Recommended Manuscript Structure

AI Pre Review reads your manuscript, splits it into sections by recognizing standard headings, and measures each one. None of the following is required — analysis runs on any document — but following it makes the measured statistics (word counts, figure and table counts, and reference count) exact rather than approximate.

File format

• PDF (.pdf) or Word (.docx) with selectable text works best. If you can highlight the text with your cursor, then our system should have no problems reading it cleanly.
• Plain text (.txt) is fine, but add the section headings below so your sections can be easily and consistently parsed.
• Scanned or photographed PDFs do not work consistently — the page is an image with no text to extract. Test by trying to select a sentence with your mouse; in order to analyze the manuscript, we have to pass it through an Optical Character Recognition (OCR) tool first which introduces an unncesessary layer of variability for document analysis. Since you presumably have the source document in a text-based format already, let us use that to get the best quality analysis.

Section headings

A heading is recognized when it is on its own line and is just the section name (a leading number is fine). Capitalization does not matter, and 1. Introduction or II. Methods both work. Easily recognized sections include:

• Abstract (or Summary)
• Keywords
• Introduction (or Background)
• Methods (or Materials and Methods, Study Design)
• Results (or Findings)
• Discussion
• Conclusion
• References (or Bibliography)
• Acknowledgments, Funding, Conflicts of Interest, Ethics, Data Availability, Author Contributions, Appendix / Supplementary

Avoid burying a section name inside a longer phrase: Methods of Statistical Analysis is read as ordinary text, while Methods is recognized as the heading. For short statement sections, a single line with a colon works too — for example Keywords: cataract, refractive outcomes or Funding: None.

Abstract

The most reliable setup is a line that simply says Abstract above the abstract text. An unlabeled abstract (common in technique or letter formats) can be detected automatically, but adding the heading removes all ambiguity.

Figures and tables

Each figure and table is counted from its caption, which must begin with the label and number at the start of a line — for example Figure 1. or Table 2. (forms like Fig. 1 and Table 3 — are also accepted). Use one caption per item, number them in sequence, and keep caption text as real document text rather than baked into an image. Mentions in the body such as "see Figure 1" are ignored, so they never inflate the count.

References

A numbered list under a References heading parses most accurately:

• Number entries in sequence starting at 1 (1., [1], and 1) are all accepted).
• Use a single-column layout for the references — two-column reference lists are the most common cause of miscounts, because the text is read out of order.
• Unnumbered author-date lists are counted by a less precise year-based estimate, so prefer numbered lists when accuracy matters.
• In Word files, references inserted as endnotes or citation-manager fields may not extract at all (you may see a count of 0). Paste the final list as plain numbered text to avoid this.

What gets measured

After parsing, AI Pre Review reports the abstract word count, the main-text word count (Introduction through Conclusion, excluding abstract, references, acknowledgments, funding, conflicts, appendix, and captions), the figure count, the table count, and the reference count. The main-text count is what to compare against a journal's body word limit.

Need help?

Start with the Getting Started guide, or see the FAQ for common questions. Still stuck? Contact us at aiprereview@gmail.com.