C2PA User Experience Guidance for Implementers

A unique aspect of this approach is that rather than attempt to determine the veracity of an asset for a user, it enables users themselves to judge by presenting the most salient and/or comprehensive provenance information. As such it is critical that users develop trust in the system itself, over the individual data presented. Exposing the C2PA Trust Model and Trust Signals in a way that balances transparency and intuitiveness is the critical design goal addressed here. There is no design pattern that can guarantee to engender trustworthiness across multiple contexts, and while a degree of contextual customization is anticipated, C2PA recommends all implementations adhere to the following general principles.

Implementations should be created using industry standard, robust user interface technologies.

Implementations should adhere to accepted, current accessibility standards to ensure no users are excluded. For an example of such criteria, see the Web Content Accessibility Guidelines (WCAG).

Wherever suitable, UX patterns should match those outlined here. In the case that this would break contextual paradigms of the platform, lean on precedent whether in the OS or app design. Users should not have to learn new paradigms or terminology in different contexts in order to access the information.

In many cases, a subset of the available information will be the most useful to a user in a given context. A link to the full information should always be made available, however.

Because the complete set of C2PA data for a given asset can be overwhelming to a user, some implementations may display only a fraction of available information. If applicable, a link to a more detailed display of information should be made available to allow the user to make a more informed judgement on the content’s trustworthiness.

C2PA requires the name "Content Credentials" to be used for provenance-enabled user experiences that follow the technical specification. It is critical that viewers of Content Credentials develop trust in the system itself, over the individual data presented. Therefore, implementors of Content Credentials must adhere to the proper usage of the Content Credentials name and visual marks to assure that each experience is consistent, coherent and cooperative.

The Content Credentials marks may be used in the following ways to provide consistency across the ecosystem:

The Content Credentials logo comprises an icon accompanied by a word mark set in Store Norsk Ja Medium, which is the cornerstone of the Content Credentials brand identity.

The icon shape is a pin, a metaphorical representation for applying (or “pinning”) Content Credentials to an asset, serves as an indicator of both the presence of C2PA data and its cryptographic validation status. It is also a navigation elements to reveal more C2PA information.

The icon is comprised of two lower-case characters, “cr,” which is a truncation of “credentials.” The characters are contained in an outlined circle with the lower-right corner angled to 90 degrees. Both the icon and brand word mark are set in Store Norske Ja Medium.

It’s proximity to other international attribution symbols, like copyright and Creative Commons, imbues the icon with a level of trust and authority. Its has been carefully considered and should not be altered or modified in any way.

Following the consistency principle, it is important in developing trust that users can learn to easily recognize the presence of the system and that their expectations are met regardless of context. As such, the L1 indicator should be consistent in its placement across all contexts and devices to the degree that it clearly represents the presence of C2PA data.

L1 can be applied using only the icon, the title, or both. If using the icon alone, ensure its appearance meets accessibility criteria so that consumers can easily identify its presence on or near the C2PA-enabled content.

For flexibility across implementations, there are several recommended placements for L1 indicators when C2PA-enabled content is present: on top of the content or somewhere close enough that its relationship to the content is clear. If L1 is positioned on top of the content, a hover state may be applied so as not to permanently obstruct the content below. Applications may differ depending on device, such as revealing L1 via long-press on mobile. Partially overlapping the L1 indicator over content is the most robust option to avoid potential misuse in the scenario that it has been purposefully added to content in a way meant to deceive. A new user experience guidance is recommended for initial implementation rollout to make clear how to identify L1 indicators.

Behaviour of L1 indicators should reveal L2 progressive disclosure, either via a hover or click interaction. Within L2 user interfaces, L1 can continue to be used to indicate the presence of C2PA-enabled ingredients.

If L1 indicates the presence of C2PA data, L2 is where consumers can begin to view and interact with the data. The minimum L2 user experience is defined as the display of nothing more than the base required C2PA manifest data: the signing entity, the claim generator and date. The signer is a top trust signal that allows the consumers to make their judgement trusting that the information available has been 'backed' by a legitimate entity. C2PA anticipates that additional varying assertion data will be included by implementers, and recommends including the manifest thumbnail, signer logo, and a link to L3 where consumers can find more provenance data if available.

L2 styles, fonts, etc. can be customized to fit the given context. Iconography and terminology used to describe assertions should be consistent wherever possible. C2PA anticipates the need for L2 generally to be space-efficient as it appears within the implementer’s context. Therefore it is suggested the data displayed be streamlined to provide enough information for the particular content, user, and context, allowing the consumer to understand to a sufficient degree how the asset came to its current state.

