Overview
Tools such as Turnitin's AI writing indicator and GPTZero estimate how likely it is that passages were generated or heavily modified by a large language model. They work on statistical patterns in the text, which means predictable, formulaic or heavily edited human writing can resemble model output. Turnitin's own guidance states that the indicator may misidentify human-written text and should not be the sole basis for action against a student, and it marks low scores (1–19%) with an asterisk because false positives are more common in that range.
Peer-reviewed research has also found that several detectors misclassify writing by non-native English writers as AI-generated far more often than writing by native speakers (Liang and colleagues, Patterns, 2023). Vendors have since reported changes to address this, but the underlying point stands: a flag is a reason for a conversation, not a finding.
Our role is to help you respond well. We explain the report in plain terms, help you gather the evidence you already have — version history, drafts, notes, sources, search history — and prepare a clear, truthful account for your instructor or integrity panel. If you did use AI tools, we help you work out how that fits your course's policy and how to disclose it. We do not rewrite, 'humanize' or otherwise alter text to change a detector result, under any circumstances.
Who it’s for and the problems it solves
This is for you if you are…
- Students whose coursework or thesis chapter has been flagged by an AI indicator and who have been asked to explain their work
- Multilingual writers and students who use grammar tools, who worry their natural style will be misread by detectors
- Students who want to set up a documented writing process before submitting, so they can show their work if questioned
- Researchers and authors unsure what their journal's AI-use disclosure policy requires
- Instructors and program leaders who want an independent explanation of what an AI report can and cannot support
Common problems we help with
- You wrote the work yourself and a report says a large share is likely AI-generated
- You used Grammarly, a translation tool or a chatbot for some tasks and are unsure whether that was allowed
- You have been invited to an academic integrity meeting and do not know what evidence to bring
- Your course policy on generative AI is vague, or different modules have different rules
- You want to disclose AI assistance correctly in a thesis or manuscript but do not know what to say
What’s included — and what isn’t
Included
- Plain-language explanation of the specific report you received: which indicator, what the score and highlighted segments mean, and the vendor's stated limitations
- Review of your institution's and course's generative-AI policy and how it applies to what you actually did
- Help assembling writing-process evidence: document version history (Word, Google Docs, Overleaf), dated drafts, outlines, notes, annotated sources, reference-manager records and search history
- Preparation for an instructor meeting or integrity hearing: a factual timeline of how the work was produced and practice explaining your argument and sources
- Guidance on writing an accurate AI-use disclosure or acknowledgment where your policy or journal requires one
- Setting up a documented workflow for future assignments so your process is visible by default
- For instructors: a briefing on detector limitations and fair ways to follow up a flag
Not included
- Rewriting, paraphrasing or 'humanizing' text to change an AI-detection score, or advising on tools that do so
- Creating, backdating or altering drafts, notes or version history — all evidence must be genuine
- Representing you as a legal advocate in disciplinary proceedings; we help you prepare, and your institution's process and appeal routes apply
- Stating that a text was or was not written by a person — no one can determine that from a detector score
What you receive
- Report explanation
- A short written note on what your specific report shows, which parts are within the tool's documented low-reliability range, and what the vendor says about how results should be used.
- Evidence checklist and timeline
- A tailored list of the genuine records you can retrieve (version history, drafts, notes, sources) and a dated timeline template for organizing them.
- Policy summary
- Your institution's and course's rules on generative AI summarized against what you did, with points to clarify with your instructor.
- Meeting preparation session
- A rehearsal in which you explain your research process, sources and key arguments in your own words, as you may be asked to do.
- Disclosure wording guidance
- Where AI assistance was used and permitted, guidance on writing an accurate statement of which tool was used, for what, and how you checked the output.
How it works
Understand the situation
You share the report, the assignment brief, the relevant AI policy and any message you have received. We clarify deadlines and what has been asked of you.
Explain the report
We explain what the indicator measures, how your score relates to the vendor's reliability notes and which segments were highlighted.
Gather genuine evidence
We help you locate and organize your drafts, version history and notes into a clear timeline. We do not create or alter any record.
Prepare your account
You draft a factual explanation in your own words; we give feedback on clarity and completeness and rehearse the conversation with you.
What detectors measure and what counts as evidence
- Turnitin AI writing indicator
- Reports the share of qualifying prose it predicts was AI-generated or AI-paraphrased. Turnitin states results can misidentify human writing, marks 1–19% scores with an asterisk because of higher false-positive rates, and requires a minimum length of long-form prose; language support is limited, so check the current guide.
- GPTZero and similar classifiers
- Give document-level probabilities and sentence highlighting. GPTZero's own site cautions that results should not be used as a final verdict or to punish.
- Known sources of false positives
- Short texts, formulaic genres (lab reports, literature summaries), heavily proofread or grammar-tool-edited prose, and writing by multilingual authors can all raise scores without any AI generation.
- Version history
- Word (OneDrive/SharePoint), Google Docs and Overleaf keep timestamped revisions that show a document growing over time. These are often the strongest evidence of process.
