The verdict in three sentences
Trade finance banks, counterparties, NGOs and recruits now put their first questions to AI assistants like ChatGPT, before they open your website or call your treasurer. The answer they get is assembled from what the public web says about your company, and a discreet Luxembourg trading house is often missing from it or described from old articles and poorly linked registry data. A dated diagnostic on around twenty questions costs EUR 2,500 to 6,000 excl. VAT as a 2026 ballpark, and it is the sensible starting point before committing a larger budget.
Who asks AI assistants about your trading house, and about what
Commodity trading sells very little through marketing. The ecosystem that conditions your business, however, checks everything: the onboarding officer at a bank, the KYC team at a refiner, the NGO researcher, the trader weighing a move from Geneva to Kirchberg. Luxembourg adds its own difficulty: many groups house a holding, a treasury, a shipping arm or a regional hub here, and the line between entities is blurred for an outsider.
| Question asked to an AI (example) | Who asks it | What the AI must be able to find |
|---|---|---|
| "Which metals trading companies are based in Luxembourg?" | Counterparty, analyst | Offices page and factual description of the business |
| "Which group does X Luxembourg SA belong to?" | Bank KYC team | Entity chart and RCS numbers |
| "Does X publish a sanctions compliance policy?" | Bank, credit insurer | Dated compliance page, no overclaiming |
| "Who runs X and since when?" | Bank, journalist | Governance page with roles and dates |
| "Does X have a recent sustainability report?" | NGO, industrial customer | Downloadable report with publication year |
| "How does X handle cobalt or cocoa sourcing?" | NGO, downstream buyer | Published responsible sourcing policy |
| "Working at X in Luxembourg: roles, languages, culture?" | Trader or operations candidate | Careers page, open roles, working languages |
| "What is the difference between X Luxembourg and X Trading in Geneva?" | Counterparty compliance team | Clear list of entities and their role |
None of these questions is about price or product. All of them are about credibility, governance and compliance. That is exactly the content trading houses publish least, or bury in PDFs that AI assistants read poorly.
Why some houses are cited and others are not
AI assistants like ChatGPT have no privileged access to your reputation. They piece together fragments: your website, registries, press releases, trade press, professional profiles, NGO reports, sector directories. When those fragments contradict each other or are missing, the answer turns vague, wrong, or names a better documented competitor.
| Factor | What helps | What hurts |
|---|---|---|
| Consistency | Same legal name, address and directors everywhere | Former entity name still online, a 2019 address |
| Factual pages | One page per topic: governance, compliance, offices, careers | A four page site with one sentence per topic |
| Third party mentions | Trade press, industry associations, registries | Only critical NGO reports, with no published response |
| Structured data | Organization markup with entities, addresses and official links | No markup, just a logo on the homepage |
| Freshness | ESG report and compliance pages dated this year | A sourcing policy dated 2021 |
| Languages | Key pages in English, consistent French and German versions | Machine translations that diverge between languages |
Two sector constraints deserve attention. First, the temptation to overclaim: a page asserting full sanctions compliance or complete traceability creates risk if reality is more nuanced, and AI assistants will repeat that text verbatim. Publish only what your compliance function has actually put in place and signed off. Second, data protection: publishing staff names and photos, or collecting CVs through a form, falls under the GDPR, supervised in Luxembourg by the CNPD. This is not legal advice: your data protection officer and counsel decide what can be published.
What the work covers and what it costs on the market
In broad terms, the work records what AI assistants say about your company today, fixes public inconsistencies (website, registries, profiles, directories), makes the pages that prove your governance and compliance readable, then tracks how answers evolve. The market is young: the ranges below are 2026 ballparks drawn from typical quotes by European agencies and consultants, with wide variation.
| Service | 2026 ballpark (EUR excl. VAT) | What moves the price |
|---|---|---|
| One off diagnostic, 20 to 30 questions in 2 languages | 2,500 to 6,000 | Number of entities, languages and audiences |
| Upgrade of factual pages | 6,000 to 18,000 | Compliance and management sign off loop |
| Structured data and consistency of public listings | 1,500 to 4,000 | Number of legal entities and offices |
| Answer tracking, dated monthly record | 800 to 2,500 per month | Number of questions and frequency |
| Trade press relations, third party mentions | 3,000 to 8,000 per month | Often handled by a financial communications agency |
| Additional German or French translation | EUR 0.12 to 0.20 per word (market assumption) | Technical trading and compliance vocabulary |
The most underestimated cost is not technical: it is internal review time. A compliance page goes through the compliance officer, sometimes group legal, sometimes the CEO. Allow four to eight weeks for that loop.
How to measure without fooling yourself
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
The only honest measure is a dated record: the same list of questions, asked in the same languages, at fixed intervals, with full answers archived. Three indicators are enough: the share of answers that mention your company, the accuracy of what is said (entity, directors, business, locations) and the sources the AI appears to rely on. Answers vary between sessions, so you reason on quarterly trends, never on a single screenshot. No serious provider can promise that an AI will recommend your company.
Mini case study
Illustrative example. Ardennes Commodities SA, a 45 person non ferrous metals trading house in Kirchberg, part of a group headquartered in Geneva. Its head of communications selects 24 questions, asked in English and French, giving 48 queries recorded each month.
In the baseline record, the company appears in 5 answers out of 48, about 10.4%. Two of those five mentions use the former legal name, and one attributes a refining activity to the Luxembourg entity that it does not carry out.
First year budget: diagnostic EUR 4,000, upgrade of the governance, compliance, offices and careers pages EUR 12,000, tracking at EUR 1,200 per month over 9 months, so EUR 10,800. Total: EUR 26,800 excl. VAT. The goal is measurable and promises nothing: correct the inaccurate mentions and compare the citation share at the month 6 and month 12 records.
FAQ
Can an AI assistant like ChatGPT recommend a trading house?
It may name one when asked which houses operate in Luxembourg in a given metal or soft commodity. Nobody controls those answers: you act on public sources, then measure, for instance on 20 to 30 questions each quarter.
Do we need to publish counterparties or volumes to be cited?
No. AI assistants mostly cite stable facts: entities, directors, locations, published policies, dated reports. Commercial discretion remains compatible with an accurate presence.
How long before anything changes?
Usually 3 to 6 months between fixing public pages and a visible change in answers, because third party sources update slowly. The internal review loop often adds 4 to 8 weeks.
Which languages should we work in from Luxembourg?
English first, since most banks and counterparties query AI assistants in English. French and German matter for local press, recruits and some authorities, with translation at roughly EUR 0.12 to 0.20 per word.
Find out how AI assistants describe your firm. A dated record of what AI assistants say about your trading house, in English and French, will tell you whether the topic deserves a budget. Request a diagnostic quote: reply within 48 hours.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
