Overview
This Framework covers the website side of the EU AI Act, the EU's law for companies that build AI (providers) and companies that use it (deployers). AI tools on your website, such as chatbots, AI-powered personalization, and tools on job or loan applications, are easy to lose track of.
This Framework helps you find them, confirm visitors know when they're interacting with AI, and flag tools that may be high-risk.
The AI Act entered into force on August 1, 2024, and applies in stages. Bans on certain AI practices have applied since February 2025, and transparency rules since August 2026. The high-risk rules, postponed by the 2026 Digital Omnibus on AI, apply from December 2027, or August 2028 for AI built into regulated products.
Each Framework aligns with ObservePoint capabilities, but it also includes recommended processes, documents, and policies that we recommend an organization consider in their pursuit of excellence. Checks reference the official text of the EU AI Act (Regulation (EU) 2024/1689), as amended by the Digital Omnibus on AI (Regulation (EU) 2026/1744).
Note: This Framework is provided for informational purposes only and does not constitute legal advice. Consult qualified legal counsel regarding your organization's specific EU AI Act obligations.
Framework
This Framework is organized into the following Policies:
AI Inventory - Finding and documenting every AI tool on your website
AI Transparency - Telling visitors when they're interacting with AI
High-Risk AI Touchpoints - Identifying AI tools on pages where the AI Act's high-risk rules may apply
Documentation & Processes - Recommended documents and governance processes
Each policy contains individual Checks.
For each Check, we include an implementation & remediation guide.
Each check is marked with the 🛠️ ✅ icons, which mean the Check requires additional configuration. Most of these need a manual review of which tags are AI tools, or a Journey that interacts with an AI tool on the page.
In addition to checks, there are recommended documents (📄) and processes (⚖️) we encourage you to maintain as part of an effective governance program.
Setting up your AI Register in ObservePoint: Several checks compare your website against your AI Register (see Documentation & Processes). Your register lives outside ObservePoint, so you mirror it by adding its AI tools to an approved Consent Category and assigning that category to your Audits. ObservePoint then flags any AI tool it finds outside the category as unapproved. When your register changes, update the category to match.
Note: ObservePoint shows you what AI tools run on your website and what data they receive. Most of the AI Act covers how AI systems are built, tested, and managed behind the scenes, such as risk management, technical documentation, and human oversight. ObservePoint can't see those, so use this Framework as one part of a broader AI governance program. The AI Act also doesn't replace the GDPR (Article 2(7)), so AI tools that collect personal data must follow its rules on consent, data minimization, and data transfers. To check these, see our GDPR Compliance Framework.
Scope & Frequency
Before diving into the specifics of each policy, we want to provide our recommendation around the scope and frequency of monitoring the AI tools on your website.
Frequency | Scope | Goal |
Daily | Pages with chatbots or AI assistants, plus job application, loan, insurance, or enrollment pages | Confirm AI disclosures are showing and no unexpected AI tools appear on your most sensitive pages |
Every Deployment | 10-15% of pages or 100% of templates | Catch AI tools added or changed by new code, tag manager updates, or vendor releases |
Quarterly | 100% of pages | Keep your AI Register complete and catch "long tail" AI tools on rarely visited pages |
Run your Audits from an EU location (for example, Dublin or Frankfurt) so they see your site the way EU visitors do.
If you're not sure how many pages exist on your domain, ask your Success Manager or our Support Team about running a Site Census scan.
AI Inventory
This policy represents checks that help you find and document every AI tool running on your website.
🛠️ ✅ Every AI tool on your website is identified
Manual review required.
You can’t meet your AI Act obligations for an AI tool you don’t know about. ObservePoint’s Cookie & Tag Database can give you a starting list of the AI tools on your site, such as chatbots and AI-powered personalization. Review the list, then record the confirmed AI tools in your AI Register.
Implementation & Remediation Guide: Link coming soon
Source citation: EU AI Act, Article 3 (definitions of "AI system" and "deployer")
🛠️ ✅ Every AI tool is approved and listed in your AI Register
Your AI Register (see Documentation & Processes) is your organization's list of approved AI tools, with each tool's purpose, owner, and AI Act risk level. This check compares the AI tools ObservePoint finds on your site against that list. Any AI tool on your site that isn't in the register is a finding.
Implementation & Remediation Guide: Link coming soon
Source citation: GDPR, Article 30 (records of processing activities); for high-risk AI, EU AI Act, Article 26 (obligations of deployers of high-risk AI systems)
🛠️ ✅ AI services called directly from your pages are identified
Some websites build their own AI features, such as a custom chatbot or AI-powered search, that send requests straight to an AI service like OpenAI, Anthropic, or Google AI. These don't appear as known tags, so they're easy to miss. Review the network requests your pages send to AI service domains, and add any AI features you find to your AI Register.
