Responsible-use guide
Checking PDFs, essays and dissertations for AI signals
Long academic documents need careful preparation. Check extraction, word limits, references and section context before interpreting an AI or source-matching result.
Start with the current document limits
Genutext's public preview accepts pasted English text from 300 characters to 350 words. Paid account scans accept pasted text or a readable PDF and require at least 300 characters.
| Option | Text limit | Document support |
|---|---|---|
| Free preview | Up to 350 words | Pasted text; AI-only document score |
| Standard paid credit | Up to 3,000 words | Pasted text or a readable PDF under 5 MB |
| Extended paid credit | Up to 8,000 words | Pasted text or a readable PDF under 5 MB |
Prepare and verify a PDF before scanning
The paid uploader extracts text from a PDF inside the browser, sanitises control characters and collapses whitespace. It does not perform optical character recognition on page images.
- 01
Confirm that the PDF contains selectable text
Try selecting and copying a paragraph. If the page is only an image, use an approved OCR workflow first and review the converted text for errors. - 02
Keep the file under 5 MB
The current uploader rejects larger files. Export a text-focused copy or paste the relevant prose rather than reducing readability with aggressive image compression. - 03
Check the extracted reading order
Columns, running headers, footnotes, page numbers and text boxes can be joined in the wrong sequence. Compare the extracted text with the original document. - 04
Remove irrelevant material when appropriate
A bibliography, contents page or repeated template can dominate the sample without helping answer the authorship question. Preserve the original file separately.
Choose sections that answer a defined review question
A dissertation is not one uniform genre. An abstract, methods chapter, literature review and reflective discussion can have very different language. Combining them into one score can hide that variation.
When to scan the complete document
Use the complete document when it fits the selected limit, the extraction is clean and the overall distribution is relevant. Keep quotations and references in mind when reading passage and source results.
When to scan a selected section
A focused section can be more interpretable when the whole document is too long or includes materially different text types. Record the exact section and do not describe its result as a score for the entire work.
What often needs separate treatment
- Reference lists, bibliographies and appendices dominated by source text
- Tables, equations, code and data outputs
- Direct quotations and supplied assessment wording
- Standard methods or ethics statements with constrained language
- Collaboratively authored or previously published sections
- Translated sections or writing produced with approved accessibility support
Interpret variation across a long document
Paid scans can show sentence-level AI signals where the provider returns them. Use local variation to generate questions, not to assign authorship sentence by sentence.
| Pattern | Possible question | Context to inspect |
|---|---|---|
| Uniformly low signals | Does other evidence raise a policy concern? | A low score does not prove a wholly human process |
| Uniformly high signals | Is the genre, translation or extraction unusually regular? | Check the complete report, sample suitability and process evidence |
| One high-signal section | Was that section drafted or revised differently? | Compare sources, drafts, voice, task demands and permitted assistance |
| Abrupt sentence variation | Is there mixed authorship, quotation or broken extraction? | Open the original pages and inspect the passage boundaries |
| High source match | Is the wording properly attributed? | Open each confirmed source; do not infer plagiarism from the total |
A document review checklist
- Define the authorship, disclosure or source-use question before scanning
- Confirm permission to process the document and remove unnecessary personal identifiers
- Choose a credit that covers the intended prose, or record the selected section precisely
- Verify PDF extraction against the original and note any omitted material
- Run AI only or AI + Plagiarism according to the review question
- Read sentence signals and source matches in their chapter and disciplinary context
- Preserve drafts, sources and the writer's explanation as separate evidence
- Describe the detector result accurately and record the human decision separately
For current prices and inclusions, see scan-credit pricing. For score conversion and storage boundaries, read the methodology and limitations.