A Bachelor's Thesis with AI: The Complete Honest Workflow from Assignment to Defense

I'm a [field] student writing a bachelor's thesis on [topic]. The
required length is [40–50] standard pages, I have [5] months, and
I'm in [country].

Propose 5 different ways to narrow this topic. For each one, give:
1. A concrete research question (one sentence, not a vague statement)
2. What data would be needed and where to get it (public database,
   my own survey, interviews, content analysis, existing dataset)
3. An estimate of how long data collection would take
4. The main risk — why this variant could fail
5. What the conclusion would look like if the result turned out to
   be "nothing interesting"

The variants must differ in approach, not just in wording. At least
one should be qualitative and at least one quantitative. Skip the
intro and summary, go straight to the five blocks.

My research question is: [question].
Field: [field]. Method I'm considering: [survey / interviews /
data analysis / content analysis]. Sample I can realistically
reach: [description, e.g. 120 students at one faculty].

Do three things:
1. Rephrase the question into a form that is unambiguously
   answerable, and explain what you changed and why.
2. Derive 2–4 testable hypotheses from it. For each one, state
   which variable is being measured, how it's operationalized, and
   what result would disprove it.
3. Be critical: point out where this question is only apparently
   answerable — what data I'd need to answer it honestly, and which
   of that I probably won't be able to get.

Don't hold back on point 3, I want to hear the uncomfortable part
too.

I'm writing a bachelor's thesis on [topic], method [method].
I received the assignment on [date], submission is on [date],
defense is on [date]. I realistically have [10] hours a week for
the thesis, except during [exam period from—to], when I have zero.

Build a week-by-week schedule from today to submission. For each
week:
- what exactly I should do (an activity, not a phase)
- what should be done by the end of the week (a checkable output)

Separately list 5 milestones I can't miss, and for each one write
what happens if it slips by two weeks.
Leave the last 3 weeks before submission free as slack for
proofreading, typesetting, and unexpected problems — don't schedule
any writing there.
Output as a table: week, date, activity, output.

Tomorrow I have my first meeting with my thesis advisor about my
bachelor's thesis. Topic: [topic]. My current research question:
[question]. Method I'm considering: [method]. What I have so far:
[status].

Prepare briefing notes for the meeting:
1. A 5-sentence summary of my plan I can read aloud at the start
2. 8 questions for my advisor, ranked by importance — ones where
   their answer will change what I do next (not questions I could
   answer myself by looking them up)
3. 3 decisions I need them to sign off on
4. What I should write down from the meeting so I don't have to
   solve it again in a month

Be concise, it should fit on one page.

Map the academic literature on [topic], focus [narrowing]. I'm
interested in the period [last 10 years] and the context
[country / EU / global].

I want:
- the main schools of thought and key authors, who's cited in this
  field
- 15 concrete sources (journal articles, monographs, research
  reports) with a full citation and a link to where the text can be
  found
- one sentence per source on why it's relevant to my topic
- which ones are freely available and which are only behind paid
  databases
- 5 search terms I can use to keep searching

Cite a source for every claim. If you're not sure a source exists,
say so instead of guessing and filling it in.

Take the source [file name] and produce a structured summary:
- the author's main thesis in 5 sentences
- method and sample used (if it's a research study)
- 5 most important findings, with a page reference for each
- 3 directly quotable passages in exact wording, with page numbers
- what's missing or weak in the text

Base this exclusively on this source. Where a page number can't be
found, say so instead of guessing.

Go through all the uploaded sources and build a literature map for
the question [research question]:
1. What main positions appear in the sources — who holds which one
2. Where the authors agree (across sources)
3. Where they directly disagree — for each dispute, give both sides
   and the source each comes from
4. Which sources are original research and which just draw on
   others' work
5. Chronology: how the view on this topic has evolved over time

Don't add anything that isn't in the uploaded sources.

My research question is [question] and I want to study it using
[method] on the sample [sample].

Based on the uploaded sources, do a gap analysis:
- what's already established and well covered on my topic
- where the findings are contradictory or insufficient
- what, specifically, none of the uploaded sources address
- which of these gaps my thesis can realistically fill, and which
  is too big for a bachelor's thesis

For each gap, cite the sources that lead you to conclude it exists.

List how the term [term] is defined across the uploaded sources.
For each source: the exact wording of the definition, author, year,
page. Then compare: where the definitions differ, which ones are
mutually incompatible, and which is most widely used in the field.
Finally, suggest which definition I should adopt in my thesis and
how to justify it.

Do the same for the terms: [term 2], [term 3].

Here's a raw list of sources I've noted down in different formats —
some copied from the web, some transcribed from a book:

[paste list]

Convert them into a single consistent format following [my required
citation style, e.g. APA / Chicago / ISO 690], sort alphabetically
by author's last name, and split into: monographs, journal articles,
theses and dissertations, web sources.

Where a field needed for the citation is missing from what I gave
you (year, publisher, pages, DOI), do NOT fill it in — write
[MISSING: year] instead, and add a list at the end of everything I
need to track down.

