How we build your research.
Every number comes from your data. Every citation comes from a live academic database. Every reference is verified before it goes into your document. AI is used in one step only: converting verified, structured material into prose.
Five stages. One uses AI.
Four stages have no AI involvement. The fifth, writing, uses AI to translate verified inputs into prose. It cannot add anything the earlier stages have not already established.
Literature retrieval
Live databases. Not AI memory.
We query live academic databases containing hundreds of millions of peer-reviewed articles. Your literature is built from what those databases return for your specific research question, timestamped to the search date.
If a paper does not exist in a live database, it does not go into your thesis.
Statistical analysis
Your data. Validated software. Exact outputs.
We run your patient data through validated statistical packages. The analysis plan comes from your approved protocol. Every number, confidence interval, and p-value is traceable to a line of code run against your dataset.
No statistic is estimated, approximated, or recalled from memory.
Reference verification
Every DOI resolved before it goes in.
Every reference is submitted to primary academic databases for DOI verification. A citation only appears in your thesis if it resolves to a real, retrievable publication.
We do not include references we cannot independently confirm.
Figure generation
Code output. Not AI-drawn.
Charts, forest plots, CONSORT flow diagrams, and heatmaps are generated by scripts running directly on your statistical output. Every axis label and data point comes from your results, exported at print resolution.
Figures are formatted to the submission standards of your target journal or university.
Writing
Verified inputs in. Academic prose out.
This is where AI is used. It receives the verified literature, statistical output, confirmed reference list, and your approved protocol, and converts those into academic prose. It cannot determine what the data shows. The earlier stages do that.
The AI adds nothing not already established by Stages 1 to 4. It translates structured, verified inputs into well-written chapters.
The constraint that matters
The AI in Stage 5 can only write what Stages 1 to 4 have already established. It has no access to your raw data, no ability to run statistics, and no way to add a citation that has not already been verified by a live database query.
What AI does and does not do
AI is a writing tool in this pipeline. It is not a research tool, an analysis tool, or a source of factual claims.
AI handles
- Writing prose from structured, verified inputs
- Sequencing arguments and transitions between sections
- Paraphrasing published literature already retrieved and cited
- Formatting tables and figure references in academic style
- Editing for clarity, register, and readability
AI does not
- Calculate statistics. Validated software does that.
- Retrieve or generate citations. Live databases do that.
- Create figures. Code scripts running on your results do that.
- Verify references. Primary database APIs do that.
- Access your patient data directly
- Add information from its own training knowledge
Turnitin and AI detection
Turnitin runs two completely separate checks. They measure different things and are often confused.
Similarity score
Checks whether your text matches existing documents. A thesis based on your own patient data and properly cited published sources should produce a score well within the acceptable range. Most South African universities target below 15%.
AI writing score
A separate system that analyses text patterns only. It cannot determine whether underlying facts were independently verified. Turnitin advises that scores below 20% are unreliable, and no institution should take action on this score alone. Polished academic prose can trigger false positives.
The methodology is the integrity argument
A thesis where every statistic came from validated software and every citation was verified by a live database query is epistemically more reliable than one where facts were typed from memory, regardless of what a text-pattern detector reports.
AI use declarations
South African universities increasingly require students to submit an AI Use Declaration with their thesis. This is not a disciplinary process.
Designed for exactly this
University AI declaration forms typically include checkboxes for Literature Search, Data Analysis, Writing Assistance, Editing, and Citation Formatting. One category for every stage in our pipeline. There is nothing to hide and nothing that does not fit.
Not punitive
University AI policies describe the declaration as a means to support reflective and responsible engagement with AI in research. It is a structured conversation with your supervisor, completed early and co-signed, not an investigation.
You can be examined on every finding
Because every result traces to your own patient data and your approved protocol, you can answer questions from examiners about any number, figure, or conclusion in your thesis. The methodology creates that confidence.
We help you prepare it
When you receive your deliverable, we can provide a clear breakdown of which AI categories apply and what was done at each stage. Your supervisor conversation will be straightforward.