Implementation & Remediation Guide: Link coming soon
Source citation: EU AI Act, Article 3 (definitions of "AI system" and "deployer")
AI Transparency
This policy represents checks that help make sure visitors know when they're interacting with AI. These rules have applied since August 2, 2026.
🛠️ ✅ Chatbots and AI assistants tell visitors they're talking to AI
Manual review required.
The AI Act requires AI chatbots to tell people they're interacting with AI, unless it's obvious. This duty falls mainly on the chatbot provider, but the chatbot runs on your site, so check that the disclosure appears. It must show up by the visitor's first interaction and meet accessibility requirements. A person still needs to review whether the wording is clear.
Implementation & Remediation Guide: Link coming soon
Source citation: EU AI Act, Article 50 (transparency obligations); European Commission: Guidelines on the transparency obligations for providers and deployers of AI systems
🛠️ ✅ Tools that infer visitors' emotions are flagged for review
Manual review required.
Some tools infer how visitors feel. If a tool does this using biometric data, such as a face or voice, visitors must be told, and it’s banned in workplaces and schools. Flag tools that mention emotions in ObservePoint’s Cookie & Tag Database, then ask your legal team to confirm whether the rules apply.
Implementation & Remediation Guide: Link coming soon
Source citation: EU AI Act, Article 3(39) (definition of "emotion recognition system"); EU AI Act, Article 50(3) (informing people about emotion recognition); EU AI Act, Article 5 (prohibited AI practices)
High-Risk AI Touchpoints
This policy represents checks that help you find AI tools on pages where the AI Act's high-risk rules may apply. The AI Act lists certain uses of AI as high-risk, such as AI that screens job applicants, scores creditworthiness, prices life or health insurance, or decides admission to education. For these stand-alone systems, the high-risk rules apply from December 2, 2027.
🛠️ ✅ AI tools on high-risk pages are identified
Manual review required.
Review the AI tools that run on pages such as job applications, loan or credit applications, insurance quotes, and school enrollment. An AI tool on one of these pages isn't automatically high-risk, but it should be reviewed by your legal or compliance team. Once the review is done, ObservePoint can flag automatically any tag that wasn't approved for these pages.
Implementation & Remediation Guide: Link coming soon
Source citation: EU AI Act, Article 6 (classification of high-risk AI systems); EU AI Act, Annex III (high-risk AI systems)
🛠️ ✅ People are told when high-risk AI helps make decisions about them
Manual review required.
When a high-risk AI system helps make decisions about a person, such as screening a job application, the deployer must tell that person. On a website, this is usually a notice on the application page or in the linked privacy notice.
Implementation & Remediation Guide: Link coming soon
Documentation & Processes
This policy represents recommended documents and processes pertaining to AI governance.
📄 Create and maintain an AI Register
A list of every AI tool your organization uses, including those on your website. For each tool, record its purpose, owner, vendor, where it runs, what data it receives, whether your organization is the provider or deployer, and its AI Act risk level (banned, high-risk, transparency rules, or minimal). This is the source of truth your ObservePoint checks compare against.
📄 Document approved AI disclosure wording
Approved wording for chatbot disclosures, AI-generated content labels, and notices on pages where high-risk AI is used. This gives every team the same language and makes disclosures easy to review.
⚖️ Review new AI tools before launch
Before a new AI tool or AI feature goes live, review it against your AI Register. Confirm its risk level, check that it isn't a banned AI practice (such as AI that manipulates people or recognizes emotions in the workplace), and confirm the right disclosures are in place.
⚖️ Label AI-generated content
If you publish AI-generated or AI-altered images, video, or audio (deepfakes), or AI-generated text that informs the public on matters of public interest, it must be labeled as such. ObservePoint can't tell whether content was created by AI, so this needs an editorial process.
⚖️ Provide AI literacy training
Staff who use or manage AI tools should understand how they work and what risks they carry. The AI Act asks organizations to support AI literacy among their staff.
⚖️ Prepare high-risk AI tools before December 2027
For any AI tool your legal team identifies as high-risk, work with the vendor to confirm how it meets the AI Act's requirements, and plan for your own deployer obligations, such as human oversight and keeping logs.
Conclusion
This framework gives you a clear view of the AI on your website. By monitoring it with ObservePoint, you can keep your AI Register complete, make sure visitors know when they're interacting with AI, and spot high-risk AI tools well before the high-risk rules apply.
As ObservePoint unlocks more data and enhances its platform, we will support more checks to be fully automated. Expect this framework to evolve over time, and talk to your Customer Success Manager if there are other checks you'd like to implement.
