AI Usage Disclosure

What we actually use

The image conversion tool uses Tesseract, a long-established open-source optical character recognition engine. Modern Tesseract uses a neural network (an LSTM model) to recognise characters, which is why describing it as AI-powered is fair rather than marketing spin.

On top of that recognition step, the tool applies conventional programmed logic - not a language model - to work out table structure: locating the header row, mapping each word into a column by its position on the page, merging wrapped lines, and removing headers repeated across pages.

The PDF tool reads text and coordinates embedded in the file directly and does not need OCR unless the PDF is a scan.

What we do not do

• We do not send your documents to any third-party AI service or external API.
• We do not use your documents, or anything extracted from them, to train or fine-tune any model.
• We do not use a large language model to interpret, summarise, or infer meaning from your financial data.
• We do not generate or invent values that were not present in your document.

Processing happens on our own server

All recognition and conversion runs locally on the server handling your request. Your document is not transmitted onward to any other provider for processing.

Known limitations of the technology

Character recognition is probabilistic and makes mistakes, particularly on low-resolution photographs, skewed pages, or unusual fonts. Table-structure detection can misplace a value when a document's layout is ambiguous even to a human reader.

We would rather tell you this plainly than imply the output is infallible. Please verify converted figures - see our Disclaimer.