AI Disclosure

This page explains exactly how AI is used in producing the With Purpose personalized PDF Report (“the Report”), what we do and don’t do with that AI, and how our process complies with the European Union AI Act and broader transparency principles.

We believe customers deserve to understand AI authorship plainly. The short version is below; the longer one follows.

The short version

Your Report is drafted by an AI (currently Anthropic’s Claude language model) and then reviewed line-by-line by a human (Adam) before being sent to you. The AI does the synthesis work. The human does the editorial work. The voice you read is the editor’s; the structure is the AI’s; the words belong to both.

We do not use AI to:

• Make decisions that affect you legally or financially
• Generate the Report and send it to you without human review
• Train language models on your survey responses
• Produce content we cannot stand behind editorially

What the AI does

When you complete the survey and pay, the following happens:

1. Your survey responses are scored against five validated psychological instruments (BPNSFS, Schwartz PVQ-21, Ikigai-9, MLQ, RIASEC)
2. The scores are matched to one of 25 business-recommendation clusters in our internal catalogue
3. Your survey responses, the matched cluster, and reference data (vendor information, your country’s tax and legal context) are sent to Anthropic’s Claude API
4. Claude generates a draft of the Report — roughly 30 pages of structured markdown — based on a detailed system prompt we have written and continuously refine

The AI is Anthropic’s most advanced model at the time of generation. The system prompt that guides the AI is roughly 9,000 tokens long and contains explicit rules about voice, structure, what to quote verbatim from your responses, what not to invent, and what country-specific information to use.

What the human does

Once the AI has produced a draft, Adam personally reviews the document line by line. The review takes 30–60 minutes per Report and includes:

1. Verification that vendors recommended in the Report exist and are priced correctly
2. Verification that legal and tax facts about your country are current and accurate
3. Editing for voice, tone, and clarity
4. Removing any AI-generated content that feels generic, formulaic, or off-brand
5. Confirming the Report quotes your survey responses faithfully and does not put words in your mouth
6. Approving or rejecting the Report against a 12-criterion rubric

Reports that don’t meet the rubric are regenerated or rewritten. You only receive a Report that has Adam’s signed-off approval.

This is the deliberate moat against AI-generated content of variable quality. Most automated $79 products send you whatever the AI produced. We don’t.

What we do NOT do

To be unambiguous:

• We do not use AI to make decisions about you. The cluster matching is a deterministic, score-based algorithm. The AI generates text from that match; it does not decide your match.
• We do not use AI for any automated decision that has legal or significant effects on you. The Report is informational; you remain the decision-maker.
• We do not use your survey responses to train AI models. Anthropic’s API policy is that customer inputs and outputs are not used for training. We additionally do not export, share, or repurpose your data for any AI training.
• We do not use AI to write the editorial review. The 12-criterion rubric and the editorial decisions are made by Adam.
• We do not generate Reports without human review. Every Report sent to a customer has been read by Adam.
• We do not use AI to generate this disclosure page. This page was written by Adam (and edited with AI assistance for clarity, in the same way many writers use spellcheckers).

EU AI Act compliance (Article 50)

The European Union AI Act, which entered into force in August 2024, requires that AI-generated content be detectable to recipients (Article 50). We comply with this requirement in several ways:

1. This AI Disclosure page is linked from our footer on every page of withpurpo.se
2. The landing page explicitly mentions AI authorship in the “Trust Signals” and “FAQ” sections, before any purchase decision
3. The Report itself does not duplicate this disclosure (per a deliberate design decision to avoid disclosure-fatigue inside a paid product), but customers have already encountered the disclosure on the website before paying
4. Generated content is not represented as human-authored without qualification — both the website and this page identify the AI’s role transparently

We do not use AI in ways that fall under the AI Act’s “high-risk” category (Annex III). We are a small, optional, informational service; we are not a recruitment platform, a credit scorer, an education-admission system, or any other regulated category.

Models we use

Currently:

• Anthropic Claude (Opus tier) — for drafting the Report
• No other AI models — we do not use OpenAI, Google, or Meta models in production
• No image generation — the visual elements in the Report (typography, ornaments, palette samples) are designed by humans

We may change models in the future as quality and pricing evolve. Material changes (e.g., switching to a different vendor, using a model trained substantially differently) will be reflected on this page within 30 days.

How your data is used in AI processing

For full details, see our Privacy Policy, Section 5.

In summary: only the data needed to generate the Report is sent to Anthropic. Your email address, payment information, and full identity are NOT transmitted. The data is sent under Standard Contractual Clauses (the EU’s GDPR-compliant transfer mechanism). The data is not used to train models.

Our editorial principles

We follow a small set of principles for AI use:

1. AI for synthesis, human for judgment. The AI synthesizes across many signals to produce a coherent draft. The human judges whether that draft is fit to send.
2. No fabricated personalization. The AI is forbidden from inventing quotes, fabricating user context, or “improving” verbatim user responses.
3. No income claims. The AI is forbidden from making promises about earnings, success rates, or outcomes.
4. Specificity is mandatory. Every operational recommendation in the Report must include a named vendor, a price range, and a verification source — drawn from our maintained reference databases, not invented.
5. Country accuracy. The AI is forbidden from naming entity types, registries, or tax rates not in the user’s specific jurisdiction reference data.
6. Transparency. When in doubt, we err toward telling you we used AI rather than hiding it.

Questions or concerns

If you have questions about how AI is used in your Report, or if you believe AI was used in a way that misrepresented your responses or led you to a poor recommendation, please email:

adam@withpurpo.se

We take these concerns seriously and will respond within 3 business days.

Contact

MB AI Konsultacijos
Ulonų g. 1-37, LT-08240 Vilnius, Lithuania
Company code: 306436438
Email: adam@withpurpo.se

This page is intended to be transparent and complete. It is not a substitute for personalized legal advice. For matters specific to your situation or compliance obligations, consult a licensed attorney specializing in EU AI law.