For content recommendations see the section on interface language.

L2 UI options are displayed along a spectrum from depth of information to breadth of provenance representation. In the figure below, the left side represents a single manifest summary and the right side represents the complete provenance summary depicting multiple manifests. However, the middle example shows a customized provenance summary wherein certain manifest assertions are highlighted. This flexibility allows implementers to show more depth for a given subset of manifests.

In order to summarize the data in a context-specific manner, it is recommended that consideration is given to what balance of depth and breadth of data will be most relevant to users. For example, if a given piece of content has only one manifest, consider displaying more of its assertion data, as in Figure 6. A single manifest summary. If there are multiple manifests, then depicting the entire provenance summary may give users a more complete understanding of the content’s history, such as in Figure 8. Provenance summary, three manifests. However, in all likelihood there will be many cases that fall in between, like in Figure 12. Customized summary combinations. The following sections will further document UI options along this spectrum.

A manifest summary only represents a single manifest. It could be the active manifest, as this is the most recent version tied to the C2PA-enabled content, or an ingredient manifest as determined by the implementer. The determination for displaying an ingredient manifest should be if there is substantive assertion data the implementer believes is more relevant to its audience than that of the active manifest. A shortcoming of displaying a single manifest summary is that it may not represent the complete history of the C2PA-enabled content, and may require consumers to navigate away from the implementer’s context to L3 to learn more.

A provenance summary represents the collection of manifests related to the C2PA-enabled content. Manifests should be presented in the order that they are referenced via ingredient assertions, starting with the active manifest at the top and ending with the origin ingredients below. Origin ingredients represent the beginning of their respective history branches.

Summary displays should show at minimum the origin ingredients and active manifests, or if screen real estate allows, with at least one manifest in between.

In order to provide users with a succinct summary and to allow for limited screen real estate, the manifests in between origin and active can be collapsed and represented as a numerical count. C2PA recommends a baseline rule for collapsing manifests if the total number exceeds four. However, this threshold can be altered according to context. In keeping with the core recommendations, a link to the full set of data in L3 should always follow the summary list of manifests.

When multiple origins are present, either as multiple ingredients in a single manifest or across multiple manifests, they can be summarized in the origin section of the UI. The L1 indicator can be used as a badge on ingredient thumbnails to distinguish C2PA-enabled assets.

L2 UI is flexible enough to allow for various combinations of manifest counts and origin assets.

Implementers should consider audience needs when weighing the balance of depth and breadth in combinatory summaries in order to concisely display the most complete and relevant representation of content provenance. For example, travel companies may opt to display location assertions from the origin manifest in a provenance summary so as to provide their users with key information. Similarly, a news publisher may choose to show action assertions from an ingredient manifest to disclose edits that occurred to published content.

There may be instances when C2PA-enabled content has been maliciously edited to tamper with C2PA data, or is signed by an untrusted entity. In this case, no additional prior data can be displayed.

For additional invalid states, see the section on interface language.

L3 should provide asset consumers as complete and comprehensive an overview of all relevant provenance data as is possible. It is here where consumers can parse through the entire provenance history and see assertion information across each manifest associated with that asset.

Implementers should weigh the display of assertion information based on the needs of their anticipated audiences. Overly technical assertions or custom assertions from third-party implementers can be left to L4 displays only. The following types of assertions can be considered appropriate for L3 to display:

In the above image, we propose a tree structure that shows each manifest or non-C2PA enabled ingredient. Each item is selectable, and shows its manifest assertions to the right. Each asset’s entire provenance chain should be viewable, and each individual manifest or ingredient inspected.

To understand changes between manifests over time, it may benefit users to have options to view two or more manifest ingredient thumbnails together for quick visual comparisons. This works well for static media like images, but will require different design solutions for temporal and non-visual media.

The L3 user experience should allow for any asset to be selected by the user locally or remotely, such as in the case where an asset is linked from an L2 display. Once selected, a manifest recovery search option can be available, which would allow users to select which "version" of the asset they want to see provenance information about. Read more about embedding remote manifests.

There are workflows where provenance data (assertions) need to be removed from previous manifests but the digital content (the media itself) is not changed. A potential scenario might be that an image capture manifest contains sensitive personal data (creator name, location) that needs to be removed before publishing an image as making that information public risks harm to the photographer. This would involve a redaction.

Redaction of data will always prompt a separate update manifest signed by the actor who performed the redaction. Update manifests signal that a change has been made, by who, to which manifest but that the media itself has not been altered. These should not be considered ‘edits’ of previous manifests and so it is recommended that UIs do not present them as such.

