AI Transparency & Use of Automated Systems

1. Our role and the EU AI Act New

We do not build or train foundation models. We integrate third-party general-purpose AI models into product features, which means that for the purposes of the EU AI Act we act principally as a deployer of AI systems, and as a provider only of the workflow features we build on top of those models.

Based on our current AI-assisted features, GBPPromote does not provide or deploy AI systems intended for the high-risk use cases listed in the EU AI Act. Our features generate and summarise business content. They are not used for biometric identification, employment or education decisions, access to essential services, creditworthiness, law enforcement, migration or the administration of justice.

2. Every AI feature, at a glance Expanded

Every feature in the platform that uses AI, which provider processes the data, whether the output can reach the public, and what human control applies.

If we add an AI-assisted feature, this table is updated before or at the point the feature becomes available.

3. The AI Connector New

The AI Connector works differently from our other AI features, so it deserves its own explanation.

Instead of us sending data to a model on your behalf, the Connector lets your own AI assistant. Claude, ChatGPT, or another assistant supporting the Model Context Protocol. Read your GBPPromote data directly, so you can ask questions about your locations in plain language.

How it works

• You initiate it. The connection is set up by you in Settings › Connections. It is off until you turn it on.
• You choose the scope. You authorise which locations the assistant can see.
• It is read-only. The assistant can read your data. It cannot change your listings, post content, reply to reviews or alter your account.
• You can revoke it at any time from the same settings screen, and from your AI assistant’s own connector settings.

If you are an agency connecting client locations, you should confirm your own client agreements permit sending that data to your chosen AI assistant. Our DPA covers our processing; it does not extend to a provider you separately contract with.

4. How automation is used for review replies

Review replies may be generated through our AI assistance module. The implementation details that matter:

• Automated generation. The system generates reply drafts based on the review text, rating and configured tone settings.
• Workflow control. You choose one of three modes:
  • Auto-post. AI replies post immediately, only when you explicitly enable this
  • Approve first. Replies are saved as drafts and must be approved by a person before posting
  • Manual only. AI suggestions are disabled and replies are written entirely by people
• Templates and editing. You can create human-written templates that the AI uses as context. Templates remain editable by people at all times.
• Audit trail. Generations, edits, approvals, rejections and posts are logged with timestamps, user IDs and action details.

Final publishing control remains with you. Auto-post is never on by default; it is a setting you must deliberately enable, and you can switch back at any time.

5. Data sent to AI providers

To generate a reply or draft, the system may send the following to the AI service:

• Review text or excerpt, with limited related metadata such as date and rating
• Business name and public profile details as configured in your account
• Selected tone, workflow rules or template instructions from your settings

Storage. Generated drafts and the minimal context used to create them may be stored in your account for auditing, editing, troubleshooting and service operation. Access is restricted to authorised personnel and your account administrators, under role-based access controls.

Providers are engaged as sub-processors under written data protection terms. See the DPA and section 10 of our Privacy Policy.

6. Do we train models on your data?

If our practices ever change, we will update this page and our Privacy Policy, adjust the relevant product settings, provide any required notices, and give customers the choices or controls the law requires. We would not make such a change quietly.

Note the exception in section 3: where you use the AI Connector, data reaches an assistant you control, and that provider’s own training and retention settings apply to your account with them.

7. Accuracy, bias & limitations

AI-generated content is produced by statistical models and may sometimes be incorrect, incomplete, inappropriate or inconsistent with your brand voice.

• Not perfect. Outputs may contain factual errors, awkward phrasing or suggestions needing review.
• Bias and moderation. Models can reflect limitations or biases present in training and system design. We apply moderation filters and content checks, but they are not foolproof.
• Context blindness. The model does not know your operational context: a reply that reads well generically may be wrong for a specific customer situation.

AI-assisted outputs may be unsuitable for a particular business, legal or customer-service context and should be reviewed before use where approval is enabled.

8. Human oversight & approval options

• Set default reply workflows: auto-post, approval required, or manual only
• Approve, edit, reject or regenerate any AI draft before publishing when approval mode is on
• Maintain and force-use custom templates written by people
• Disable AI assistance entirely at account or workflow level where supported

You decide whether AI content requires approval before publication. That decision sits with you, not with us, and it can be changed at any time.

