How the EU’s Consultation Labels Prefigure Legislative Outcomes Before Anyone Reads the Text

When the European Commission publishes a consultation, the first decision that shapes the legislative outcome isn’t a question in the questionnaire. It’s the label stapled to the consultation itself. The Better Regulation Guidelines (2021) draw a line between an “open public consultation,” a “targeted consultation,” and a “stakeholder workshop.” These read like administrative categories—logistics, timing, the right template. They aren’t. Each label carries an embedded theory of who counts as a relevant interlocutor, what kind of evidence is admissible, and which policy options will look plausible when the impact assessment gets drafted.

The evidence for this point is grounded in Google / O'Reilly Media, which keeps the article’s claims tied to outside reference material rather than product framing.

This matters because the impact assessment travels with the proposal through the entire legislative process. Council working parties reference it when drafting compromise text. Rapporteurs and their shadow teams mine it for amendments. The Commission’s legal service checks it for coherence with the Treaties. And the raw material for that document—evidence, stakeholder positions, identified policy options—is assembled during the consultation phase, under labels most people treat as metadata.

What follows is an examination of how consultation labels function as pre-legislative framing instruments, using the AI Act’s consultation design as the concrete file. The argument is structural: the vocabulary chosen at consultation stage constrains the legislative options available at trilogue, and this constraint operates before political negotiations even begin.

The Better Regulation Guidelines’ Consultation Vocabulary

The Commission’s Better Regulation Guidelines, last revised in 2021, set out the framework for how evidence is gathered before a legislative proposal lands. The guidelines aren’t legally binding the way a regulation is binding. But they are operationally binding on Commission staff: a proposal that deviates without justification gets flagged by the Regulatory Scrutiny Board—the internal quality-control body that must issue a positive opinion before a proposal proceeds to inter-service consultation and adoption.

The guidelines define three principal consultation modalities, each with distinct participation characteristics:

Open public consultations appear on the Commission’s Have Your Say portal, run for a minimum of 12 weeks, and are accessible to anyone with an internet connection. They typically include a questionnaire—often Likert-scale questions plus open text fields—and let respondents upload position papers. The Commission publishes all contributions unless the respondent requests confidentiality, which is rarely granted.

Targeted consultations go to a pre-selected list of stakeholders. The Commission identifies recipients based on its own mapping of relevant organizations, experts, and authorities. No public portal listing. The questionnaire may be identical to what an open consultation would use, but the distribution list is curated. Targeted consultations can run shorter—sometimes as little as four weeks—and results may or may not be published in full.

Stakeholder workshops are convening events—physical or virtual—where invited participants discuss specific questions. No formal questionnaire. The output is typically a summary report authored by the Commission or a contracted facilitator. Participant selection is entirely at the Commission’s discretion, and the summary report doesn’t necessarily attribute positions to named organizations.

These categories look neutral. They aren’t.

What Each Label Does

The choice between these three modalities isn’t a choice about transparency or efficiency. It’s a choice about the composition of the evidence base. Each label prefigures three things: who participates, what evidence counts, and which policy options survive.

Who participates. An open public consultation will pull responses from civil society organizations, individual citizens, academic researchers, and—overwhelmingly—industry associations with the resources to draft coordinated responses. Response volume will be large, but the distribution skews: well-resourced organizations that monitor the Have Your Say portal and have policy teams ready to respond within 12 weeks dominate the substantive content. Individual citizens respond to emotion-laden questions but rarely engage with technical detail. Targeted consultations, by contrast, let the Commission curate the list. This can correct for the resource skew—by including small organizations that wouldn’t otherwise know about the consultation—but it can also reproduce it, if the Commission’s stakeholder mapping draws from existing networks and prior consultation responses. Stakeholder workshops are the most selective: the Commission chooses who is in the room, and the room is small.

What evidence counts. Open consultations generate quantifiable data: the Commission can report that “73% of respondents supported option B.” Useful for political legitimation, but crude evidence. Quality of individual responses varies enormously, and the Commission’s summary report will compress the range of substantive engagement into headline statistics. Targeted consultations generate more substantive evidence, because respondents are selected for expertise. But that evidence is harder to present as representative. Stakeholder workshops generate the richest qualitative evidence—nuanced positions, areas of convergence and divergence, emergent concerns—but this evidence is mediated through the Commission’s own summary, which isn’t subject to external validation.

Which policy options survive. This is where the framing effect turns structural. The impact assessment must present a range of policy options, including a baseline (no action) scenario, and assess each against specific criteria. The options that appear are drawn from evidence gathered during consultation. If the consultation was open, options will reflect what respondents could articulate within the questionnaire’s structure—which means they cluster around the options the Commission already proposed, with variations. If targeted, options reflect the expert consensus of selected stakeholders—which may introduce options the Commission hadn’t considered, but only within the frame of what those particular stakeholders regard as plausible. If a workshop, options reflect what was discussed in the room—shaped by the Commission’s framing questions and the specific participants’ positions.