It is recommended that both the update manifest and the updated manifest (that where the redaction tool place) signify their state and reference one another. It is recommended that consistent patterns (language, iconography, interaction etc.) are used in the display of both manifests.

NB: For simplicity, both the update and the updated manifest are visible here within an L2 display. This will not always be the case however depending on how many steps there are in a provenance chain and the distance between the 2 manifests.

Note the lack of thumbnail image in the update manifest. Since update manifests do not involve changes to the digital content, they should not include any assertions that could suggest that i.e. actions, or a thumbnail.

It is up to individual implementations as to whether or not the details of the redaction are shown at L2, depending on use case. They must be displayed in L3 however where the additional context and a rationale for the additions can be given. For more on this see the actions section of the technical specification. Update manifests should contain onward journeys to L3 directing the user to the relevant manifest for closer review.

NB It is not recommended that updates/redactions are indicated at L1.

As generative AI reshapes the content we see, hear and otherwise consume, questions of authenticity and transparency arise. Audiences have the right to know when and how AI influences the information they consume.

Direct disclosure is recommended for all content where generative AI has been used as part of the production process.

This framework looks to communicate objectively what has happened technically. All disclosure should steer clear from attempting to suggest whether the use of generative AI in the production process has or could have impacted the meaning of a piece of content.

The table below presents a variety of examples of labels and descriptions for content disclosure. It provides a breakdown of how IPTC digital source types and C2PA actions are used for programmatic execution. Implementers should adopt as they deem most appropriate for their audience.

Note on AI-enhancements: Introducing a label for AI-enhancements was examined, however, it was concluded that such enhancements are not only anticipated but will also occur frequently. Given this frequency, the proposed label would be surfaced often, leading to its overuse. Over time, this could result in the label becoming redundant and losing its significance.

Note on Camera-captured: The Camera-captured label should be used to indicate a direct capture from a camera with no alterations. In reality, many cameras today use on-device software to make slight adjustments. These adjustments will most likely fall into the AI-enhancements category. C2PA can be implemented on the device to explicitly capture these steps.

Implementers may use color to attract the attention of the user in line with the design recommendations outlined in this chapter. If doing so, maintain a consistent color throughout the steps of progressive disclosure to provide a sense of continuity for the user and to reinforce the connection between related information.

All label categories should be the same color to depict a sense of neutrality. They should not be perceived to be on a scale, as part of a traffic light system nor infer any positive or negative meaning, for example, green and red.

Too many colors can increase cognitive load, create a disruptive interface and pose accessibility issues. Color perception varies among individuals, including those with color blindness. Lastly, as more labels are introduced in the future, it will become even more difficult for users to distinguish between them based on color alone.

At L1 only one label should be showcased at once.

In this section, the term creator means a party creating manifests. This could be the maker of an original work, an editor, or anyone else utilizing C2PA-enabled tools to add to the provenance chain.

As per the C2PA Guiding Principles, C2PA implementations should provide a mechanism for creators of a given content to assert, in a verifiable manner, any information they wish to disclose about the creation of that content and any actions taken since the asset’s creation. As such, the creator experience requires the following:

C2PA recommends an opt-in flow that concisely represents these requirements and can be opted-out of just as easily at any time. Once opted-in, creators should be able to distinguish between non-removable information as defined by the C2PA specifications and information that can be adjusted according to the user’s preferences.

Creators should be able to control the information in their assertions as much as is allowed by C2PA specifications. The Harms Modelling document covers the reasoning behind why creator control is imperative. To provide coverage against harms and misuse, C2PA suggests the following types of assertions be manageable by the creator:

To be accommodating to the user’s preferences, C2PA recommends presenting a UI wherein these assertion categories can be toggled on or off on a per document basis. To assist the creator in understanding the tradeoffs they are making, C2PA also suggests displaying a manifest preview that concisely and accurately depicts what information will be added into the manifest.

Some creators may want to reserve the right to not disclose the actions they’ve taken on a given piece of content. While some use cases, like photojournalism, should always show transparently what actions took place, this is less important for more creative and artistic applications. Those creators may want to protect their particular creation process and should therefore be allowed to opt out of including actions in their manifests. However, an exception to creator preferences may be required, such as disclosures for content where generative AI has been used. The signing entity should determine when the need for AI transparency supersedes creator preferences.

