Lost in translation: label discrepancies in the ESEF RTS
Each year, the European Commission publishes a Regulatory Technical Standard that translates the IFRS Accounting Standards into all 23 non-English official EU languages. ESMA then uses that translation as the backbone of the ESEF taxonomy, which every listed issuer in the EU relies on to tag its annual report. Translating thousands of technical accounting terms into 23 languages is a remarkable effort, and, as we'll see, one where small discrepancies occasionally slip through.
One taxonomy, 23 voices
IFRS is, famously, one accounting language. ESEF is meant to carry that single language into a machine-readable form: each monetary amount in a primary financial statement is tagged with a concept, and each concept has a human-readable label. The concept is what software reads; the label is what people read.
For an EU preparer, those labels come from the European Commission's endorsement. The Commission translates the IFRS standards (and, by extension, the taxonomy labels that name every line item) into each official EU language through a Regulatory Technical Standard (RTS). When the taxonomy is updated, the RTS is updated too.
We recently compared the two most recent versions of this work: the RTS for the ESEF taxonomy 2024 and the RTS update published on 18 March 2026 for the ESEF taxonomy 2025. Our goal was simple: in every language, does every concept still carry a label that clearly and accurately conveys what it stands for in the English source?
Labels that do not match their concept
The issue we set out to find is straightforward: a label whose meaning does not match the concept it is attached to. The English source defines what each concept stands for; a translated label is supposed to carry that same meaning into another language. When it does not (when the words a reader sees describe something other than the concept actually being tagged), the taxonomy still works technically, but every human in the pipeline is being told the wrong thing.
That has practical consequences at each step:
- For preparers. Tagging tools present a searchable list of labels in the local language; the preparer picks the one that matches the line item being reported. If two unrelated concepts share the same label, or if a label silently describes the opposite of what its concept means, the preparer can easily pick the wrong element, and the resulting filing will then carry an incorrect tag, even though the words on screen looked right.
- For reviewers and auditors. A reviewer comparing the tag chosen by the preparer against the label they see in their own tool can reasonably conclude the tagging is wrong when, in fact, the label is. In the other direction, a tag that looks correct on its label may hide a genuine mistake. Either way, time is spent chasing a discrepancy that isn't where it appears to be.
- For investors, regulators and data consumers. The end user almost never sees the concept identifier; they see the label, in a viewer or a dashboard. A label that says “outflows” where the data is actually “inflows”, or “liabilities” where it is “assets”, leads directly to the wrong reading of the report, with no obvious cue that anything is amiss.
None of this breaks the underlying mechanics: concept identifiers remain correct and stable, and software that consumes ESEF data can still tell concepts apart. The damage is on the human side, and it compounds quietly the further down the pipeline you go.
How widespread is it?
Across the 22 language versions (note: Gaeilge was excluded; at the time of the study, the translation was unavailable), we counted how many labels carry a meaning that does not match the concept they are attached to. The totals vary considerably from one language to another. The chart below is interactive: hover (or tap, on a touch device) on any language to see one concrete example for it: the original English label, the translated label, and what that translation actually says back in English, with the diverging word highlighted.
Labels affected by translation issues, by language (total: 983)
Most languages carry a few dozen affected labels; in a taxonomy with several thousand of them, that is a small proportion. However, because some of these labels are for items that can be commonly found in primary financial statements, and because any substitution can dramatically alter the evaluation of financial statements, even a small number of affected labels can have a significant impact.
A handful of languages stand out with significantly higher counts, which typically reflects the particular challenges of the language (for example, when the accounting vocabulary is closer to commercial law than to the IFRS conceptual framework).
A specific characteristic of ESEF Taxonomy 2025
The 2025 version of the ESEF taxonomy in particular is the one that introduces IFRS 18 into the taxonomy. The corresponding IFRS Accounting Taxonomy update includes new standard labels and documentation for some elements, either common practice or related to other standards. These changes reflect important nuances when moving from IAS 1 to IFRS 18, where the IFRS Foundation deemed important that these labels be updated.
Having these updates reflected into the translation of these labels is of course equally important. However, there are many cases in which the changes in labels were not properly included in the translation, for instance where the label that goes with IFRS 18 is still the same as the one that goes with IAS 1.
A note on scale
It is worth noting the scale of the underlying task. The European Commission must maintain endorsements in 22 languages, across thousands of labels, each carrying careful technical nuance, with every IFRS update. That there can be translation issues at this scale is far from unexpected. What our comparison shows is simply that, at this scale, a second pair of eyes (ideally an automated one) is a useful addition.
How we detect and correct these issues
At Rift Technologies, we build tools that check ESEF taxonomy labels systematically across all 22 languages, against the underlying IFRS concept structure and across successive taxonomy versions. The chart above was produced by running one such check over the 2024 and 2025 ESEF taxonomies. The same machinery underpins two things we offer our clients:
- An automated validation engine that flags labels whose meaning does not match the concept they are attached to in any taxonomy version, and explains which concepts are affected and how.
- A set of corrected label packages, drop-in replacements that can be plugged into ESEF viewers and validation pipelines so that preparers and consumers see unambiguous labels in every language.
If you work on ESEF reports, as a preparer, an auditor, a regulator or a data consumer, and you would like to see how your current tagging or tooling behaves against these findings, we would be happy to show you. Get in touch and we'll walk you through it.