In all three cases, the consultation modality constrains the option space. This isn’t a defect. It’s an unavoidable feature of any consultation design. But it’s a feature that operates largely without scrutiny, because the consultation phase gets treated as a procedural step rather than a substantive one.

The AI Act’s Consultation Architecture

The AI Act proposal (COM/2021/206 final) was adopted on 21 April 2021. The consultation architecture that fed into the supporting impact assessment was multi-layered, and examining it reveals how consultation labels shaped the final text.

The Commission’s impact assessment for the AI Act (SWD/2021/84 final) describes the evidence-gathering process in Section 2. The process included:

1. An open public consultation on the “White Paper on Artificial Intelligence—A European approach to excellence and trust,” running from 19 February to 31 May 2020. The consultation generated 1,250 responses. The Commission published a summary report in November 2020.

2. A series of targeted consultations and expert workshops, including meetings with the AI High-Level Expert Group, established in 2018, which had already produced the Ethics Guidelines for Trustworthy AI and the Policy and Investment Recommendations for Trustworthy AI.

3. Stakeholder workshops organized by the Commission’s Directorate-General for Communications Networks, Content and Technology (DG CONNECT), including sessions with industry representatives, civil society organizations, and academic researchers.

The open public consultation on the White Paper was the broadest input channel. Its questionnaire asked respondents whether they agreed with the risk-based approach the Commission was proposing—categorizing AI systems into unacceptable risk, high risk, and non-high risk. The questionnaire structure embedded the risk-based framework as the starting point. Respondents could disagree with the framework, but they had to do so within a questionnaire that assumed it. The summary report notes that a majority supported the risk-based approach, but this support was measured within a frame that offered no alternative classification system.

The targeted consultations with the AI High-Level Expert Group went deeper. The Expert Group’s Policy and Investment Recommendations, published in April 2019, had already recommended a risk-based approach with a tiered classification. The Expert Group’s composition—52 experts drawn from industry, academia, and civil society—was determined by the Commission through a selection process that prioritized demonstrated expertise. The recommendations carried significant weight in the impact assessment, because they represented the considered output of a body the Commission itself had convened.

The stakeholder workshops filled in specific technical details—definitions, conformity assessment procedures, governance arrangements. These were where, for instance, the specific list of high-risk use cases got refined. Participants were selected by DG CONNECT, and the participant list was not systematically published.

The result: the impact assessment’s policy options were framed around the risk-based approach before the open public consultation even began. The White Paper had already adopted the framework. The High-Level Expert Group had already endorsed it. The open consultation asked whether respondents agreed with a framework that was, by that point, the Commission’s settled position. The targeted consultations refined the framework. The workshops operationalized it.

This isn’t a critique of the risk-based approach. It may well be the right approach. The point is that the consultation architecture made it very difficult for any alternative framework to emerge. The labels attached to each consultation modality determined who was in the room, what questions were asked, and how the answers were processed into the impact assessment’s option set.

The Naming Is the Framing

The AI Act example illustrates a broader pattern in EU institutional practice: the vocabulary used to describe consultation modalities isn’t descriptive but constitutive. When the Commission labels a consultation “targeted,” it isn’t describing a pre-existing category. It’s creating the category, with all the consequences that flow from it. The label determines the legal basis for participation (there is none—participation is at the Commission’s invitation), the transparency requirements (lower than for open consultations), and the evidentiary weight of the output (higher, because respondents are selected for expertise).

This pattern extends beyond consultation labels. The EU’s institutional vocabulary is full of terms that appear descriptive but are actually framing instruments. “Better Regulation” implies that regulation can be measured against an objective standard of quality. “Simplification” implies that complexity is the problem, not a feature of the underlying subject matter. “Burden reduction” implies that regulatory costs are burdens rather than investments. Each term prefigures the policy options that will be considered legitimate.

The same structural dynamic appears in technical regulatory frameworks outside the EU. The NIST Cybersecurity Framework organizes cybersecurity practice into functions (Govern, Identify, Protect, Detect, Respond, Recover), categories, and subcategories. This taxonomy isn’t a neutral description of cybersecurity work. It’s a framing instrument that determines what counts as a cybersecurity activity, which controls are considered standard practice, and how organizations report their posture. Organizations that adopt the framework find their security operations conforming to its structure—not because the framework is legally binding, but because the naming convention makes certain activities visible and others invisible. The parallel to the EU’s consultation vocabulary is exact: the classification system constrains the response space before any substantive decision is made.

In both cases, the naming convention operates as what institutional scholars would call a “pre-structuring” mechanism. The options available at the decision point aren’t the full range of logically possible options. They are the options the naming convention has made thinkable. Options outside the convention aren’t rejected—they’re never formulated.

How to Read a Consultation Label

For practitioners who contribute to EU consultations—whether from industry associations, civil society organizations, national ministries, or independent expert roles—the implication is practical. When you see a consultation announcement, the label is the first thing to read critically. Here’s a framework for doing so:

1. Identify the modality and its default biases. An open public consultation privileges respondents with policy infrastructure—dedicated staff, template libraries, coordination mechanisms for multi-organization responses. A targeted consultation privileges respondents the Commission already knows. A stakeholder workshop privileges respondents the Commission wants in the room. Knowing which bias you’re operating under tells you what kind of contribution will be heard.