Granularity of actions is worth considering in creator implementations. In some cases, grouping actions into high level categories may be more understandable for consumers, versus presenting a list of detailed, creation-specific actions that may be unique to the implementation. C2PA recommends striking a balance between clarity and information overload based on the intended audience of the implementing platform.

Ingredient assertions represent a form of non-removable information because they are the key to the establishment of the provenance of an asset and may themselves contain provenance. As such, ingredients are a requirement to be displayed in the manifest and its creator-side preview.

Validation of an ingredient’s manifest is equally important to convey to creators prior to producing a new manifest. Within the manifest preview UI, C2PA recommends displaying a list of ingredient thumbnails and their validation states to ensure the user working with those ingredients is aware of additional provenance data. This is particularly important for cases when ingredients are unable to validate or contain an invalid manifest. A clear example might be a news editor who receives a piece of user-generated content that purports to depict a controversial scene - alerting the editor to the validation state of that image will give them stronger assurances of whether that content is trustworthy.

It is important to provide clear guidance about the different manifest storage options available to creators. Creators want or need to know where their manifest data is being stored, either locally or remotely, when they export their content. The following options may be presented to creators regarding the location of their manifest data:

Describe each export action clearly and specifically. The action of storing remotely is more akin to “publishing to,” “storing in,” “saving to” remote storage or “publishing,” “storing,” “saving” remotely, compared to the action of storing locally or “attaching” a manifest to a file.

The tradeoffs between local and remote storage should be conveyed to creators, whether that be in-product or in separate documentation.

Attaching manifests to files directly generally keeps them more private. However, this comes at the expense of increased file sizes. Manifest data attached to files directly can also be stripped from those files when they are published on any platform online that processes uploaded files for size reductions.

Storing manifests remotely can generally make them more resilient, persistent, and recoverable, with the added benefit of not contributing to file size. However, because remote manifest recovery is achieved through searching for soft binding matches, remotely stored manifests and their respective content can potentially be viewed by anyone. This means that remote manifest storage is inherently less private.

The C2PA Implementation Guidance recommends that a manifest be created for an asset when a significant event in the lifecycle of the asset takes place, such as its initial creation or an "export" operation from an editing tool. This is in part due to the underlying technical process of digitally signing the manifest, but also aligns with natural creation workflows. When a creator is ready to export their work, they should be able to decide whether or not to attach the accumulated assertion data to their content.

It is important to align any interface language with the recommended terminology and phrases. This lets terminology retain the same meaning and weight across experiences, which is best for user comprehension. It also ensures the value and adoption of content provenance, thus making C2PA data easier to understand and promote across a large ecosystem.

For customization of UI elements, use the recommended terms or customize with user comprehension in mind.

In the following table, we recommend terminology for consumer-facing L1 UI:

Please refer to L1 interface language for a reminder of the importance of adhering to recommended terms and descriptions. The following table contains terminology for UI elements such as commonly used categories and validation states:

The names and descriptions of C2PA actions have generally been written with implementers in mind. Users who are less familiar with the C2PA will benefit from clear labeling and descriptions of actions that capture as much of the related actions that may apply as possible.

Here we provide a matrix of current C2PA actions with recommended labels and optional descriptions for them in the "Edits and activity" section of consumer manifest UIs.

Note that when displaying C2PA actions ("Edits and activity") in a manifest, descriptions may accompany them in the UI or be made available in separate docuentation.

When writing user-facing error or warning messages, space may be limited. Prioritize clearly stating what is wrong, missing, unavailable, etc., and providing next steps a user can take to potentially correct the situation when applicable. Explanation of why an error or warning occurred, or more detailed guidance on correction,may be left to separate help documentation provided through a "Learn more" link or similar.

Following are a few example messages for some of the most common types of issues and errors C2PA is aware of today, which you may use as a starting point and reference for your own implementations.

Some C2PA terms and concepts may be particularly unfamiliar, hard to grasp, or contentious for users, or be inherently lengthy to explain (running the risk that explanations will not be read and understood).

Here we list a few known examples and share possible alternatives. These suggestions are currently untested, but you may take them as directional guidance toward simpler, more user-friendly interface language.

Correctly identifying and displaying trust signals is of paramount concern for our overall user experience. C2PA strives to understand the value consumers will apply to content attribution through ongoing user research studies and usability testing.

Pending user research, C2PA will provide implementors with recommended L2 customizations based on combinations of content, audiences, and platforms. Examples may include news publishers, social sites, e-commerce and retail, travel, and entertainment platforms. The recommendations for video UX will also continue to expand in scope and complexity, along with the introduction of other media formats like audio and streaming content.