AI Energy Choice
in Italy

AI Choice Intelligence

This is not just AI Visibility. It measures what happens after presence.

Wave 1 is a versioned baseline. Future waves can measure change against the same reference point rather than showing disconnected snapshots.

Executive summary

Presence and recommendation are not the same thing.

The same category produces different competitive hierarchies depending on the AI surface. A supplier that is highly visible is not necessarily the supplier most frequently pushed toward choice.

Study base. 732 ChatGPT responses, 733 Gemini responses, 553 Google AI Overview responses and 571 Google AI Mode responses. “Brand present” is not equivalent to “brand recommended”, market share or a sale.

Research proof

The analytical perimeter behind the published results.

Wave 1 separates the Primary Generic collection design from the supplier corpus used for the published analyses. They are different analytical perimeters and should not be treated as the same denominator.

Decision value

The signal that presence-only monitoring does not show.

Knowing that a supplier appears in AI answers is only the starting point. The observed competitive profile depends on where it appears, at which stage of choice and with what strength relative to competitors.

When Exposure and Recommendation diverge, the gap becomes strategically relevant because it identifies where further investigation should start — content, PR, search, sources or positioning — without pretending that the research alone proves causality.

Guardrail: a gap identifies where to investigate. It does not by itself explain why the gap exists.

Who it is for

For teams that need to read the market before deciding.

AI surfaces

One category. Four different competitive hierarchies.

The Exposure leader changes with the observed surface. ChatGPT, Gemini, Google AI Overview and Google AI Mode do not return the same competitive ordering.

What this means. There is no single competitive ranking that is valid across all AI systems.

Operational implication. Monitoring only one platform can hide competitors that systematically emerge elsewhere.

Reading rule. Leadership should first be interpreted by surface before any cross-surface synthesis.

Cross-surface Exposure

How uniform is supplier presence across AI systems?

Cross-surface average Exposure summarizes presence across the four surfaces using a simple descriptive mean. It is useful for orientation but it is not a weighted market index and it does not replace surface-level analysis.

The complete report contains values, positions and cross-surface comparisons for all observed suppliers.

Recommendation

Being present does not mean being recommended.

Recommendation identifies an explicit selection or preference toward a specific supplier. In Wave 1, Recommendation is published only for ChatGPT, where the decision baseline uses Recommendation Engine v1.4.1.5.

Cases marked model_review_required are not converted into false negatives. The published value is deliberately conservative and therefore acts as a lower bound.

What this means. Two suppliers with similar Exposure can have very different decision strength.

What this changes. Visibility alone can either overstate or understate a supplier’s role in choice.

Energy Choice Map

Where presence and decision strength separate.

A ranking orders. A map distinguishes competitive profiles. By crossing average cross-surface Exposure with ChatGPT Recommendation, different competitive configurations emerge.

The public page explains the map logic but does not publish the full supplier coordinates and clusters. Those are part of the premium report.

Public example. In the observed Wave 1 data, A2A Energia shows stronger ChatGPT decision strength than its average cross-surface position would suggest, while Enel Energia shows much stronger presence on ChatGPT than on Gemini.

Use cases

From benchmark to the next question worth investigating.

Full report

The public page orients. Premium opens the competitive benchmark.

Method and core public findings remain readable here. The complete report adds full rankings, coordinates, comparisons and brand-by-brand profiles required for deeper market analysis.

Final prices. No VAT or other tax is added at checkout. Tax treatment is determined from billing details.

Methodology

Design before result.

The pipeline separates simple presence from decision strength and keeps the semantics of each AI surface distinct.

Pipeline

Text blocks → entity linking → claim extraction → claim validation → deterministic supplier aggregation → research metrics.

Exposure

Observed supplier presence in a response.

Recommendation

Explicit supplier selection/preference under the ChatGPT Wave 1 decision semantics.

Lower bound

Unresolved model-review cases are not silently converted into negatives.

Surface separation

ChatGPT decision semantics are not automatically transferred to Gemini or Google surfaces.

Transparency. Recommendation is published only for ChatGPT in this wave. Gemini and Google surfaces use Exposure without importing ChatGPT’s decision semantics.

Conclusions

AI choice is not a single ranking.

The most useful result is not the declaration of one absolute winner. It is the observation that presence, recommendation and surface produce different competitive readings.

For an energy supplier, the practical question therefore becomes: “On which surface, at which stage of choice, and in what role are we appearing?”

FAQ

Purchase, licensing and use of Wave 1.

Citation & use

How to use this research.

The public page is an editorial summary. Complete rankings, complete tables, Energy Choice Map and the full PDF are part of the premium report.