# Will OpenTelemetry mark its semantic conventions for generative AI as stable before 31 December 2027?

> As of 17 November 2026, Prolepsia puts the chance of yes at 46%.

- Forecaster: Prolepsia (https://prolepsia.com)
- Forecast made: 2026-11-17
- Question type: Yes / no
- Resolves: 2027-12-31
- Status: Open
- Forecasts: 1
- Page: https://prolepsia.com/forecasts/momentum-pendulum-spectrum/will-opentelemetry-mark-its-semantic-conventions-for-generative-ai-as-stable-bef
- Open in the app: https://app.prolepsia.com/library/0cba4e18-cae2-433e-b7e1-f40bd4c6ad16

## Forecast

- Yes: 46%
- No: 54%

## How Prolepsia reasons

## TL;DR
The forecast is **46.4% Yes and 53.6% No** for OpenTelemetry officially designating its general GenAI semantic conventions Stable before December 31, 2027. Treat No as the slight favorite, not as a strong conclusion. The decisive distinction is between stabilizing individual components and adopting an authoritative Stable designation for the general foundation.

## Context
OpenTelemetry’s GenAI semantic conventions define how observability systems describe generative AI operations. As of the November 17, 2026 forecast date, the [overarching documentation still says Development](https://github.com/open-telemetry/semantic-conventions-genai/blob/main/docs/gen-ai/README.md). The dedicated repository supports a focused stabilization effort, but its existence does not itself establish specification maturity.

The qualifying event is an adopted official declaration covering the general or foundational GenAI conventions. Stable SDKs, production use, release candidates, and isolated Stable attributes do not qualify. The deadline is strictly before December 31, 2027, at 00:00 UTC. Experimental extensions can remain after a qualifying foundational declaration; the question does not require every GenAI feature to graduate together.

## Evidence
The historical record gives two lessons. OpenTelemetry can finish a bounded stabilization effort within the available window. It can also spend much longer resolving scope and compatibility, even after announcing a push toward stability. The relevant comparisons are small and selected, so they guide timing rather than supply a reliable statistical success rate.

For HTTP, the final stabilization push began January 30, 2023. A feature-freeze release followed July 13, and the core Stable release arrived November 3, followed by the November 6 declaration. That was 277 days from the push to release and 113 days from freeze to release. The announced sequence totaled 84 days, so actual release came 193 days later than that simple schedule implied ([stabilization plan](https://opentelemetry.io/blog/2023/http-conventions-stability/); [Stable release](https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.23.0); [declaration](https://opentelemetry.io/blog/2023/http-conventions-declared-stable/index.md)). This is evidence that a focused push can succeed, but also that short public schedules deserve caution.

Database conventions show why the first candidate matters more than the last. A preparatory transition-plan notice appeared February 12, 2024. The first released candidate arrived October 8, 2024, the second March 31, 2025, and core stability May 2, 2025. The first-candidate-to-Stable interval was 206 days; the second-candidate interval was only 32 days. The earlier preparatory notice preceded stability by 445 days ([transition-plan notice](https://github.com/open-telemetry/semantic-conventions/commit/183647b9e55ecccb320b76bab325416507e151b9); [first candidate](https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.28.0); [second candidate](https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.32.0); [Stable release](https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.33.0)). Reading only the last interval would hide most of the stabilization work.

RPC followed a similarly substantial path. Its public kickoff was June 2, 2025, followed by a formal-project milestone August 29, a first candidate February 19, 2026, the Stable merge October 28, and the Stable release November 2. The public-kickoff-to-release interval was 518 days, while candidate-to-release took 256 days ([kickoff](https://opentelemetry.io/blog/2025/stabilizing-rpc-conventions/); [project record](https://github.com/open-telemetry/community/commit/459a9dcb45d5efc8f5410a023cd7665575e7c5be); [candidate release](https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.40.0); [Stable release](https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.45.0)).

Unfinished efforts balance these successful examples. Messaging and system conventions remained Development in the November 2, 2026 release snapshot, despite earlier stabilization activity. Process conventions also illustrate that component-level candidate progress need not establish general stability ([messaging status](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/messaging/README.md); [system status](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/system/README.md); [process metrics](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/system/process-metrics.md)). These are unfinished cases, not proof of permanent failure. Excluding them would make stabilization look easier than it is.

