AI models can produce convincing Indigenous-style images and stories in seconds, but are those outputs research data, evidence of cultural harm, or both? In an ABC News investigation published 22 August 2025 (www.abc.net.au/news/2025-08-23/calls-to-protect-indigenous-intellectual-property-from-ai-cultur/105680182), ChatGPT and Midjourney were used to generate 'Dreamtime' artworks and narratives. Experts including Dr Terri Janke and Ngunnawal elder Jude Barlow warn the results homogenise diverse cultures and can appropriate language and sacred stories without consent. This post shows researchers, UX teams and policy analysts how to run a reproducible qualitative analysis of AI cultural appropriation: what to collect, how to code, and how to surface findings stakeholders trust, using tools like www.evidano.com to automate ingestion, custom-language transcription/translation, thematic and cross-segment analysis while protecting sensitive data.
Fast take + source
What happened: ABC asked ChatGPT and Midjourney to create Indigenous-style art and Dreamtime stories; models produced dot-painting imagery and invented Ngunnawal words in under a minute. Experts call this cultural appropriation and potential theft.
- Source: ABC News investigation, published 22 August 2025, www.abc.net.au/news/2025-08-23/calls-to-protect-indigenous-intellectual-property-from-ai-cultur/105680182
- Voices: Dr Terri Janke (legal authority), Jude Barlow (Ngunnawal elder), Wiradyuri artist Birrunga
- Policy context: Productivity Commission discussions (Aug 2025) and prior 2022 report estimating $78M spent by tourists on Indigenous-style art in 2019–20; up to 75% of Indigenous-style art on the market may not be First Nations-made
Findings snapshot
| Date / Item | Metric / Fact | Value / Note | Source |
|---|---|---|---|
| Published | ABC investigation | 22 August 2025 | www.abc.net.au |
| AI test | Time to generate image + story | < 1 minute | ABC (ChatGPT, Midjourney) |
| Market context | Tourist spending on Indigenous-style art (2019–20) | $78 million | Productivity Commission (2022) |
| Market authenticity estimate | Share of Indigenous-style art not made by First Nations people | Up to 75% | Productivity Commission (2022) |
| Models mentioned | Generative systems used in test | ChatGPT, Midjourney | ABC |
How the test worked (plain English)
ABC's team supplied prompts such as “Create an Aboriginal Dreamtime story artwork about Ngunnawal country” to ChatGPT and used a ChatGPT-generated prompt to run Midjourney. The models returned dot-painting style images and a narrative that included reconstructed Ngunnawal words.
Key mechanics: modern generative models pattern-match on large scraped corpora; they can replicate visual tropes and lexical fragments without cultural context, consent, or the living protocols that govern Indigenous expression.
- Observed harms: homogenisation of diverse art styles; use of community language without consent; easy commercialisation via marketplaces (Etsy, Adobe Stock, Temu).
- Expert warnings: Dr Janke (“AI has no Dreaming, it has no kinship, country, or cultural obligations.” Jude Barlow) AI ‘homogenises’ and appropriates reconstructed language.
Implications for researchers, UX teams and policy analysts
For qualitative researchers
Treat AI outputs as a new data type: capture prompt, model version, generation timestamp and marketplace links. Code AI-origin vs community-origin artifacts as separate segments.
Validate linguistic claims with community experts rather than relying on model assertions; document consent and provenance for each item.
For UX & product teams
Do not rely on 'authentic-looking' outputs as substitutes for community co-design. Use rapid qualitative tests to surface whether an output misrepresents or flattens culture before productising it.
Map user journeys that could monetise AI-generated cultural content and run risk assessments (legal, reputational, wellbeing).
For policy and legal teams
Quantify scale and commercial pathways: which marketplaces host AI-generated Indigenous-style art, and what revenue flows result? Use reproducible corpus analysis to inform regulation.
Policy asks: include community consultation requirements, provenance metadata, and penalties for sales that misrepresent community authorship.
For curators & cultural organisations
Prioritise community-led authentication and create public guidance about AI-generated works. Track how often community-specific symbols or language appear in AI outputs to inform education campaigns.
Do more, faster with Evidano
Ingest & preserve provenance
Collect model outputs, prompts, marketplace listings, interview transcripts and community statements into one corpus using www.evidano.com. Store generation metadata (prompt text, model, timestamp, image file) alongside scraped marketplace URLs for traceability.
Custom-language transcription & translation
Use Evidano's transcription/translation with custom dictionaries to accurately capture reconstructed or endangered words (Ngunnawal, Wiradyuri). That prevents automated mistranslation and preserves linguistic nuance for coding and validation.
Thematic, frequency & cross-segment analysis
Run thematic coding across segments (AI-generated vs community-authored) to quantify themes like 'homogenisation', 'sacred content', or 'language misuse'. Evidano computes frequencies, co-occurrence networks, and hierarchical code → subcode structures to surface patterns quickly.
Stakeholder-ready outputs with security
Generate shareable visuals and exportable codebooks for community review. Data is encrypted and Evidano does not use customer data to train third‑party models, important when handling sensitive cultural material.
Checklist: reproducible workflow for analysing AI cultural appropriation
Step-by-step to run in 2–4 weeks:
- 1) Collect: scrape marketplaces, download AI images, save prompts + model metadata, record community responses and expert quotes.
- 2) Prepare: transcribe interviews, add custom dictionary entries for local languages, redact PII where required.
- 3) Code: import a starter codebook (AI-origin, cultural markers, language misuse, commercialisation), run AI-assisted coding and validate with community reviewers.
- 4) Analyse: run thematic, frequency, and cross-segment comparisons (AI vs community), produce co-occurrence maps and quote samplers.
- 5) Report: create an executive brief and a community-facing summary; include provenance tables and recommended actions for product/policy teams.
Conclusion, next moves
AI-generated Indigenous-style art raises urgent research, policy and ethical questions. For teams studying cultural harm or building safeguards, reproducible qualitative analysis turns impressions into evidence you can act on.
Start by assembling a provenance-first corpus and running a segmented thematic analysis; you can do both and scale with www.evidano.com. If you need a pilot: upload a small sample (marketplace listings + a few community interviews) and run a 2‑week analysis to quantify harm vectors and inform policy or product decisions.
Ready to convert contested outputs into defensible research? Visit www.evidano.com to start a secure pilot and build a reproducible workflow that respects communities and documents provenance.
Keep reading
- Commentary on NewsTwo Definitions: Climate Change Acceptance for UndergradsHow a PLoS One Delphi study (Aug 25, 2026) defined climate change acceptance for undergraduate science students, and how AI-enabled qualitative analysis applies it.
- Commentary on NewsResearcher-in-the-loop: AI-enabled UX researchHow the researcher-in-the-loop model governs AI-enabled UX research. Learn practical governance, stats from the August 2026 piece, and how Evidano supports this workflow.
- Commentary on NewsResearcher-in-the-Loop: Governance for AI UX ResearchGovern AI in qualitative UX research with the researcher-in-the-loop model from Jennifer L. Bowie (Aug 25, 2026): practical rules, risks, and tool mappings.