Iframely Publisher Policy

Effective October 1, 2026

This page explains how Iframely accesses and uses publisher content, including URL previews, rich media embeds, and automated processing, including AI-related use. Our goal is simple: make content work across the web while respecting how publishers want it to be used.

1. How Iframely works

Iframely helps applications display links and content in a structured, consistent way. When a URL is shared, we retrieve and format the content so it can be embedded, previewed, or represented in user-facing systems.

We don’t create content and we don’t take ownership of it. Our role is to act as a technical intermediary between publishers and applications, applying publisher rules consistently across all customers and helping our customers understand those rules when needed.

This means you don’t have to manage dozens of integrations or enforce policies individually. Iframely acts as a common layer that represents and enforces your preferences to a wider audience.

2. Two different use cases

Not all access to content is the same, and we treat it that way.

Preview and embed use is about showing content to users, like link previews, cards, or rich media embeds. For this, we focus on structured data such as metadata, summaries, and embed-ready formats, relying on open protocols designed for this purpose: Open Graph, oEmbed, and similar standards.

Automated processing, including AI-related use, involves processing content for enrichment, indexing, or grounding in automated systems, including AI systems. This type of access is handled separately and may be more restricted.

These two use cases are kept clearly apart for each party involved.

3. Three different bots

Iframely runs three separate web robots: the Iframely Preview Agent, the Iframely Processing Agent, and the Iframely Training Agent. These robots use different request patterns depending on the use case, and publishers can detect and apply rules to each independently.

The Preview Agent handles preview and embed use. The Processing Agent handles automated processing other than Training Use — indexing, enrichment, contextual grounding, and AI-assisted retrieval. The Training Agent handles Training Use specifically.

Using a separate identity for Training Use means a publisher’s training preference doesn’t affect their indexing, search, or assist availability — each can be allowed or restricted independently, without one affecting the other.

Preview, processing, and training requests can be distinguished through signals such as user-agent identifiers, request headers, and verification mechanisms like signed requests, including emerging standards such as Web Bot Authentication. All three robots can be verified as Iframely-run using Web Bot Auth, reverse DNS checks and IP allowlisting.

These requests represent different types of access and can be identified separately, allowing publishers to apply different rules to each use case, for example permitting previews but restricting automated processing, including AI-related use. Where supported, you can also target specific applications or customers using request-level signals or configuration in the Iframely dashboard.

Refer to the detailed Iframely Preview Agent, Iframely Processing Agent and Iframely Training Agent documentation for full technical details.

4. Crawling etiquette

Iframely does not perform broad web crawling and instead operates on a request-driven basis; we act on behalf of our customers, one URL at a time. Iframely agents respect your compute capacities and bandwidth.

The Iframely Preview Agent usually fetches the page’s metadata section and may validate referenced images, videos, iframes, or audio files by requesting only headers or the first few bytes. This keeps the impact on your servers minimal.

The Iframely Processing Agent respects your robots.txt directives and does not attempt to fetch sensitive paths. It does not retrieve full content unless permitted based on explicit publisher signals or technical indicators of how content is made available, interpreted conservatively and subject to other applicable restrictions.

The Iframely Training Agent accesses only the content made available for Training Use, as described in Section 6. By default, this is limited to structured, publisher-declared metadata and referenced media links; access to anything beyond that for Training Use requires an explicit Publisher Control permitting it. The Training Agent respects your robots.txt directives and any training-specific restrictions, and does not attempt to fetch sensitive paths.

5. Your data, control and signals

You remain in control of how your content is accessed and what data is provided to each robot, including targeting individual Iframely customers.

Iframely looks at a combination of page signals to understand your preferences, including robots directives, HTTP headers, structured metadata, open data protocols, licensing information, and delivery configuration. We interpret these signals practically and apply them consistently so that your rules don’t need to be reimplemented for every customer integration.

You can configure how your preferences are applied via Iframely’s dashboard or by contacting us for clarification.

This includes the ability to opt out of Training Use specifically — independent of your preview or general automated-processing preferences — by providing a Publisher Control that restricts or prohibits Training Use, or via Iframely’s dashboard where available.

Refer to content access document for technical description of page signals and default Iframely behaviour.

6. Automated processing boundaries

Iframely supports limited automated processing use cases, including AI-related use, such as indexing, categorization, enrichment, contextual processing, and Training Use.

Structured, publisher-declared metadata and referenced media links (for example, titles, descriptions, and images) that are already available through preview and embed use are available by default across every mode of automated processing, including Training Use. Access to anything beyond that — additional structured data, excerpts, or fulltext — requires an applicable Publisher Control permitting it for the relevant use.

