Our Methodology

Last updated: February 2026

We believe you deserve to know exactly how our reports are created, what they can and can't do, and when you should seek additional help. This page explains our approach in plain language.

1. How Our Reports Are Generated

WorkGlobal reports are generated using large language models — specifically, Anthropic's Claude. When you use our tool, you provide key details about your work setup: your country of residence, where you work or earn income, your work type, payment method, and income range. These details are sent to the AI, which generates a structured analysis based on its training data about international tax law, bilateral tax treaties, and compliance requirements.

The AI does not search the internet in real time or query government databases. Instead, it draws on broad knowledge of tax frameworks, treaty networks, and filing obligations that was included in its training data. The output is then formatted into the structured report you receive.

2. What the AI Knows (and Doesn't)

The AI has broad knowledge of tax treaties, filing requirements, residency rules, and compliance obligations for many country pairs around the world. For well-known corridors — such as US-UK, US-Canada, and US-India — the analysis tends to be more detailed and more accurate, because these jurisdictions are heavily represented in the AI's training data.

However, tax laws change frequently. New treaties are signed, rates are adjusted, filing deadlines shift, and new compliance requirements are introduced. The AI's training data has a cutoff date, which means it may not reflect the very latest changes. For less common country pairs, the AI may have less specific knowledge and may provide more general guidance rather than precise details.

3. Our Guardrails Against Hallucination

A known limitation of large language models is "hallucination" — generating information that sounds plausible but is incorrect. We take this seriously and have built specific guardrails into our prompts to reduce this risk:

No invented specifics. The AI is instructed to never invent specific tax codes, form numbers, treaty article numbers, or filing deadlines it is not confident about. If it is unsure of a specific detail, it will say so rather than fabricate one.

Hedging language when uncertain. When the AI is not fully confident in a piece of information, it uses qualifying language such as "typically", "generally", or "verify with your local tax authority" to signal that independent verification is warranted.

Honest treaty coverage. The AI will explicitly state when a tax treaty may or may not exist between two countries, rather than assuming or guessing. If it cannot confirm a treaty is in force, it will tell you.

No invented statistics. The AI avoids fabricating statistics, percentages, or numerical claims that it cannot support. If a specific rate is uncertain, it will note that you should verify the current rate with your tax authority.

Country-pair specificity. The AI only flags compliance issues that are specific and relevant to your particular country pair, rather than listing generic warnings that may not apply to your situation.

4. Understanding Confidence Ratings

Our paid compliance guides include per-section confidence ratings: HIGH, MEDIUM, and LOW. These ratings help you understand where the AI is most and least confident in its analysis.

HIGH means the AI has strong, well-established knowledge of this area for your specific country pair. The information is drawn from well-represented jurisdictions and common tax frameworks.

MEDIUM means the information is likely correct but should be verified. This may apply to areas where rules are more nuanced, where recent changes may have occurred, or where the AI's training data is less comprehensive.

LOW means the AI has limited knowledge of this specific area for your country pair. You should treat this section as a starting point only and definitely consult a qualified tax professional before acting on it.

It is important to understand that these ratings are the AI's self-assessment of its own certainty. They are not guarantees of accuracy. Even HIGH-confidence sections should be verified for critical decisions.

5. What We Recommend

Regardless of the confidence ratings in your report, we always recommend the following:

Use our reports as a starting point, not a final answer. Our guides are designed to give you a structured overview of your cross-border tax situation and help you ask the right questions — not to replace professional advice.

Verify specific details independently. Always confirm specific tax rates, filing deadlines, and form numbers with your local tax authority or a qualified professional. These details change regularly and vary by jurisdiction.

Consult a professional for complex situations. If your situation involves multi-country structures, high income levels, business entity decisions, or other complexities, a qualified tax professional is essential.

Re-run your analysis periodically. Tax laws and treaty provisions change over time. If your work setup remains the same, consider re-running your analysis annually to capture any updates in the AI's knowledge.

6. Why We're Transparent About This

We believe in honesty over hype. Many AI-powered tools present generated content as authoritative fact without disclosing how it was created or what its limitations are. We take a different approach.

We include confidence ratings in every paid report. We add disclaimers where they matter. And we publish this methodology page because we want you to make informed decisions about how to use our guidance. Cross-border tax compliance is complex and the stakes are real — penalties, double taxation, and missed obligations can have serious financial consequences. Being upfront about what AI can and can't do is the responsible approach, and it is the standard we hold ourselves to.

7. Disclaimer

WorkGlobal provides educational guidance only. Our content is not legal, tax, financial, or professional advice. All reports and analyses are generated by artificial intelligence and may contain errors, omissions, or outdated information. Tax laws are complex, jurisdiction-specific, and subject to frequent change. You should always consult a qualified tax professional or legal advisor before making decisions based on our content. WorkGlobal assumes no liability for actions taken based on our reports.

8. Contact

Questions about our methodology? Email us at support@workglobal.it.com.