ESAP Goes Live 10 July 2027: Making Your Filings Machine-Readable and Discoverable

On 10 July 2027 the European Single Access Point opens, and for the first time a supervisor, an analyst or a machine can pull your disclosures from a single portal. The question is not whether you will file. It is whether what you file can be found and read once it lands there.

The European Single Access Point (ESAP), established by Regulation (EU) 2023/2859 and the omnibus Directive (EU) 2023/2864, is the EU’s answer to a genuine problem: financial and sustainability information about European entities is scattered across national registers, officially appointed mechanisms and regulator websites, in formats that range from tagged XBRL to scanned PDFs. ESMA is building a single portal, managed centrally, that collects all of it and exposes it through a search interface and a public API. Most of the commentary treats this as a disclosure-transparency story. For the technology function it is something narrower and more concrete: a change in what a valid filing has to carry.

What actually changes on 10 July 2027

Collection of information begins on 10 July 2026, and the portal opens to users on 10 July 2027. The rollout is phased by instrument. The first phase covers disclosures under the Transparency Directive, the Prospectus Regulation and the Short Selling Regulation. A second phase, from 10 January 2028, brings in accounting, market abuse and sustainability disclosures among others; a third, from 10 January 2030, adds the remaining bodies of law across banking, insurance, securities financing and alternative investments. Voluntary submission of additional information also opens from 2030.

You do not file directly into ESAP. You file, as you do today, with a collection body — an officially appointed mechanism, a national competent authority or an EU agency — and that body forwards the disclosure to ESAP with the required technical envelope. That indirection is precisely where firms lull themselves into thinking there is nothing to do. The collection body relays what you gave it. If what you gave it is a scanned PDF with no identifiers, that is what ESAP receives, and that is what a data user cannot use.

Submitted is not the same as discoverable

The regulation draws a distinction that matters technically. Information must be supplied in a data-extractable format — one that allows the content to be copied or exported without retyping, which rules out scanned images — and, for certain disclosures, in a machine-readable format, meaning a structured format software can process without conversion. Annual financial reports already fall under the ESEF regime, so they arrive as Inline XBRL and satisfy the higher bar. A half-yearly report supplied as a text-based PDF meets the extractable bar but not the structured one. The gap between those two states is the gap between a filing that appears in a search and one that appears in an analyst’s dataset.

This is the same discipline that turns a build artefact into a genuinely usable deliverable rather than a file that merely exists. We have written before about producing a machine-readable software bill of materials for the Cyber Resilience Act, and the underlying lesson transfers directly: a document a regulator can technically open is not the same as a document a downstream system can parse. ESAP makes that distinction supervisory.

The metadata your filing has to carry

Discoverability on ESAP is driven by metadata, not by the document body. The implementing technical standards, adopted as Commission Implementing Regulation (EU) 2025/1339 of 10 July 2025, specify the descriptors the collection body attaches. In practice your filing needs, at minimum:

• A valid legal entity identifier for the issuer, consistent with the GLEIF register. A lapsed or mistyped LEI is the single most common way a filing becomes an orphan record.
• The type of information — annual report, half-yearly report, inside information, major-holding notification and so on — drawn from a controlled classification, not free text.
• A size classification for the entity, aligned to the Accounting Directive, and an industry-sector code from the prescribed taxonomy.
• The period or reference date the disclosure covers, the language, and an indicator of whether the submission contains personal data.

None of this is exotic. The trouble is that most firms hold these attributes in different systems — the LEI in treasury or company secretarial, the sector code nowhere in particular, the format flag implicit in whatever the reporting tool happened to export. The work is assembling them, correctly, at the moment of filing rather than reconstructing them afterwards. That case for building the descriptors into the pipeline rather than bolting them on is one we have made in more detail in a companion piece on the metadata and format requirements behind the portal.

Building the metadata at the point of filing

The right place to attach and validate this is the step that hands the disclosure to the collection body — not a spreadsheet a person fills in later. A small validation gate catches the two failures that matter most: an invalid identifier, and a format that does not meet the bar the information type requires.

"""Attach and validate ESAP metadata before a filing reaches the
collection body. Run at the point of submission, not afterwards."""

import re
from dataclasses import dataclass, asdict

LEI_RE = re.compile(r"^[A-Z0-9]{18}[0-9]{2}$")  # ISO 17442 structure

# Formats ESAP treats as machine-readable; anything else is accepted
# only as a data-extractable document, never a scanned image.
MACHINE_READABLE = {"xhtml-ixbrl", "xbrl", "xml"}


@dataclass
class EsapMetadata:
    lei: str                    # issuer LEI, must match the GLEIF register
    info_type: str              # e.g. "annual-financial-report"
    size_class: str             # per the Accounting Directive
    sector: str                 # prescribed sector taxonomy
    period_end: str             # ISO 8601 date the disclosure covers
    language: str               # ISO 639-1
    file_format: str            # e.g. "xhtml-ixbrl", "pdf"
    contains_personal_data: bool

    def format_flag(self) -> str:
        return ("machine-readable" if self.file_format in MACHINE_READABLE
                else "data-extractable")

    def validate(self) -> list:
        errors = []
        if not LEI_RE.match(self.lei):
            errors.append("LEI is not a valid ISO 17442 identifier")
        if self.info_type == "annual-financial-report" 
                and self.file_format not in MACHINE_READABLE:
            errors.append("annual reports must be iXBRL, not a flat PDF")
        return errors


record = EsapMetadata(
    lei="529900T8BM49AURSDO55",
    info_type="annual-financial-report",
    size_class="large",
    sector="K64.19",
    period_end="2026-12-31",
    language="en",
    file_format="xhtml-ixbrl",
    contains_personal_data=False,
)

problems = record.validate()
if problems:
    raise ValueError("; ".join(problems))
print(record.format_flag(), asdict(record))

The point is not the exact fields, which the implementing standards fix; it is that the check runs before the disclosure leaves your control, and that a failure blocks the submission rather than surfacing months later as a record nobody can find.

Sequencing it against the date

The date that binds you is not 10 July 2027 but 10 July 2026, when collection begins for the first phase. If your instruments sit in phase one — transparency, prospectus, short selling — your filings are being drawn into the pipeline a year before the portal is public, and the metadata discipline needs to be live by then. Treat 2026 as the readiness deadline and 2027 as the day the world sees the result. That framing is the same one worth applying to the other 2027 go-lives clustering on the calendar, including the EU market-risk regime under FRTB: the reporting change lands well before the headline date.

Practically, the sequence is short. Confirm which of your disclosures fall in which phase. Reconcile your LEI against GLEIF and fix any lapses now, because renewal is not instant. Establish which information types require machine-readable output and which of those you already produce. Then move the metadata assembly and validation upstream, into the filing step, so it is generated by the system and not typed by a person under deadline. None of this is large. All of it is the difference between disclosures that are present and disclosures that are usable.

A filing that clears the collection body but carries a dead LEI and a scanned page is, on ESAP, indistinguishable from a filing you never made.