Paperless-ngx + AI: Automatically tag and classify documents

Since version 3.0 Paperless-ngx can aLanguage model to suggest title, sender, document type and keywords for new documents and to answer questions about documents. The model can running locally via Ollama – documents then do not leave your company. The AI supplements the existing assignment based on rules and learned patterns; human oversight remains important.

How Paperless-ngx currently classifies documents – and where AI helps

Paperless-ngx classifies documents in the traditional way using rules (for example, “contains ‘Stadtwerke’ → sender Stadtwerke”) and via a learning-based method, which derives patterns from already classified documents. This works well, but it requires examples and maintenance. A language model understands the content even without training data: It recognizes that a letter is a reminder from the tax office, even if no similar reminder is yet in the system.

What the AI takes over in Paperless-ngx

• Suggest title: e.g. “Electricity bill 2026 – municipal utilities” instead of “scan_0042.pdf”.
• Recognize sender and assign.
• Determine document type: invoice, contract, notice, quotation …
• Assign keywords:for example cost center, project, or status.
• Document chat:Answer questions such as “When is this invoice due?” directly on the document.

New documents can first land in the inbox, where the suggestions are reviewed and accepted with one click — a good middle ground between automation and control.

Local or cloud?

Paperless-ngx can use a model via the OpenAI interface or locally via Ollama to use. In invoices, contracts and official notices, personal or confidential data is usually included – the local version (or a model service in Germany) is therefore usually the better choice. The hardware required for this is explained by What hardware does your own AI need?.

Community extensions

Even before the built-in AI, extensions from the community emerged, such as paperless-gpt or Paperless-AI, which use language models for classification and text recognition. Since AI is now included in Paperless-ngx itself, it is usually worth first taking a look at the built-in functions.

Best practices

• Clear structure first:A manageable list of senders, document types, and keywords helps both people and AI.
• Have suggestions reviewed:at least in the first few weeks and for tax-relevant documents.
• Combine rules and AI: Clear cases via rules, unclear ones via AI.
• Measure quality: How often are the suggestions correct? Where is improvement needed?
• Document data protection: Which model processes which documents, where does it run?

More than tagging

In combination with an automation such as n8n additional steps can be connected: pass invoice data to accounting, enter deadlines in the calendar, notify responsible persons. An example shows Automatically capture invoices by email.

bettersorted relies on open, transparent building blocks – operated on servers in Germany and bundled in the automaisa Hub. We provide vendor-neutral advice, are a BAFA-registered consultant and authorized INQA coach. The consulting and guided implementation can be handled via the INQA-Coaching with 80% funding; the Funding Check shows the right path. For a non-binding initial consultation: Contact.

Example: An association with changing volunteers

An illustrative scenario: An association receives notices, donation receipts, insurance documents, and invoices. Filing changes with the volunteers — and with it, the system. With AI suggestions in Paperless-ngx, documents are consistently named and tagged in the inbox; the treasurer only has to confirm them. The local model ensures that member data does not leave the association’s server.

How to measure whether AI helps

• Share of suggestions adopted unchanged
• Time from receipt to filing
• Number of misfiled documents in spot checks
• Satisfaction of the people who file documents

Costs and funding

For AI-supported document filing, no license fees apply when using open-source tools. Costs arise for setup (planning, installation, integration, testing), operation (server or hosting in Germany, updates, monitoring, data backup) and support (training, rules, contact persons). We do not quote fixed prices because scope and starting point vary greatly.

Eligible for funding is the consulting and guided implementation: The INQA-Coaching covers 80% of coaching costs nationwide (up to €11,520, vouchers until 30.06.2028); a preliminary analysis is subsidized by the BAFA consulting grant with 80% in the new federal states, Lüneburg and Trier, otherwise 50% – for applications until 31.12.2026.

Frequently asked questions

Does Paperless-ngx have AI?

Yes, since version 3.0 Paperless-ngx can optionally use a language model to suggest titles, senders, document types, and tags, and to answer questions about documents.

Can I run the AI in Paperless-ngx locally?

Yes, via Ollama on your own hardware or on a server in Germany. In that case, the documents do not leave your company.

How reliable is the automatic tagging?

For clear documents, very reliable; for unusual correspondence, less so. An inbox with review of the suggestions is therefore useful.

Do I need a graphics card for Paperless-ngx AI?

For a local model with acceptable speed, a graphics card is recommended. Alternatively, a model service in Germany can be connected.

Status: October 2026. Models, versions and licenses change quickly – before making a decision, check the current license with the provider.