How to Check Your Thesis Before Submission: A Researcher’s Final Audit
Before submitting a thesis, audit it in layers: argument, research-question alignment, methods, results, interpretation, references, ethics and technical consistency. Proofreading is only the last layer. The central question is whether the thesis tells one coherent story from the problem and objectives through the methods, findings and conclusion, and whether every claim is supported by the study you actually conducted.
Your margins are correct.
Page numbers align.
References are in APA.
Excellent.
Now:
Does the conclusion answer the research question?
That question can do more damage than the formatting checklist.
A thesis can be perfectly typeset and methodologically broken.
Audit 1: the argument
Read only:
- title;
- abstract;
- final paragraph of introduction;
- research question/objectives;
- conclusion.
Do they describe the same study?
Look for thesis drift.
The introduction may promise:
access to family planning.
The objectives measure:
service readiness.
The Discussion focuses on:
contraceptive uptake.
The conclusion recommends:
improving autonomy.
All related.
Not necessarily aligned.
Your thesis should be able to tell the same core story at four different lengths: title, abstract, chapter summary and full text.
Audit 2: question → objective → methods alignment
Draw:
Problem → Question → Objectives → Design → Data → Analysis → Conclusion
Check every arrow.
Ask:
- Can the design answer the question?
- Does each objective appear in the analysis?
- Did you collect the data needed?
- Does every result answer an objective?
- Does the conclusion stay within the design?
If the answer breaks anywhere, fix the alignment before the punctuation.
Audit 3: methods reproducibility
Could another researcher understand what you did?
Check:
Setting
- where;
- when;
- relevant context.
Population
- eligibility;
- exclusions;
- recruitment.
Sampling
- frame;
- method;
- sample-size justification;
- response/non-response.
Measurement
- variables;
- definitions;
- instruments;
- validity/reliability where relevant.
Data collection
- who collected;
- procedures;
- timing;
- quality control.
Analysis
- outcome/estimand;
- models/tests;
- adjustment;
- clustering/design;
- missing data;
- qualitative analytic procedure.
If your supervisor circles:
“How exactly?”
the Methods section is not finished.
Audit 4: results completeness
Make a table:
| Objective | Where answered? | Main estimate/finding | Table/Figure | Complete? |
|---|---|---|---|---|
| Objective 1 | Results 4.1 | 56% prevalence | Table 2 | Yes |
| Objective 2 | Results 4.2 | aPR 1.18 | Table 3 | Yes |
| Objective 3 | , | , | , | No |
Every objective should have an answer.
A result can be:
- uncertain;
- null;
- qualitative;
- incomplete because data failed.
But it should not disappear.
Audit 5: interpretation strength
Search your Discussion for:
- caused;
- led to;
- resulted in;
- proved;
- demonstrated;
- predictor;
- determinant.
Now ask whether the design supports each word.
Cross-sectional association?
“Associated with” may be safer than “caused.”
Uncontrolled pre-post study?
“Symptoms declined after treatment” may be justified.
“The treatment caused the decline” needs a counterfactual argument.
Qualitative study?
“Participants described...” may fit.
“All women believe...” probably does not.
Your Discussion became more causal than your Methods.
This happens often.
Interpretation should not acquire new powers after data collection.
Audit 6: tables and figures
For every table or figure ask:
- Can it stand alone with its title/notes?
- Is the denominator clear?
- Are abbreviations defined?
- Are effect measures labelled?
- Are confidence intervals shown?
- Are units visible?
- Are decimal places sensible?
- Does the narrative contradict the table?
- Is the same result repeated unnecessarily?
Then verify every number appearing in:
- abstract;
- executive summary;
- conclusion;
- tables.
One changed analysis can leave an old number hiding in four places.
Audit 7: reference integrity
Use the citation audit.
For every high-risk citation ask:
- Does it exist?
- Is the metadata correct?
- Is this the correct published version?
- Does the source support my claim?
- Does every in-text citation appear in the reference list?
- Does every listed reference appear in the text where required?
AI-suggested citations get manual verification.
No exceptions.
Audit 8: ethics and research integrity
Check:
- ethics approval body;
- approval/reference number;
- required renewals or amendments;
- participant consent/assent;
- parental permission where relevant;
- confidentiality;
- data storage;
- permissions for secondary data;
- trial registration where applicable;
- conflicts of interest;
- funding;
- AI-use disclosures required by institution/journal.
Do not invent ethics language.
