% | 16% to 22% | 88 | 14.7% | 15 |
| PwC | 108 | 18.0% | 15% to 21% | 50 | 8.3% | 18 |
| EWR Digital † | 106 | 17.7% | 15% to 21% | 127 | 21.2% | 7 |
| HexaGroup | 101 | 16.8% | 14% to 20% | 106 | 17.7% | 11 |
| Fifth Ring | 77 | 12.8% | 10% to 16% | 83 | 13.8% | 6 |
| ADI Analytics | 66 | 11.0% | 9% to 14% | 66 | 11.0% | 10 |
| Jasper Colin | 59 | 9.8% | 8% to 12% | 69 | 11.5% | 8 |
| Adcetera | 49 | 8.2% | 6% to 11% | 43 | 7.2% | 5 |
| Kimberlite Research | 39 | 6.5% | 5% to 9% | 47 | 7.8% | 8 |
| Aranca | 28 | 4.7% | 3% to 7% | 38 | 6.3% | 3 |
| Stramasa | 27 | 4.5% | 3% to 6% | 48 | 8.0% | 4 |
| Mordor Intelligence | 21 | 3.5% | 2% to 5% | 38 | 6.3% | 3 |
| Leadline Marketing | 20 | 3.3% | 2% to 5% | 20 | 3.3% | 4 |
| Umbrex | 20 | 3.3% | 2% to 5% | 72 | 12.0% | 2 |
| Alexander Group | 19 | 3.2% | 2% to 5% | 42 | 7.0% | 3 |
| AlphaSense | 15 | 2.5% | 2% to 4% | 37 | 6.2% | 3 |
Advisory and research are two different markets
The 25 questions split into two kinds: who to hire for go-to-market advisory (360 answers), and who to buy research and buyer intelligence from (240 answers). Almost nobody wins both.
| Firm | Named in advisory questions | Named in research questions |
|---|---|---|
| Enverus | 1.9% | 44.2% |
| Rystad Energy | 21.7% | 57.9% |
| Wood Mackenzie | 25.6% | 61.2% |
| Kimberlite Research | 0.3% | 15.8% |
| McKinsey & Company | 41.4% | 47.9% |
| BCG | 43.9% | 41.2% |
| ModalPoint † | 42.5% | 27.5% |
| HexaGroup | 27.8% | 0.4% |
Enverus moves from 1.9% of advisory answers to 44.2% of research answers. HexaGroup runs the opposite way, 27.8% to 0.4%. An engine that recommends a firm for strategy will not necessarily recommend it for data, and buyers asking the two questions see two different markets.
The two engines do not answer the same way
Across 577 distinct websites cited as sources, 85 were cited by both engines. 274 appeared only in ChatGPT answers and 218 only in Perplexity answers.
Part of that is mechanical: the two engines attach sources differently, so the citation pools are not strictly comparable. The naming counts are, and they diverge too. Leadline Marketing is named in 20 ChatGPT answers and none from Perplexity. Stramasa is named in 27 Perplexity answers and none from ChatGPT. "Are we visible in AI search" has no single answer.
Where the evidence comes from, and a network of sites behind it
Five domains account for a large share of the sources behind these answers, and they are not research firms. They appear in 155 of the 600 answers: 141 of the 300 Perplexity answers and 14 of the 300 from ChatGPT.
| Domain | Answers citing it | ChatGPT | Perplexity |
|---|---|---|---|
| worldmetrics.org | 86 | 5 | 81 |
| zipdo.co | 63 | 9 | 54 |
| roadtooffer.com | 53 | 0 | 53 |
| gitnux.org | 51 | 0 | 51 |
| wifitalents.com | 26 | 0 | 26 |
We fetched the specific pages the engines cited. Four of the five publish the same article under the same URL pattern: worldmetrics.org and zipdo.co both serve a page headlined "Top 10 Best Oil And Gas Marketing Services of 2026", and gitnux.org and wifitalents.com both serve "Top 10 Best Energy Market Research Services of 2026". All four share an identical site taxonomy. None of them discloses a sample size, a method, or who conducted the research. One of the four advertises "proprietary research and verified data" on the same page.
This matters to a buyer for one reason. If you ask an AI assistant who the best energy research firms are, a large share of the evidence behind the answer you get may trace to a templated network of listicles rather than to anyone who has done the research. Checked 2026-09-18; these sites can change.
How Report #1 was measured
The Index reports observed AI answers, never an on-page readiness score. Every number above traces to a stored answer record, dated and pinned to a model version.
- 600 answers, 300 from chatgpt, 300 from perplexity.
- 25 buying questions, each asked 12 times per engine. Repetition matters: AI answers are probabilistic, so a single ask is an anecdote.
- Engines and models: ChatGPT (gpt-5-mini) and Perplexity (sonar), web search enabled, clean-slate calls with no chat history.
- Named and cited are counted separately and never averaged into one score.
What this cannot tell you. At this sample size the Index resolves differences of roughly 11 points; anything smaller is indistinguishable from resampling noise. Real users see further personalisation from their own history, location and account, so treat these as directional visibility rather than a promise of any one person’s answer. Report #1 is a baseline, not a trend: there is no comparable prior period, and the first trend claim will come with Report #2.
Scope. This edition measures the energy consulting, research and marketing segment, which is where these questions land today. It is not a census of operators or oilfield service companies.
Want to know how AI assistants describe your company?
We will tell you which of these questions your company appears in today, and which it is missing from.
Frequently asked questions
What does the Index measure?
How AI answer engines represent energy firms when buyers ask real purchasing questions: who gets named, who gets used as a source, and who never appears. It reports observed answers, dated and model-pinned. It is not an on-page score.
Which engines, and how often?
Report #1 covers ChatGPT and Perplexity, 12 repeats per question per engine, on a monthly measurement cycle. Models are pinned per run and any change is disclosed.
Can a company pay to change its position?
No. Every firm is measured by the same pipeline. ModalPoint and EWR Digital are measured by it too and are marked with a dagger wherever they appear.
Why is being cited without being named a problem?
Because the buyer never sees you. The engine reads your page, uses what it learned, and recommends somebody else in the sentence the buyer actually reads.
Talk to ModalPoint
A 30-minute call to see if ModalPoint is the right firm and whether the timing makes sense. No obligation either way.