03 · What You Need to Know
What AI Can Check, What It May Miss, and How to Verify Compliance
What Do Journal Author Guidelines Usually Cover?
Author guidelines, sometimes called instructions for authors or submission guidelines, describe the requirements that researchers must follow when preparing and submitting manuscripts.
They may specify acceptable article types, manuscript structure, word limits, abstract formats, reference styles, artwork requirements, reporting standards, declarations, and submission procedures.
Some requirements concern the manuscript itself. Others concern separate files or information entered into the journal's online submission system.
Elsevier's author support documentation emphasizes that researchers should begin with the target journal's Guide for Authors because individual journals have their own requirements. Springer Nature likewise directs authors to journal-specific submission guidelines rather than treating publisher-wide information as sufficient. Elsevier's manuscript preparation guidance makes this distinction explicit.
Which Requirements Can AI Help Check?
AI can compare explicit instructions with the information available in a manuscript and associated documents. Its strongest applications involve identifying potential discrepancies and explaining what the author should inspect.
| Requirement |
Possible AI Assistance |
Necessary Verification |
| Article type |
Compare the manuscript's stated category with supplied article-type descriptions |
Confirm eligibility and the correct category with official journal instructions |
| Word limits |
Identify relevant sections and flag apparent excess length |
Count words using a reliable tool and the journal's counting rules |
| Abstract structure |
Check for required headings and content categories |
Verify wording, word count, and scientific accuracy |
| Manuscript sections |
Identify potentially missing headings or declarations |
Confirm whether each section is required for the article type |
| References |
Flag apparent formatting inconsistencies or incomplete entries |
Verify bibliographic details, citation correspondence, and journal style |
| Tables and figures |
Review captions, numbering, and textual references |
Inspect actual image files, resolution, dimensions, and accessibility |
| Ethics and disclosures |
Identify whether relevant statements appear to be present |
Verify the underlying approvals, facts, and required wording |
| Supplementary materials |
Compare supplied file lists with stated requirements |
Confirm files are complete, correctly named, and uploadable |
| Submission metadata |
Check consistency across supplied titles, author details, and abstracts |
Compare with actual fields in the submission portal |
This distinction is important. AI may correctly identify that a figure caption is missing while being unable to establish whether the corresponding image satisfies the journal's technical specifications.
A useful AI-generated report should therefore separate verified observations from suspected problems and requirements that could not be assessed.
Why Should You Provide the Official Guidelines Instead of Asking AI to Recall Them?
Journal instructions can change. Requirements for article length, graphical abstracts, reporting checklists, declarations, and acceptable file formats may be revised without the journal changing its title or publisher.
An AI model may recall an earlier version of the instructions or combine requirements from different journals.
For this reason, outdated AI-generated author guidelines can create avoidable submission problems.
The safer approach is to retrieve the current instructions from the journal's official website and provide the relevant material to the AI system.
Include the journal name, article type, source URL, and date of access. If the guidelines refer to separate publisher policies or reporting standards, those materials may also be necessary.
Even with official instructions supplied, AI may misunderstand them. Access to the correct document reduces one source of error but does not eliminate the need for verification.
How Should AI Handle Requirements That Depend on Article Type?
Many journals publish several manuscript categories with different requirements.
An original research article may require a structured abstract, while a commentary or perspective may use an unstructured summary or no abstract at all. Word limits, reference limits, and supplementary-material requirements may also differ.
AI should therefore establish which article type is being assessed before applying a checklist.
When the intended category is unclear, the system should identify the ambiguity rather than automatically select the category with the most familiar structure.
A manuscript that satisfies the requirements for one article type may still be unsuitable for another.
Can AI Reliably Check Word Counts and Character Limits?
AI can identify sections that appear excessively long and help locate material that may be condensed. However, language models should not be treated as authoritative counting tools.
Exact word and character counts are better obtained using a word processor, script, or another deterministic counting method.
The counting rules also matter. A journal may exclude references, figure legends, tables, or supplementary materials from particular limits, while another may include them.
Before reporting a violation, establish what the journal actually counts and apply that rule consistently.
If the instructions are ambiguous, the AI should flag the uncertainty rather than declare the manuscript compliant.
Can AI Check Reference Formatting and Citation Accuracy?
AI can help identify apparent inconsistencies in reference punctuation, author-name formatting, journal abbreviations, and citation style.
It may also flag references that appear incomplete or in-text citations without obvious corresponding bibliography entries.
However, a correctly formatted reference is not necessarily a real or accurate reference.
AI may overlook incorrect titles, fabricated publication details, or mismatched digital object identifiers. Verification should use authoritative bibliographic records, publisher pages, or other reliable sources.
