AI Transparency Notice

This notice explains how Linqyou uses artificial intelligence, what it sends to the model, and what role it plays in the numbers and advice you see. We use a general-purpose model from a third-party provider; we do not train models. None of our uses is high-risk under the EU AI Act: in particular, Career helps a job seeker with their own search and is never used by an employer to select candidates.

The rule we build to

The AI never computes a number. Every statistic on your dashboard — impressions, growth rates, comparisons — is calculated by ordinary, deterministic software from the data in your uploaded LinkedIn exports. The AI model is shown those already-calculated numbers and asked only to write about them: to explain what they might mean and suggest what to do next.

This means a number on your dashboard is never a guess, an estimate, or something the model inferred from a data field it wasn’t given. If a figure isn’t present in your export, we say so — we do not ask the model to fill the gap.

What is AI-generated

Content written by the model is visually marked throughout the product with an “AI-written” label, and always states which of your own metrics it is based on. AI-generated content includes:

• Written interpretation of your metrics (why a number might matter)
• Suggested next actions
• Answers in the AI chat feature
• Choosing a topic for each of your posts from a fixed list; ordinary software then does the counting
• Content: suggested post ideas (the engine picks them from your own results; the model only rewords them) and a first draft of a post, written from the stories you choose, a few of your own posts (or, with fewer than eight posts, a writing sample or comments you wrote) as style examples, a short list of measured writing habits such as sentence length and punctuation, and your company profile. You can also ask the model for other opening lines, or to change a draft as you describe (shorter, less formal, in Dutch); it works only from what the draft was written from and what you typed. A campaign draft also gets a short model-written note on how it fits the campaign brief. A draft is marked “AI-written first draft”. It is checked by ordinary software so that it states no number, name or quotation that is not in what you supplied, but it is a draft: you read it, change it and post it yourself. Linqyou never posts anything for you.
• Career: a short message to one person on your shortlist, only when you choose “Write with AI”. It is written from your own headline and declared skills, the roles you typed, and that person’s first name, headline and employer as LinkedIn shows them; nothing else about them is sent. It is checked by ordinary software so that it states nothing that is not in what you supplied, and you edit and send it yourself. Separately, job titles that our rules cannot place (for example “barista”) are sent on their own, without any name or company, so that each title can be filed under a field and a level; the answer is stored against the title, not against a person.
• Labelling each comment people left on your posts with one comment type and up to two of your own subjects, chosen from a fixed list, for counting purposes. The comment text is sent for this and the label is stored; the text is not kept in the label table, and the model is never asked to describe or judge the person.

Content that is not AI-generated, and never carries the label, includes every chart, every raw count, every percentage, and every comparison shown elsewhere on the dashboard.

Where inference happens

Requests go from our servers directly to Emby (api.emby.ai), the model provider named on our Sub-processors page. If a request to one Emby model keeps failing we try the other one, so no further provider is involved. We only use a model that runs in the EU, under a written agreement that the provider does not keep what we send after answering and does not use it for training. Our software refuses any model that is not on that confirmed list. We do not use batch modes, which usually involve keeping the content.

Two kinds of input are sent for interpretation: pre-computed metric summaries — for example { "reach.median_daily": 2.5, "reach.mean_daily": 31.8 } — and the text of your own posts, when the analysis is specifically about how those posts are written. For Content the stories you chose to write from are sent too, and short audience questions taken from comments on your posts may be used to suggest an idea; a draft never quotes or names a commenter. Comments by other people are sent one at a time to label them and, in Leads, to help you reply to one. For research, the profile facts we stored are sent to write one message. What is never sent: raw uploaded export files, connection lists, private messages, or contact details.

How AI input is kept separate from benchmarks

We do not use raw posts, prompts, AI chat transcripts, or user-specific AI narratives as a benchmark dataset or to train third-party models. Where privacy review enables benchmark work, it uses normalized, benchmark-safe features under the separately disclosed data-use model—not source wording. Generalized text guidance fails closed if it overlaps source wording, and it is never a reproduction or sale of a member’s post.

No fully automated decisions with legal or similarly significant effect

Linqyou does not use AI to make decisions about you that produce legal effects or similarly significantly affect you. Generated insights are informational suggestions for you or your workspace to act on at your own discretion; the product does not, for example, automatically act on your LinkedIn account or make hiring, credit, or eligibility decisions. Ranking people in People to engage and Leads is profiling, used only as a suggestion to a person who decides what to do.

Feedback

Every AI-written insight can be rated. That feedback is kept with your account, used only to improve the quality of future insights, never for advertising and never shared with a third party.