- Process artifacts
- Outlines, handwritten notes, annotated PDFs, reference-manager libraries, library search histories and supervisor feedback on earlier drafts.
- Oral explanation
- Being able to explain your argument, choices and sources in conversation is frequently what instructors rely on most, and it is something you can prepare for.
Reading a flag calmly
A detector score is a model's estimate, and the organizations that build these tools describe them as one input to human judgment. That does not mean a flag can be ignored: many institutions will ask you to explain your process, and some will look at other evidence. It does mean that you are entitled to have the report considered alongside everything else you can show.
We start by checking basics that are often overlooked: whether the document met the tool's length and language requirements, whether the score falls in the range the vendor labels less reliable, which exact passages were highlighted, and whether those passages are formulaic sections (definitions, methods boilerplate) that tend to score higher.
Building a truthful record of how you wrote
The most persuasive evidence already exists on your devices and accounts. Version history in cloud editors records when text appeared; drafts saved as separate files show development; notes, outlines and annotated sources show thinking that preceded the prose. We help you find these, arrange them into a timeline and decide what to present, and we advise writing future assignments in a tool with automatic revision history.
Every item must be genuine and unaltered. Fabricating or backdating drafts would turn a disputed flag into clear misconduct, and we will not assist with it.
- Export or screenshot version history showing dates and incremental changes
- Collect outlines, notes and early drafts with their file dates
- List the sources you used and where you found them
- Note any tools used (grammar checkers, translators, chatbots) and for which tasks
Using AI tools within the rules
Policies range from total prohibition to encouraged use with disclosure, and they can differ between modules at the same university. Publishers generally agree that AI tools cannot be listed as authors and that authors must disclose substantive use and remain responsible for the content; the Committee on Publication Ethics (COPE) sets out this position. We help you map what you did or plan to do against the rules that apply to you and draft an honest acknowledgment where one is needed.
If you are a multilingual writer who relies on language tools, we can help you ask your instructor in advance which kinds of assistance are acceptable. Legitimate language support is also available through English-language editing that respects your institution's rules.
What the work can look like
Illustrative example — fictional data
Sample excerpt: report explanation note
An international master's student's 2,400-word essay received an AI indicator score, with highlighted segments concentrated in the introduction and a paragraph of definitions. All details below are invented to show the format of our note.
What the report shows
The indicator predicts that roughly a third of qualifying prose resembles AI-generated text, concentrated in two sections. The vendor states that results can misidentify human writing and should not be the only basis for action. The definitions paragraph uses standard textbook phrasing, a pattern that can raise scores.
Evidence available
Google Docs version history shows the essay built over nine days with 60+ sessions; an outline file predates the first draft; Zotero shows sources added before writing began. The student used a grammar checker for spelling, which the course policy permits.
Suggested next step
Reply to the instructor offering the version history and outline, and be ready to explain the essay's argument and source choices in a short meeting. The student drafts the reply in their own words.
Illustrative example only. The scenario, score description and evidence are fictional and do not describe a real client or outcome.
What you’ll need to provide
- The AI-detection report or a screenshot of the score and highlighted text
- The assignment brief or submission guidelines and your course's or institution's AI policy
- Access to your own version history, drafts, outlines and notes (you retrieve them; we help organize)
- An honest account of which tools, if any, you used and for what
- Any correspondence from your instructor or integrity office, with dates
You don’t need everything ready to ask for a quote — tell us what you have and we’ll say what’s missing.
What affects price and turnaround
Price depends on
- Length of the flagged document and number of highlighted passages
- How much evidence needs to be located and organized
- Whether meeting or hearing preparation is included
- Urgency, for example a meeting already scheduled
- Whether you want ongoing workflow set-up for future assignments
Turnaround depends on
- How quickly you can retrieve your version history and drafts
- The date of any scheduled meeting or response deadline
- Availability of your institution's written AI policy
Every project is quoted individually. How we price our work.
Questions people ask before starting
Can you prove I did not use AI?
No one can prove authorship from a detector score, in either direction. What we can do is help you present genuine evidence of your process and explain the report's limitations accurately, which is what fair decision-makers look for.
Can you rewrite my text so it is not flagged?
No. We never alter text to change a detector result. Doing so would misrepresent your work and would likely breach your institution's rules. We focus on understanding the report and documenting how the work was actually produced.
I did use ChatGPT for part of the work. Can you still help?
Yes. We help you understand what your policy allows, prepare an honest account of what you used it for, and draft a disclosure if appropriate. We will not help you conceal use that the policy required you to declare.
Does using Grammarly or a translator make detectors flag my writing?
It can. Heavy editing by language tools can make prose more uniform, which some detectors associate with AI text. Whether such tools are permitted depends on your course policy, so it is worth checking and, where needed, disclosing them.
Will you contact my university for me?
No. We help you prepare, and you communicate with your instructor or integrity office. If you have access to a student advocate or union adviser, we recommend involving them for procedural questions.
How can I protect myself before I submit next time?
Write in a tool with automatic version history, keep outlines and notes, record any AI or language-tool use as you go, and ask about the policy early if it is unclear. We can set this workflow up with you in one session.