Report what actually occurred and what the approving body required.
Audit 9: AI use
Institutional and journal policies vary.
Current ICMJE guidance expects authors to be transparent about relevant AI use and keeps responsibility for accuracy, integrity and attribution with human authors.
Your audit should therefore ask:
- Did I use AI in drafting/editing?
- Does my institution require disclosure?
- Does the target journal require disclosure?
- Did I upload confidential/unpublished data into a system without appropriate safeguards?
- Did AI generate or suggest references?
- Did I independently verify them?
- Did AI change scientific meaning during editing?
Do not let an AI detector become the ethics committee.
Detector scores can be fallible.
The human responsibility is the real issue.
Audit 10: the technical layer
Only now do the formatting sweep.
Check:
- headings;
- numbering;
- tables;
- figures;
- appendices;
- cross-references;
- abbreviations;
- spelling;
- tense;
- citation style;
- margins;
- page numbers;
- table of contents;
- lists of tables/figures;
- file naming.
Technical polish matters.
It just should not be mistaken for scientific quality.
the 30-minute red-team read
Give yourself one final exercise.
Pretend you are Reviewer 2.
Write the five strongest criticisms of the thesis.
Examples:
“The primary outcome is poorly defined.”
“The cross-sectional design cannot establish the temporal sequence implied.”
“The qualitative sample-size justification is generic.”
“Adjusted variables are not justified.”
“The conclusion overstates the findings.”
Now answer each criticism before submission.
Your job is not to make the thesis criticism-proof.
That document does not exist.
Your job is to remove the criticism you can already see.
Automated pre-submission checking
Check My Thesis currently presents itself as a pre-submission tool for citation verification and AI-writing review, including checks for potentially fabricated references and preprint updates.
That can fit inside Audit 7 and parts of Audit 9.
It does not replace:
- methodological review;
- statistical review;
- argument review;
- supervision;
- human source checking.
Affiliate disclosure: Methods Bench may earn a commission if you purchase Check My Thesis through our affiliate link, at no extra cost to you. We have not yet independently benchmarked the product's detection accuracy.
What do I actually do 48 hours before submission?
48 hours
- alignment audit;
- methods audit;
- results-objective audit;
- reference audit;
- red-team critique.
24 hours
- tables/figures;
- abstract numbers;
- conclusion;
- formatting;
- cross-references.
Final hours
- export PDF;
- open the final file;
- inspect page breaks;
- click internal links if relevant;
- confirm appendices;
- confirm filename;
- submit the file you actually checked.
The last instruction sounds ridiculous until you upload THESIS_FINAL_FINAL2_old.docx.
Open your thesis and answer eight yes/no questions:
- Does the conclusion answer the question?
- Does every objective appear in Results?
- Could another researcher reproduce the main methods?
- Are important estimates reported with uncertainty?
- Does the Discussion stay within the design?
- Have high-risk references been manually verified?
- Are ethics and AI-use disclosures accurate?
- Do abstract numbers match the final analysis?
Any “no” becomes today's task.
Frequently asked questions
- What should I check first before submitting a thesis?
Start with argument and methodological alignment, not formatting.
- Should I proofread before or after checking methods?
Do the substantive audit first. There is little value polishing a paragraph you later need to rewrite.
- Can software check my whole thesis?
Software can help with references, grammar, formatting and some consistency checks. It cannot reliably decide whether your design, analysis and conclusion are scientifically aligned.
- Should I use an AI detector before submission?
A detector can flag text for review, but its output should not be treated as proof of authorship or misconduct.
- How early should I do the final audit?
Ideally before the final formatting day. Major methodological inconsistencies need time to repair.
Download the Methods Bench Final Thesis Audit as a one-page checklist.
Use it before any automated checking tool.
References and further reading
- 1.ICMJE Recommendations. Preparing a Manuscript for Submission and Use of AI in Publishing.
- 2.EQUATOR Network reporting guidelines relevant to study design.
- 3.STROBE, CONSORT, COREQ or other design-appropriate guidance.
- 4.Walters WH, Wilder EI. Citation fabrication and errors generated by ChatGPT. Scientific Reports. 2023.
- 5.Check My Thesis website for current vendor-described product features.
Take it further
Better research, once a week.
Practical methods guidance, research tools and funding opportunities from Methods Bench.
Get practical research notes and new opportunities
Short, useful emails for researchers in Uganda and East Africa. No spam.