GolOps Lab · Research center

Language models.
How do they choose?

We study how language models shape perceptions of companies, industries, and countries — and build a methodology for measuring it externally. This affects trust, supplier shortlists, and economic choice.

Get the research Propose a partnership

6language models
under observation

12+industries
being monitored

Quarterlyindex map updates

Scroll down

Flagship research 2026

63%

of ChatGPT responses about Russia's largest companies contain a negative assessment when foreign partners inquire

10 companies · 80,000 queries · 6 language models · 4 marketsGolOps Lab · April 2026

63%

average negative sentiment of responses

48%

contain a direct «avoid» recommendation

4.2×

sentiment gap vs. Western competitors

companies that systematically manage their narrative in language models

Methodology

A methodological foundation for a new field

GolOps develops an open methodology for measuring reputation and choice in the language model environment. At its core — an external, model-agnostic evaluation of AI representation that accounts for geography, choice scenarios, and the competitive landscape.

01 · Standards

GolOps-LLM standards

Open standards for describing and verifying AI representation.

LLM-01 — Audit methodology
LLM-02 — Sampling protocol
LLM-03 — Reporting standard

02 · Indices

Index system

An index system for measuring an entity's position in the AI environment.

CCI — Choice Control Index
PI — Participation Index
WI — Win Index

03 · Foundation

Mathematical foundation

The models that underpin external, model-agnostic evaluation of AI representation.

Weighted sentiment model
Narrative bias analysis
Cross-geography comparison

04 · Application

Applied layer

Tools that apply the methodology for companies and institutions.

Corporate monitoring
Management reporting
Government monitoring

Subject of research

How language models become a choice mechanism

Shaping perception

How language models prioritize and shape the image of companies, industries, and countries for foreign audiences.

Economics of trust

How AI output influences commercial decisions, shortlisting, and strategic partner selection.

A new layer of reputation

Language models become a new layer of business and cultural reputation — on top of media, analytics, and the public field.

Lab focus areas

Three levels of one environment

Companies

How AI systems represent businesses in a competitive environment. A company's image in model responses for foreign partners, investors, and clients.

Industries

How key sectors of the Russian economy appear in international AI recommendations. The structural gap with global counterparts.

Countries

How AI representation of national economies and the export layer is formed. The geopolitical lens as a systemic effect.

Publications and research

Two formats. One methodology.

Industry research sets the market context and benchmarks for the entire sector. Individual research gives a company a precise measurement tool and an action plan — for a specific company, market, and decision-maker.

GolOps Lab · Industry research · 2026

2026 Q1

How ChatGPT describes Russian business to foreign partners

April 2026 · 10 companies · 80,000 queries · 6 language models · 4 markets

63%

of responses contain a negative assessment

«Language models frame responses through the lens of sanctions and geopolitics for any foreign partner query — regardless of topic.»

48%

direct «do not engage» recommendations

«The algorithm makes a commercial decision before first contact with the company — before a call, before a meeting.»

4.2×

gap with Western competitors

«Shell — 12% negative responses. Russian companies — 63%. The gap is structural, not situational.»

Gazprom · Rosneft · Lukoil · Norilsk Nickel · Sber · VTB · Rosatom · Russian Railways · Severstal · Rostec

View research

GolOps Lab · Individual research · 2026

2026 Q1

Gazprom's image in language model responses to German top executives

April 2026 · Germany · GPT · Persona: CFO · 30 queries

87%

of responses — purely geopolitics

«Technology, the LNG program, and infrastructure — are not mentioned in any of the 30 simulated queries.»

18/30

«do not engage» recommendations

«60% of CFO queries received a direct negative recommendation — with no analysis of the commercial proposal.»

23×

gap with Equinor on the reliability metric

«Equinor: 71% positive reliability mentions. Gazprom: 3%. The gap is systemic.»

View research

Get research for your company or industry

Industry research is published quarterly. Individual research is on request.

Request research Industry report

Publications

What the lab publishes

01Company research
02Reviews of sectors of the Russian economy
03Russian export indices across geographies
04Methodology descriptions
05Regular measurements and time-series comparisons

Team

Who stands behind the research

Behind every study — specialists with experience in reputation management, quantitative methods, and data analysis.

Alexander Palchikov

Founder, GolOps

Founded GolOps with a focus on managing brand AI selection in the language model environment. Initiated Russia's first systematic study of the largest corporations' reputation in AI output.

Alexey Arsenyev

Head of research

Specializes in quantitative methods of text data analysis and sentiment evaluation in language models. Responsible for methodological verification of research and the algorithmic basis of GolOps-LLM standards.

Alexander Abramov

Director of corporate solutions

Experience in strategic communications and corporate reputation management. Responsible for adapting the methodology to corporate clients' needs and developing monitoring systems for large organizations.

Who it's for

Three groups for whom this matters

Companies

To understand and strengthen your position in the AI environment. See the gap with competitors and manage the narrative before it becomes systemic.

Government and institutions

To see where the country and its industries are losing or strengthening their representation in AI recommendations. Shape information environment policy based on data.

Academic partners

To jointly develop a new discipline of AI representation measurement — with methodology verification and publications in peer-reviewed journals.

PARTNERSHIP AREAS

Priority areas for partnership

GolOps Lab is open to partnerships with research centers, universities, analytical teams, and development institutions. We are interested in joint projects at the intersection of data, AI, reputation, exports, and choice theory.

SOCIOLOGY · CULTURE

Sociology and cultural studies

Research into how language models shape cultural perceptions of companies, industries, and countries, and how this affects trust and choice.

AI · MODEL BEHAVIOR

AI and language model behavior

Analysis of how language models prioritize, generate recommendations, and reproduce stable choice patterns across scenarios.

STATISTICS · MEASUREMENT

Statistics and reputation measurement

Developing reproducible approaches to measuring the reputational position of companies, industries, and countries in the language model environment.

NARRATIVES · HUMANITIES

Narratives and cultural shifts

Studying how language models shift narratives about national brands, industries, and economic roles across markets.

ECONOMICS · CHOICE THEORY

Economics of trust and choice

Research into the link between AI representation, trust, shortlist mechanics, and economic decisions at the level of companies and markets.

MATH · ALGORITHMS

Mathematical models of narrative bias

Building models to detect bias, run comparative analysis of AI responses, and externally evaluate reputation across geographies and choice scenarios.

Partnership & Research

GolOps Lab creates the language for measuring AI representation

We translate the language model environment into a system of indices, research, and strategic conclusions — for companies, industries, and countries.

Get the research Contact the lab

Language models.How do they choose?