AI content transparency

What Viewfy generates

Viewfy is a growth agent for founders. On a customer's behalf it drafts blog posts, the hero and infographic images inside them, social posts, replies in community threads and outreach emails. Text is written by Claude models from Anthropic. Images are produced by image models hosted on fal.ai, currently Google's Gemini image models and Recraft as a fallback. Viewfy is the provider of this generating system. The founder who publishes the output on their own channels is its deployer.

How text is marked

Claude models embed an imperceptible statistical watermark in the text they generate. Viewfy keeps that text as generated apart from typographic clean-up (quote normalisation, dash replacement, a short list of word swaps), so the watermark survives publication. Viewfy does not run a second model over finished text to disguise it. Anthropic publishes a checker for files at claude.com/check-content and a detection API for regulators, researchers and organisations with Article 50 duties.

How images are marked

• A signed C2PA manifest (Content Credentials) is embedded in every published image. The untouched model output is attached as the parent ingredient, so the image model provider's own manifest and any pixel watermark it applied (such as Google's SynthID) are preserved through Viewfy's resizing and compression.
• IPTC XMP metadata sets DigitalSourceType to trainedAlgorithmicMedia and a credit line reading "AI-generated image".
• When the founder has switched on the reader-facing label for their blog, a visible "AI generated" badge sits on the hero image and the post states that its images are generated.

Verify any image at contentcredentials.org/verify or with an EXIF/XMP reader.

How posts are labelled

Each blog has a "Show AI label on posts" setting. The machine-readable marks below are always present; the reader-facing notice, hero badge and footer line appear when the founder has switched the label on. Whether a visible label is required for a given post is the founder's responsibility as the deployer.

• Reviewed posts. When the founder or a team member reads a draft, checks its facts and approves it, the published page says "Drafted with AI. Read, fact-checked and approved by [name] on [date] before publishing." Viewfy records who approved it and when. If a model later rewrites the post, the review is void and the post returns to draft until a person approves it again.
• Auto-published posts. When the founder has switched on auto-publish, the page says "Generated with AI and published automatically" and names the company responsible for the content.
• Editorial responsibility. Every blog footer names the person editorially responsible for it and how to reach them.
• The same facts are published in machine-readable form in each page's JSON-LD (creditText, accountablePerson, image creditText) and in the blog's llms.txt.

Social posts and community replies

Social posts and thread replies are short and are approved by the founder before posting unless they have chosen auto-publish for that channel. Images attached to them carry the same C2PA and IPTC marks as blog images.

Contact

Questions about marking, labelling or a specific piece of content: email m@viewfy.ai. Terms covering AI-generated content are in the Terms of Service.