Reference-management software and structured bibliographic validation may be more dependable than a language model for mechanical checks.
Authors should also determine whether the journal requires strict reference formatting at initial submission. Some journals allow flexible initial formatting and apply their preferred style later in the publication process.
Can AI Check Figures, Tables, and Supplementary Files?
AI can inspect supplied figure captions, table titles, numbering, and references within the manuscript. With suitable document and image-processing capabilities, it may also identify some visible formatting problems.
Nevertheless, a text-based comparison cannot establish whether an image meets requirements concerning resolution, color mode, dimensions, embedded fonts, or acceptable file format.
Those properties require inspection of the actual files, preferably using tools that can read their technical metadata.
Similarly, AI cannot confirm that supplementary files are complete or accessible merely because the manuscript refers to them.
A comprehensive submission check should include both manuscript content and the actual files intended for upload.
Can AI Verify Ethical Approval and Required Declarations?
AI may identify whether a manuscript contains statements about ethical approval, informed consent, competing interests, funding, author contributions, or data availability.
But finding a declaration is not the same as verifying its truth.
For example, AI may confirm that a manuscript contains an ethics approval number. It cannot establish that the number corresponds to a genuine approval covering the reported research unless the relevant records are independently checked.
Similarly, the presence of a data availability statement does not prove that the dataset is accessible under the stated conditions.
Authors should distinguish document-completeness checks from factual verification of the underlying declarations.
Watch Out
Never ask AI to fill missing ethics approvals, conflict-of-interest declarations, author confirmations, or data-sharing arrangements with plausible standard wording. If the required facts have not been established, the correct result of the check is "verification required," not an invented declaration.
Can AI Check Reporting Guidelines Such as CONSORT, PRISMA, or STROBE?
AI can help compare a manuscript with a reporting checklist when the relevant guideline and its requirements are supplied.
For example, it may identify whether a randomized trial manuscript describes allocation procedures or whether a systematic review reports a search strategy.
However, the presence of a relevant phrase does not establish that the reporting item has been adequately addressed.
Reporting guidelines also have different scopes. CONSORT concerns randomized trials, PRISMA addresses systematic reviews and related evidence syntheses, and STROBE concerns observational studies.
Authors should use the appropriate guideline and current extension where applicable. A generic AI-generated checklist may omit important design-specific requirements.
AI can assist with identifying possible reporting gaps, but methodological adequacy and accurate checklist completion require informed researcher review.
Can AI Detect Every Missing Submission Requirement?
No. An AI system can only assess requirements that it has correctly identified and materials it can meaningfully inspect.
Some requirements may appear only in the submission portal, conditional questions, publisher-wide policies, or article-type-specific instructions.
Others may concern administrative details such as author verification, reviewer suggestions, permissions, or separate declaration forms.
These limitations explain why identifying every submission requirement is a different task from comparing a manuscript with an already established checklist.
AI should not claim complete compliance when the full set of requirements has not been verified.
What Should a Useful AI Compliance Report Look Like?
A compliance report should identify the requirement, its official source, the relevant manuscript evidence, and the action required.
It should also distinguish between confirmed discrepancies and matters requiring further investigation.
| Assessment Status |
Meaning |
Appropriate Action |
| Appears compliant |
Available evidence matches the supplied requirement |
Confirm the finding during final review |
| Possible discrepancy |
The manuscript may not satisfy the requirement |
Inspect the cited manuscript section and guideline |
| Missing information |
Required content or evidence was not found |
Determine whether the material is genuinely absent |
| Cannot verify |
The required evidence or technical capability is unavailable |
Use an appropriate tool or obtain authoritative confirmation |
| Not applicable |
The requirement does not apply to the selected article type or study |
Confirm the basis for excluding it |
A report organized in this way is more useful than a single overall compliance score. It makes uncertainty visible and gives the author a concrete basis for correcting the manuscript.
What About Journal Policies on AI-Assisted Manuscript Preparation?
AI use during manuscript preparation may itself be subject to publisher or journal requirements.
Springer Nature's current AI manuscript-preparation guidance distinguishes lower-risk language and presentation assistance from uses that shape scholarly content and from uses that replace accountable human research contributions. Its policy emphasizes verification, transparency, and author responsibility. The official policy should be consulted for its detailed requirements.
Authors should also consider whether uploading an unpublished manuscript to an external AI service is permitted under applicable confidentiality, intellectual-property, institutional, and publisher requirements.
A compliance check should not create a new confidentiality or research-integrity problem while attempting to solve formatting issues.