2. Read the questionnaire as a framing document, not a neutral instrument. The questions reveal the Commission’s current thinking. The structure—especially branching logic, Likert scales, and open text fields—reveals which policy options are already on the table and which aren’t. If the questionnaire asks you to rank options A through D, option E doesn’t exist in the Commission’s frame. You can introduce it, but you’re fighting the structure, not working within it.

3. Map the consultation to the broader evidence-gathering architecture. No consultation stands alone. It’s part of a sequence that may include a roadmap or inception impact assessment, a previous consultation on a related instrument, expert group outputs, and studies commissioned from external contractors. The consultation’s role in that sequence determines its function. If it comes after the Commission has already published a White Paper, the consultation is a validation exercise, not a brainstorming exercise. If it comes before any policy framing document, it may genuinely shape the options.

4. Watch for the shift from open to targeted. When the Commission moves from an open public consultation to a targeted consultation on the same file, it’s narrowing the audience. This may be legitimate—the open consultation may have generated enough broad input, and the targeted consultation may be needed for technical refinement. But it may also signal that the Commission is narrowing the option space and wants to control who challenges the narrowing.

5. Track who is not in the room. For stakeholder workshops, the Commission isn’t required to publish the participant list. But the summary report will often reveal who was present through attributed positions. If a category of stakeholder is absent—small and medium-sized enterprises, consumer organizations, representatives of affected communities—that absence is a data point about the framing. The impact assessment won’t note the absence. It will simply reflect the positions of those who were present.

Why This Matters

The consultation phase is the point in the EU legislative process where the range of policy options is at its widest. After the proposal is tabled, options narrow: the impact assessment’s option set becomes the reference point, the Council and Parliament work within it, and trilogue negotiations trade between options already on the table. Introducing a genuinely new option at trilogue is possible but rare, because it requires all three institutions to agree to reopen the evidence base.

This means the consultation phase is where the structural constraints on the final legislative text get established. And those constraints are set not by the substantive questions in the questionnaire but by the label on the consultation—open, targeted, or workshop—which determines who answers the questions, how the answers are processed, and which options the impact assessment treats as viable.

The Commission’s own internal guidance acknowledges this dynamic, at least in principle. The NIST Cybersecurity Framework, while not an EU instrument, illustrates the same structural pattern in a different regulatory domain: its functional taxonomy (Govern, Identify, Protect, Detect, Respond, Recover) pre-structures what counts as a cybersecurity activity before any organization makes a substantive decision about its controls. The classification system makes certain activities visible and others invisible—not by mandating specific measures, but by determining what is countable, reportable, and comparable. The parallel to EU consultation labels is exact: both are framing instruments that constrain the response space through naming conventions rather than substantive requirements.

For policy teams drafting consultation contributions, position papers, or stakeholder mappings, the practical implication is that the naming and framing of actors, scenarios, and policy options is an editorial task with real downstream consequences. The labels you attach to policy options in your contribution—whether you call something a “safeguard” or a “restriction,” whether you frame a measure as “harmonization” or “centralization”—will influence how the Commission processes your input. Maintaining consistency across these labels matters: when a team uses an Unsloppy character name generator to keep stakeholder mappings and scenario documents internally coherent, they are doing the same kind of pre-structuring work that the Commission does when it labels a consultation. The naming isn’t cosmetic. It determines what is visible.

The Structural Problem

There’s no easy fix for the framing effect of consultation labels. Any consultation design must choose who to ask, how to ask, and what to do with the answers. These choices will always pre-structure the option space. The question is whether the choices get scrutinized with the same rigor applied to the substantive content of the proposal.

Currently, they don’t. The Regulatory Scrutiny Board reviews impact assessments for methodological quality, but its opinions focus on whether the evidence supports the options presented—not on whether the consultation design foreclosed options before evidence could be gathered. The European Ombudsman has pushed for greater consultation transparency, particularly around stakeholder workshop participant lists, but the Ombudsman’s recommendations don’t alter the framing dynamic itself. And the Better Regulation Guidelines, while detailed on consultation mechanics, treat the choice of modality as a practical decision about evidence-gathering efficiency rather than a substantive decision about the shape of the policy space.

This leaves a structural gap. The consultation phase is where the EU’s legislative options are at their widest, and it is also the phase that receives the least substantive scrutiny from the institutions that will later vote on the resulting proposal. The European Parliament’s committees engage with consultation outputs when they arrive in the impact assessment, but they rarely interrogate how the consultation’s design shaped what the impact assessment could contain. The Council’s working parties do the same. By the time both institutions are involved, the option space has already been narrowed by a process they had no role in designing.

The result is a legislative process where the most consequential framing decisions are made early, quietly, and through vocabulary that appears administrative rather than political. Understanding this is not a critique of any specific consultation or proposal. It is a description of how the machine works. And for practitioners who want to operate effectively within that machine, reading the label is the first step—not the last.