Against that backdrop, GenAI has stronger positive evidence than its Development label alone suggests. The stabilization tracker opened September 11, 2026, and its November 6 checklist records merged work on agent/workflow boundaries, operation-specific metrics, token instruments, structured content, and exception events. This is a concrete effort to settle a foundation, not merely a broad aspiration ([stabilization tracker](https://github.com/open-telemetry/semantic-conventions-genai/issues/511)).

The same tracker is the clearest evidence against assuming quick completion. Naming, nested instrumentation, usage attribution, schema compatibility, tool-result representation, capture configuration, reference validation, and migration work remain unresolved. It explicitly keeps the core in Development ([remaining work](https://github.com/open-telemetry/semantic-conventions-genai/issues/511)). These are not all equally difficult, and a checklist count would not measure the time left. Some decisions define the contract that implementations and telemetry consumers must share.

Schema work makes that point concrete. The input/output schema discussion separates stability blockers from improvements that can wait until afterward. That is good prioritization, but it also confirms unfinished foundational design ([schema stability discussion](https://github.com/open-telemetry/semantic-conventions-genai/issues/527)). A stable contract needs more than working instrumentation: producers and consumers must agree on what the emitted data means and how changes preserve compatibility.

Scope control is the strongest route around delay. The proposal to exclude retrieval from initial stability shows maintainers considering a narrower first foundation ([retrieval proposal](https://github.com/open-telemetry/semantic-conventions-genai/issues/637)). My reading is that the project is trying to distinguish essential commitments from later expansion. This supports timely completion without assuming every AI operation must first stop changing.

Practical validation is also advancing. The repository’s reference scenarios exercise inference, agents, tool execution, and metrics, and validate emitted telemetry against the conventions. That creates a way to expose implementation disagreements before promotion, though it does not establish a final stable contract ([reference scenarios](https://raw.githubusercontent.com/open-telemetry/semantic-conventions-genai/main/reference/README.md)).

Release evidence remains weaker. The dedicated repository has no published releases, and its changelog contains only an Unreleased section ([release history](https://github.com/open-telemetry/semantic-conventions-genai/releases); [changelog](https://raw.githubusercontent.com/open-telemetry/semantic-conventions-genai/main/CHANGELOG.md)). This supports caution about maturity. It is not a formal barrier: an authoritative documentation change or adopted announcement could qualify without a GitHub release.

## What's non-obvious
**Scope reduction can help the technical schedule while still failing the question.** Excluding retrieval or other unsettled areas can make a stable foundation easier to finish. But separate Stable metrics, spans, or inference conventions do not automatically amount to a Stable general specification. The breadth of the adopted declaration matters alongside the engineering work.

The reverse mistake is to require universal stability. RPC’s overarching documentation says Stable unless otherwise specified, providing a precedent for general stability with exceptions. HTTP and database parent documents remain Mixed despite core-stability milestones, showing why the authoritative announcement matters as well as the parent label ([RPC status](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/rpc/README.md); [HTTP status](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/http/README.md); [database status](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/db/README.md)). The right test is neither “everything is stable” nor “something is stable.” It is whether OpenTelemetry explicitly establishes a stable general foundation.

## Uncertainties
The largest gap is timing. The evidence does not establish a committed current completion date or a comprehensive assessment of remaining blockers. That does not mean no internal or working-group timetable exists. An adopted foundational freeze, settled compatibility rules, and completed reference validation would make the remaining path much clearer.

The next gap is the intended designation. A bounded package can satisfy the question if the project explicitly calls the general foundation Stable. The same technical work can fall short if maintainers promote only separate components while retaining a Development umbrella and making no qualifying general declaration. Clear adopted language about the first Stable scope would close this gap.

Historical comparisons also have limits. Kickoffs, preparatory notices, freezes, candidates, merges, and releases start different clocks. Successful cases share project practices and sometimes maintainers; they are not independent trials. There is no established universal minimum candidate waiting period to add mechanically, and instrumentation migration support is not a pre-Stable countdown ([convention requirements](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/how-to-write-conventions/README.md)). The forecast therefore remains a close judgment: meaningful progress supports Yes, while unfinished foundational agreement and the strict general-designation requirement leave No slightly ahead.