9. Safety, filtering & content moderation

• Pre- and post-generation checks for profanity, hate speech, personal data leakage and other disallowed categories
• Monitoring and rate limits to detect unusual generation patterns that may indicate misuse
• Escalation rules so replies touching potentially sensitive content, legal claims, health references or threats, can be flagged for mandatory human review regardless of your workflow setting

10. Data protection considerations

• Data minimisation, only the minimum context reasonably needed is sent to the provider
• Legal basis, processing may rely on contractual necessity, legitimate interests, consent or another appropriate basis depending on the feature and jurisdiction. See section 3 of our Privacy Policy
• Processor agreements. AI providers and other sub-processors are engaged under contracts requiring appropriate technical and organisational measures
• Assessment: where required or appropriate, we assess automated processing risks and document mitigation measures, as set out in our GDPR Compliance statement

11. Opt-out & user controls

• Disable AI suggestions, turn off AI-assisted replies and content generation in settings where available
• Require approval, set replies to approval-required so nothing publishes without human sign-off
• Disconnect the AI Connector, revoke it from Settings › Connections at any time
• Request deletion, of AI-generated drafts and associated context, subject to legal, security and operational retention requirements
• Training preference. If any optional training or improvement workflow were ever introduced, you may contact us to record and enforce a preference to exclude your data

12. Labelling of AI-generated content New

Inside the product, AI-generated and AI-assisted drafts are labelled before review or publication, so users can identify machine-assisted content and decide what to do with it.

Content published to third-party platforms

Where AI assists in drafting content that is then published to a third-party platform such as a Google Business Profile, two limitations apply and we would rather state them plainly:

• Google does not currently provide a machine-readable field for marking a review reply or post as AI-assisted. We cannot apply a disclosure Google does not support, so this is a platform limitation rather than a choice we have made.
• The published content is yours. Whether AI assisted in drafting it, and whether that is disclosed in the text itself, is a decision that sits with you as the account holder, as does responsibility for the content, under section 18.

If you operate in a jurisdiction requiring disclosure of AI-generated communications to consumers, you should reflect that in the content you approve, and you should use approval-required mode rather than auto-post so you retain sight of everything published from your account.

Where a machine-readable disclosure standard becomes available and supported by the destination platform, we will implement it and update this section.

13. Auditing, logging & retention

For transparency and accountability we maintain:

• Logs of when AI generation was invoked and which user or system action triggered it
• Limited prompt and context records, truncated or minimised where appropriate
• Audit records of edits, approvals, rejections and postings, with timestamps and user identifiers
• Records of the model family or service version used, where the provider supports it

These records are retained only as long as necessary for security, support, compliance, dispute handling and service operation. Retention follows the periods in section 11 of our Privacy Policy, and generated drafts are removed or anonymised within a reasonable period following a deletion request or account termination.

14. No automated decision-making GDPR Art 22

Our AI-assisted features generate suggestions, drafts, summaries and recommendations. Publication behaviour depends on the workflow you configure, and where automation is enabled you retain the ability to disable it, require human review, edit or override output, or switch to manual workflows at any time.

We do not use AI for decisions relating to creditworthiness, employment, legal enforcement, eligibility determination, or profiling that materially affects the rights or freedoms of individuals.

15. No emotion or psychological profiling

Our features are limited to content generation, summarisation, text classification and related workflow assistance, based on textual input provided by users and publicly available business information. They are not designed or intended to perform emotion recognition, emotional inference or psychological assessment.

16. Special category personal data

You should not submit special category personal data. Including health data, racial or ethnic origin, religious or political beliefs, trade union membership, genetic or biometric data, or data concerning sex life or sexual orientation. Into AI generation features, unless you have a lawful basis and appropriate safeguards.

Where relevant, the system may flag potentially sensitive content for mandatory human review, but that is a safety net rather than a substitute for your own assessment.

17. AI literacy & internal oversight

We provide internal guidance and operational oversight for staff involved in deploying, supporting, monitoring or reviewing AI-assisted features. This covers awareness of system limitations, escalation handling, customer controls and responsible use expectations, consistent with the AI literacy obligation in Article 4 of the EU AI Act.

18. Liability & responsibility

AI suggestions support your workflows. You, as the account owner or authorised user, remain responsible for the final content published from your account. Including content drafted with AI assistance and content published automatically where you have enabled auto-post.

Our Terms & Conditions set out the liability limits and other contractual provisions applying to use of the platform.

19. Requests, complaints & escalation

For any AI-related request, human review, deletion of drafts, opt-out, or questions about an AI-assisted output, contact us using the details below.

Please include your account ID, the affected business profile, and the date and time of the relevant AI activity if known. That lets us locate the record in the audit log rather than asking you for it later.

If you have concerns about an output, workflow, moderation result or automation setting, you may contact us for clarification, support, review or escalation. Where appropriate we may ask for further detail to verify the request and investigate.

20. Contact & changes to this page

We may update this page as our AI-assisted features evolve or as legal requirements change. Material changes are posted with an updated effective date and, where appropriate, communicated to account administrators. When we add an AI-assisted feature, the table in section 2 is updated before or at the point that feature becomes available.