How to Check Your Essay Before You Submit It

What should you check before you submit an essay?

Check five things, in this order: the version history, the similarity report, the AI detector score, the citations and any required disclosure, and the copies you keep afterwards. Together they take about 30 minutes. The point is to make sure the record of how you wrote it exists before the file leaves your hands.

Every menu name below comes from the vendor's own help page as of November 2026. The routine belongs to the authorship record guides on this site, and the other step-by-step pages sit in the how-to section.

1. 1

   Step 1: Name the version history

   Open version history and name the current state, for example Final before submission, with the date.

   Mark step 1 done
2. 2

   Step 2: Run a similarity check

   Use your school's draft-check option, if it has one, and read each match against your citations.

   Mark step 2 done
3. 3

   Step 3: Run an AI detector once

   Check the final draft once, then write down the tool, the score and the date. Do not edit the essay to move the number.

   Mark step 3 done
4. 4

   Step 4: Fix citations and disclose

   Refresh the reference list, match it to the in-text citations, and add any AI disclosure the course rules ask for.

   Mark step 4 done
5. 5

   Step 5: Keep the proof

   Put a PDF of the final, your notes and both reports in one dated folder, and back it up.

   Mark step 5 done

How do you save and name the version history?

In Google Docs, click Last edit at the top right, choose a version, then click More > Name this version. In Word, the file must be stored in OneDrive or SharePoint: click the file title at the top and select Version history. This is the strongest part of your record.

Google's help page says you can add up to 40 named versions per document and that you need edit permission on a file to browse its history. Microsoft's support page says version history only works for files stored in OneDrive or SharePoint, and that a personal Microsoft account can retrieve the last 25 versions. With a school or work account, the number kept depends on how the school configured the library, so do not assume old versions are still there. If you have not set this up yet, the guide to keeping draft version history covers it from the start.

How do you run a plagiarism check on your own draft?

Ask your instructor or library whether students can submit a draft for a similarity check before the final hand-in, then read the report match by match. A similarity report shows which passages overlap with text in the checker's database. It does not say whether those passages are plagiarised, and the percentage alone is not a verdict.

Turnitin's own blog is blunt about this. It says the software compares a submission to a database of internet, academic and student paper content and reports the percentage of the work that is similar, and it states that similarity does not equal plagiarism and that the company leaves that judgment to educators. The same post mentions a drafting product that gives students feedback on similarity and citations. Whether you have access to anything like that depends on your school, so ask.

For each match, check three things: quoted words are in quotation marks, the source is cited next to them, and paraphrases are in your own sentence structure with a citation. The page on how a plagiarism checker differs from an AI detector explains why the two reports answer different questions.

How do you run an AI detector on your own essay and read the score?

Run the detector once on the final draft, record the result, and treat the score as an estimate. A detector does not know who wrote the text. It can label writing that is fully human as machine-written, which the vendors call a false positive, so a number is a reason to keep your record, not a judgment on you.

Paste the draft into an AI detector and note the score and the date. One option is the AI detector from Walter Writes AI; whichever one you use, read the number as an estimate, because no detector is certain. Save a screenshot of the result next to the similarity report. One run on the final draft is enough: repeated runs give you more numbers, not more certainty.

Turnitin's 2023 post on its AI writing indicator defines a false positive as incorrectly identifying fully human-written text as AI-generated. It says the company aimed for a false positive rate under 1 percent, acknowledges that the rate is not zero, and says Turnitin does not make a determination of misconduct. Cornell University's Center for Teaching Innovation goes further: its guidance says detection tools cannot provide evidence for their claim, that research shows continued false positives, and that it does not currently recommend automatic detection for academic integrity cases.

For background, read what an AI detector score measures. If a flag arrives after submission, showing your writing process after an AI flag uses the record you are building here.

How do you fix citations and disclose AI help?

Open the reference list last, after the text is final. Check that every in-text citation has an entry and every entry is cited, refresh the list from your reference manager, and add an AI disclosure if the course rules require one.

If you use Zotero with Word, its documentation says the plugin adds a Zotero tab with Add/Edit Citation, Add/Edit Bibliography and Refresh buttons, and that the bibliography updates automatically as you add or remove citations. The documentation warns that Unlink Citations is irreversible and should usually only be used in a final copy, after you save a backup. Course rules on AI help differ between instructors, so check the syllabus. The guide to disclosing AI writing help has wording you can adapt.

What proof should you keep after you submit?

Keep one dated folder with a PDF of the final essay, the named versions in the original document, your notes and outline, the similarity report and the AI detector result. Do not delete or tidy the original file after hand-in. Then copy the folder somewhere other than the laptop, as described in the guide to backups and data loss.

What this routine does not do

The routine does not prove authorship by itself. It collects dated evidence a reviewer can weigh. A low detector score is not clearance: your school may use a different tool and reach a different number. A high score is not guilt: both Turnitin and Cornell's teaching center say a score needs human judgment and other evidence. None of these checks replaces reading your course's own rules.