Here are summaries of 15 sources I have for my bachelor's thesis:

[paste summaries — for each: author, year, title, 3–5 sentences on
content]

My research question: [question]. Method: [method].

Rank the sources by relevance to my question and split them into
three groups:
A) must-read in full — sources I can't write the thesis without
B) abstract and conclusion are enough — context, one claim, filler
C) skip — why, specifically

For group A, note what I should focus on while reading and which
part of the thesis it'll cover. For group C, justify the exclusion
in one sentence. Finally: is there a type of source you think I'm
missing?

You're an experienced bachelor's thesis advisor in [field]. Below
you'll find summaries of the sources I've studied and my research
question.

Research question: [question]
Required length per assignment: [e.g. 40–50 standard pages]
Type of thesis: [theoretical / empirical with original data
collection / literature review]

Source summaries:
[paste summaries from the literature review phase]

Propose an outline for the thesis:
- chapters and subchapters in a logical sequence,
- 3–5 bullet points per chapter on exactly what it should cover,
- an estimated length in pages for each chapter,
- for each chapter, list which of my sources fit there and why.
At the end, note what's missing from my sources to tighten up the
outline.

thesis/
  sources/     PDFs and scanned articles
  notes/       your notes and summaries from the literature review
  data/        CSV from the survey, exports
  charts/      PNGs generated by the script
  text.md      the thesis itself
  analyze.py   script that computes the numbers

Work within this project's folder.
Go through the sources/ folder and my notes in the notes/ folder.
In chapter 2, in the file text.md, after the paragraph about
[concept] add one paragraph (max 200 words) summarizing how the
authors in my material approach this concept.

Rules:
- draw exclusively on what's in the sources and notes,
- don't invent or add anything from general knowledge,
- cite every claim in the form [Last name year, p. X],
- where your source material is missing something, write a TODO
  comment instead of a claim,
- match my style from the surrounding text, no new headings.

Here's the header of my CSV (first 5 rows, semicolon-separated,
UTF-8 encoding, decimal comma):

[paste 5 rows including column names]

Write me a Python script (pandas, matplotlib) that:
1. loads this file as data/survey.csv,
2. cleans the data: [drop incomplete responses, standardize how
   regions are written, flag ages outside the 15-99 range as
   missing],
3. computes descriptive statistics (n, mean, median, SD) for the
   variable [variable] broken down by [group],
4. tests the hypothesis [hypothesis] with an appropriate statistical
   test, and writes a comment explaining why that particular test is
   appropriate and what assumptions the data must meet,
5. saves the numeric results to results.csv,
6. saves the charts to the charts/ folder as PNGs at 300 dpi.

For each step, write a comment explaining WHY it's done.
The script must run even when there are empty cells in the data.

Loaded 812 rows, 41 columns.
Removed 47 incomplete responses, 765 remain.

Descriptive statistics: time_online by age_group
                    n    mean  median      sd
age_group
18-24             213    4.82    4.50    1.91
25-34             241    3.47    3.20    1.64
35-44             186    2.91    2.75    1.38
45+               125    2.14    2.00    1.22

Kruskal-Wallis: H = 118.42, p < 0.001
Results saved: results.csv
Chart saved: charts/fig1.png (300 dpi)

Explain this script to me line by line, as if I were a first-year
student who's never seen Python before.

For each block, write:
- what it does, in plain English,
- why this step is necessary,
- what would happen if I left it out.
For every statistical test, also explain why this particular test
was chosen, what assumptions it makes, and what I'd have to replace
it with if the data didn't meet those assumptions.
At the end, give me 5 questions a reviewer might ask me about this
analysis, and how to answer them.

[paste script here]

Update the charting part of the script so all charts share a single
style suitable for a printed bachelor's thesis:
- sans-serif font, labels at least 11 pt,
- axis labels in [my language] including units, e.g. "Time online
  (hours/day)",
- chart titles in [my language], numbered as "Figure 1: ...",
- no unnecessary color: shades of gray, color only where it
  distinguishes groups, and still legible when printed in black and
  white,
- no background grid, no 3D effects,
- add n (sample size) to each chart's caption,
- export to charts/ as PNG, 300 dpi, white background.
Put all the settings in one place at the top of the script.

Here are my notes for chapter 2 [chapter title], including a
citation with every note. Assemble them into a continuous draft
covering these topics in this order: [topic A], [topic B],
[topic C].

Rules:
- use ONLY claims from the notes; don't add anything from your own
  general knowledge, even if it's common knowledge
- keep the citation for every claim in the exact form it appears in
  the notes
- where the notes aren't enough for a smooth transition, or where
  support is missing, write TODO: [what to look up] on its own
  line — don't invent filler
- don't reconcile conflicting notes, flag the contradiction instead
- academic [my language], no superlatives

At the end, list all the TODOs.