Where publishers allow or do not restrict indexing or similar automated access through technical signals (such as robots directives, metadata, licensing information, or delivery formats), Iframely relies on those signals to determine whether and how content may be accessed for automated processing for such purposes.

Iframely may process content to extract structured data, generate previews, or perform limited transformations such as summarization, solely to support these permitted use cases.

Training Use follows the same signal-based model: the default Training Use described above is available unless restricted by a Publisher Control. Where a Publisher Control permits broader Training Use, Iframely may make additional content available for that use. Where no such Publisher Control is present, access beyond the default is not available.

Publishers may express permissions or licensing preferences through metadata, machine-readable signals, or commercial mechanisms such as pay-per-use access. Iframely preserves and communicates these signals and applies them operationally, but does not treat them as independently granting any legal rights in content.

Iframely may impose additional technical, contractual, or pricing conditions on Training Use beyond what a Publisher Control permits. Iframely does not assume responsibility for a customer’s compliance with publisher terms.

In all cases, Iframely acts solely as a technical intermediary and does not grant rights in publisher content.

7. Full content and default access

By default, Iframely relies on structured representations of content — including metadata, images, summaries, and embed-ready formats — which are generally sufficient for previews and interoperability.

The Iframely Preview Agent does not access fulltext content.

Iframely Training Agent does not access fulltext by default. Access to fulltext content by the Iframely Processing Agent is limited, conditional, and purpose-specific. Where permitted, such access is used only to support indexing, categorization, summarization, or representation of content, and not for storage, reuse, or redistribution beyond what is operationally necessary for those functions.

Iframely relies on explicit or commonly recognized technical signals — such as robots directives, HTTP headers, metadata, licensing information, or controlled delivery formats — to determine whether fulltext access is appropriate. Where such signals restrict or prohibit automated access, Iframely does not access full content.

Content made available in machine-readable formats (such as APIs or structured document formats) may support automated processing for the purposes described above, but does not, by itself, imply permission for unrestricted access, reuse, or Training Use beyond the scope provided by Iframely.

Fulltext access granted for one purpose does not imply permission for another. For example, fulltext access permitted for indexing or summarization does not itself authorize Training Use, and vice versa — each is governed by its own applicable Publisher Controls, as described in Section 6.

Publishers may explicitly allow or restrict fulltext access for any given use, including Training Use, through technical signals or direct configuration.

Iframely minimizes content retrieval wherever possible and prefers partial or structured representations when sufficient.

In cases of ambiguity, Iframely applies restrictive defaults and may limit or avoid fulltext access.

8. Attribution and integrity

We preserve source information, links, and metadata. Where licenses or publisher terms require attribution or other conditions, we pass that information along so it can be respected downstream.

9. Enforcement and responsibility

Iframely applies your preferences consistently across all customers. If signals indicate that certain uses are not allowed, we restrict or block access accordingly. If we detect misuse or inconsistencies in how content is requested or processed, we may take action to enforce your intended rules; however, responsibility for compliant use ultimately remains with the application accessing the content.

We prioritize enforcement of explicit publisher restrictions where technically feasible.

You can always reach out to report concerns or clarify how your content is being handled.

10. Evolving standards

The way content is accessed — especially for automated processing, including AI-related use — is evolving rapidly. New standards for identifying bots, expressing permissions, and delivering machine-readable content are emerging. We follow these developments closely and adopt them where they help better represent publisher intent.

This includes emerging standards for expressing permissions for automated processing, including AI-related use (such as training or reuse signals), identity-aware access controls, and monetization mechanisms. Iframely may adopt and support such standards to better reflect publisher intent and enable compliant use of content in automated systems, including AI systems.

11. In short

Iframely helps applications display and embed web content in a structured, consistent way while respecting publisher preferences.

We separate preview/embed use, automated processing, and Training Use, running three distinct robots — the Preview Agent, the Processing Agent, and the Training Agent — that can be identified through headers, user agents, or dashboard signals.

By default, we make structured content — including publisher-declared metadata, links, and preview or thumbnail images — available across our automated processing modes, including Training Use. Access to additional content, such as additional structured data, excerpts, or full text, depends on the applicable use and Publisher Controls.

Training Use beyond the default is available where a Publisher Control permits it. The Training Agent is separate so that a publisher’s training preference can be applied independently of indexing, assist, or other automated processing.

Publishers retain all rights, and Iframely acts solely as a technical intermediary, communicating and enforcing publisher permissions without granting independent rights.

We minimize content retrieval wherever possible, preserve attribution and metadata, and favor restrictive interpretations in cases of ambiguity.

All uses remain subject to publisher permissions, and Iframely does not grant content rights independently.