## Forecast history

| Date | Forecast |
|---|---|
| 2026-11-17 | 46% chance of yes |

## Sources Prolepsia read

- [semantic-conventions-genai/docs/gen-ai/README.md at main · open-telemetry/semantic-conventions-genai · GitHub](https://github.com/open-telemetry/semantic-conventions-genai/blob/main/docs/gen-ai/README.md)
- [Final push to HTTP semantic convention stability \| OpenTelemetry](https://opentelemetry.io/blog/2023/http-conventions-stability)
- [messaging status](https://raw.githubusercontent.com/open-telemetry/semantic-conventions/v1.45.0/docs/messaging/README.md)

## The question

As of November 17, 2026, OpenTelemetry's semantic conventions for generative artificial intelligence (GenAI) remain officially classified as 'Development.' These conventions standardize how observability systems represent telemetry from generative AI applications, including large language model inference, embeddings, AI agent invocations, tool execution, retrieval operations, and related metrics, spans, events, and attributes. In May 2026, OpenTelemetry moved its GenAI semantic conventions from the main semantic-conventions repository into a dedicated repository, open-telemetry/semantic-conventions-genai, to support independent versioning and faster iteration. The dedicated repository's GenAI documentation continues to identify the overall specification as 'Status: Development.' This question asks whether the OpenTelemetry project will officially promote its generative AI semantic conventions from Development to Stable before December 31, 2027. The qualifying period begins November 17, 2026, and ends December 30, 2027, inclusive. The forecast concerns the stability designation of the GenAI semantic conventions themselves, rather than the maturity of OpenTelemetry SDKs, instrumentation libraries, or the broader OpenTelemetry specification. The relevant official sources are the [OpenTelemetry GenAI semantic conventions documentation](https://github.com/open-telemetry/semantic-conventions-genai/blob/main/docs/gen-ai/README.md) and the [dedicated GenAI semantic conventions repository](https://github.com/open-telemetry/semantic-conventions-genai).

### How it resolves

The question resolves YES if, on or after November 17, 2026, and before December 31, 2027, the OpenTelemetry project officially designates the overarching generative AI semantic conventions specification as Stable. Qualifying evidence includes an official change to the GenAI specification's top-level status from Development to Stable in the OpenTelemetry documentation, an official release or changelog explicitly announcing that the GenAI semantic conventions as a whole have reached Stable status, or an equivalent authoritative declaration from OpenTelemetry's specification maintainers or governing bodies. The designation must apply to the general GenAI semantic conventions, not merely an individual attribute, metric, span, event, provider-specific convention, or instrumentation implementation. The question resolves NO if no qualifying official stability designation has occurred by the end of December 30, 2027, measured in UTC. The primary resolution source will be the official open-telemetry/semantic-conventions-genai GitHub repository, including its documentation, release history, changelog, and merged specification changes. The official OpenTelemetry website and the main open-telemetry/semantic-conventions repository will serve as supplementary authoritative sources. If these sources disagree, an explicit official announcement that the overall GenAI semantic conventions have reached Stable status takes precedence over an outdated documentation status, provided the announcement represents an adopted project decision rather than a proposal or future intention. Historical commits and release timestamps will establish whether the qualifying designation occurred before the deadline. If the specification is renamed, reorganized, or transferred to another official OpenTelemetry repository, the successor specification will be evaluated using the same criteria. If the project is discontinued or never reaches Stable status before the deadline, the question resolves NO. If the available historical record is insufficient to determine whether an official designation occurred before the deadline, the question will be annulled rather than resolved based on speculation.

### Fine print

For this question, 'before 31 December 2027' means strictly earlier than December 31, 2027, at 00:00 UTC; a stability designation first made on December 31 does not qualify. A Release Candidate designation, an announcement of plans to achieve stability, or a statement that the conventions are production-ready without an official Stable designation does not qualify. Stability of only selected GenAI components or signals does not qualify while the overarching GenAI specification remains in Development. Conversely, the introduction of new experimental GenAI features after the core specification is officially designated Stable does not reverse a qualifying YES outcome. If OpenTelemetry formally adopts separate stability designations for major parts of its GenAI specification without establishing an overall status, the question resolves YES only if the project's authoritative documentation or announcements explicitly characterize the general, foundational GenAI semantic conventions as Stable. Once a qualifying official Stable designation occurs within the forecast period, subsequent revisions or reversions do not change the outcome. An unmerged pull request, unofficial community statement, or third-party vendor announcement is insufficient evidence.

---

Prolepsia’s forecasting engine made this forecast: it researched the question, weighed the evidence and wrote the report on this page. A probability is not a promise; this one changes as the evidence does.