Write the methodology chapter based on this outline of my approach:

Design: [e.g. survey study, quantitative, cross-sectional]
Sample: [who, how many, how selected, where and when collected]
Instrument: [survey with X questions, scales, where the items were
adapted from]
Data collection: [how distributed, response rate, excluded
responses and why]
Processing: [software, tests, significance level]
Ethics: [informed consent, anonymization]
Limitations: [what I know is a weakness]

Requirements: academic style, past tense, descriptive, no
superlatives and no evaluation of the quality of my own approach.
Describe every decision so that someone else could replicate it.
Where the outline is missing something, write TODO: [what to add].

Here are the results of my analysis from [results.csv] and the
chart descriptions [file names]. Hypothesis H1 was: [wording of H1].

Don't write chapter text. Answer in three blocks:
1) What the data literally say about H1 — no interpretive
   overreach.
2) What alternative explanations exist (confounding variables,
   sampling bias, sample size, chance)?
3) What I CANNOT claim from this data, even if it sounds tempting —
   especially anywhere it would mean mistaking correlation for
   causation, or generalizing beyond the sample.

For each point, state which number or chart you're basing it on.

Go through this text and flag passages that sound like generic AI
prose:
- filler and padding ("in today's world", "plays a key role")
- unsupported superlatives ("crucial", "groundbreaking")
- empty summary sentences that add nothing
- strings of three synonyms instead of one precise word
- throat-clearing openers at the start of paragraphs

Just FLAG them, and for each one briefly explain why it's
suspicious. Don't rewrite anything and don't suggest replacements —
I'll rewrite it myself.

Proofread this text as an academic-style editor in [field]. Don't
rewrite it, just list your findings with a quote of the affected
passage:

1. Person and voice: am I consistently using [first person /
   editorial "we" / passive voice]? Flag places where it breaks.
2. Terminology: terms I refer to with different words (sometimes
   "respondent," sometimes "participant") — suggest which term to
   standardize on.
3. Tense: theory in present tense, my own procedure in past tense —
   flag deviations.
4. Abbreviations: is each one defined on first use?
5. Colloquial or journalistic phrasing.

Here's the text of my thesis [filename.md] and the reference list
[references.md]. Do a cross-check and return four lists:

1. Citations that appear in the text but are missing from the
   reference list.
2. Entries in the reference list that the text never cites.
3. Mismatches in the details of the same source (different year,
   different last name, different author order between the text and
   the list).
4. Direct quotes in quotation marks that are missing a page number.

For each finding, give the chapter and passage. Don't fix anything.

Format this list of sources according to [my required citation
style, e.g. APA / Chicago / ISO 690], style [numeric reference /
author-date] per the conventions of [my university].

The sources are of different types: monograph, book chapter,
journal article, thesis, web page, legal statute.

- don't fill in details that aren't in what I gave you; mark missing
  fields as [MISSING: year] etc.
- don't change titles, names, or years
- sort alphabetically by first author's last name
- append a list of entries with missing fields at the end

Go through this chapter and list every checkable factual claim in a
table: numbers, proportions, years, names, references to other
people's research.

Columns: claim | where in the text | cited source | type
(number/date/name) | how hard it is to verify.

Don't claim whether it's true — just list what needs to be verified
against the source.

Convert this chapter from markdown to LaTeX for our faculty's
template. I'm attaching the preamble and a sample chapter from the
template — follow its conventions, don't add new packages.

- headings to section/subsection by level
- images into a figure environment with \label and \caption, leave
  the captions as they are
- tables into whatever environment the template uses
- convert citations to \cite with keys from the attached .bib file
- properly escape special characters (%, &, _, quotation marks)
- where you're not sure about the mapping, leave a % TODO comment
  and the original text

Here's the structure of my thesis (table of contents + first
paragraph of each section) and here are our faculty's requirements:

[paste the exact wording of the policy: required sections, length,
order, appendices, how AI use must be disclosed, abstract format,
number of keywords]

Go through the requirements one by one and mark each: MET / NOT MET
/ CANNOT VERIFY FROM STRUCTURE. For anything not met, say what's
missing and where it belongs. At the end, list what I still need to
check by hand in the typeset version.

You're a strict reviewer of bachelor's theses in [field]. Read the
attached thesis and write the report you would actually submit —
skip the polite preamble and any praise that doesn't say anything
concrete.

1. Three weakest points in the methodology (what specifically, and
   why it matters).
2. Three holes in the argument — claims that don't follow from what
   precedes them, or that aren't backed by data or a citation.
3. Five questions I would ask at the defense, starting with the
   nastiest.
4. For each question, describe what a good answer would look like —
   and what answer you'd consider evasive.

Base this only on the text of the thesis, don't guess at what I
probably did.

Prepare me for the defense. From the attached thesis, generate:
- 15 questions the committee might ask, split into: methodology,
  interpretation of results, positioning within the literature,
  practical implications, "why didn't you do it differently"
- for each one, a skeleton of a good answer in three points
- three questions my thesis doesn't have a good answer to, and a
  suggestion for how to admit that honestly

Ask about my numbers and my decisions, not the field in general.