When AI Can Write, Edit, Draw and Publish for You: Can One Person Become a Media Company?

# When AI Can Write, Edit, Draw and Publish for You: Can One Person Become a Media Company?

Author: 邱俐瑜 Liyu Chiu｜公民科 x AI 資訊科技媒體素養教案獎得主、AI 教育學院成員、經營 FB 粉絲專頁「快樂鯉魚秋老虎：運算思維與程式設計」、關注 AI 媒體素養、數位公民與科技教育

Producing media once required a small team for reporting, writing, photography, editing, design, subtitling and distribution. With AI, one person can now complete many of these steps, but the core question is still who gets to decide how a story is told.

Key facts

• 2026 creator-economy research treats AI as a co-creator and distinguishes augmentation from substitution in creative work.
• For Indigenous media, practical benefits include transcription, translation, summarization, editing, subtitles, visuals and publishing support.
• The central issue is narrative agency: who films, asks, edits and titles a story shapes what audiences see.
• Oral traditions may already contain rules about speakers, occasions, relationships and responsibility, offering a useful lens for AI-era media literacy.

Why this matters to ordinary readers

This is not a topic only for specialists. It touches everyday decisions about health, storytelling, food, crops, risk and community responsibility. The goal is not to bury readers in jargon, but to make the underlying question understandable and useful.

Supporting evidence and its limits

A story-writing experiment by Doshi and Hauser offers a concrete caution: access to generative AI story ideas improved creativity and reader evaluations of some outputs, while AI-assisted stories became more similar to one another. The study tested a writing task; it did not demonstrate that one person could run a profitable media company. For Indigenous creators, tools can help with ideas, while local expression, reporting relationships and cultural judgment remain human responsibilities.

It is more practical to separate a one-person media operation into a workflow than to imagine AI as an all-purpose assistant. Before reporting, it can organize public background material and a draft interview guide. Afterwards, it can help create a preliminary transcript and flag names or dates for checking. Before editing and publishing, someone still needs to return to the recording, confirm a contributor’s permission and check whether subtitles have changed the tone. These are different tasks. Tools speed up organization and version comparison; they do not automatically complete verification, consent or accountability.

Greater speed also changes the speed at which errors travel. If one unchecked summary is reused in a post, a short-video caption and a graphic, it can look like several independent sources when all versions actually come from one mistake. Keeping original material, quotations cleared for publication and confirmed versions separate—and retaining a final step that returns to the source—helps a small team see which sentence needs to be checked even under time pressure.

What modern science can contribute

Modern science is strong at measurement, comparison and verification. It can test whether a pattern appears repeatedly across places and times, while also showing where evidence is still weak. Yet research only measures what its design includes, so relationships, norms and lived context may still be missed.

For media work, this kind of accountability can be concrete. Does a study describe participants, platform data or a particular writing task? Is an interview one person’s experience, or evidence for describing an entire community? Stating the type of evidence, its time frame and its limits does not weaken a story. It lets readers tell the difference between a person’s voice, a research finding and a question that still needs to be asked.

Two-Eyed Seeing: two-way knowledge does not make both sides identical

Indigenous traditional knowledge is better described here as accumulated norms, taboos, observations, experiential rules and practices. Two-way knowledge does not force these into the mold of Western science, nor does it ask science to validate culture. It places different ways of forming judgments beside each other so that similarities, differences and unanswered questions become visible.

AI should organize, not judge

AI can index interviews, images, environmental records, public research and historical material. It can help detect patterns in time, place and events. But it should not decide cultural meaning for communities, generalize one person’s account to an entire people, or turn correlation into causation.

Bringing the question back to Indigenous communities in Taiwan

The useful move is not to copy an overseas answer, but to turn it into a local question that can be observed. Do similar patterns appear in different communities? Do elders and younger people notice the same environmental changes? Where do sensor data, public records and lived experience intersect?

For independent media in Indigenous communities in Taiwan, a practical start can be one small, manageable story: an event, a local service or a daily record that is suitable for public sharing. A reporter can first agree with the speaker which names, places, images and words may be public and which should remain within the community. Once complete, the speaker should know the title, extracts and platforms through which the story will appear. This does not turn every report into an elaborate procedure; it returns an existing responsibility of relationship to a fast production process.

Speech-to-text and translation require special care because pauses, Indigenous-language words and terms of address can be turned into sentences that sound fluent but are wrong. Uncertain passages can be flagged for the speaker or someone familiar with the context to check, rather than being completed by the model. That is not a rejection of tools; it keeps AI output in a place where people can correct it and return to the original context.

Preserve how conclusions were formed

If these materials are to be used by future AI systems, the context of each record matters: time, place, speaker, practice, exceptions and source. Preserving this context prevents cultural material from being reduced to decontextualized fragments and helps readers understand how a conclusion was reached.

Moving beyond the international case

The purpose of this feature is not to copy an overseas case into Taiwan and assume it applies unchanged. It is to turn the case into a question that Indigenous readers can understand, discuss and verify locally. 2026 creator-economy research treats AI as a co-creator and distinguishes augmentation from substitution in creative work. For Indigenous media, practical benefits include transcription, translation, summarization, editing, subtitles, visuals and publishing support. The central issue is narrative agency: who films, asks, edits and titles a story shapes what audiences see. Oral traditions may already contain rules about speakers, occasions, relationships and responsibility, offering a useful lens for AI-era media literacy. Future fieldwork or databases should preserve who said what, when, where and under what conditions, so later comparisons remain meaningful.

What this means for Yuan Media AI

Yuan Media AI should not translate every traditional practice into scientific jargon. Its role is to maintain a reading framework that can hold modern research and local experience together. When science provides data, we ask what lived context may have been omitted. When local experience provides a norm or taboo, we ask under what conditions it formed and whether exceptions exist. AI can index and compare; contextual judgment remains human.

Conclusion

AI can help one person work more like a small media organization, but it cannot assume cultural and public responsibilities on that person’s behalf. Mature AI-assisted independent media is not about generating the most content. It means knowing what needs verification, what requires asking people, and what must not be used to represent others merely because doing so is convenient.

Main source

• 《行銷分析期刊》｜生成式 AI 與創作者經濟研究
• Doshi 與 Hauser｜生成式 AI、創意與作品多樣性實驗
• Customer Needs and Solutions, published 2026-07-17; examines AI co-creation in the creator economy, including augmentation/substitution, privacy and disclosure.

How general readers can approach this topic

When reading topics that cross cultures and technologies, begin with four questions. First, is the article describing an observation, an association, or evidence of causation? Second, in what environment and relationships did the local experience develop? Third, does modern technology offer better measurement, more data, or simply faster generation? Fourth, if the case is brought into your own community, which aspects can be compared directly and which require new fieldwork? Reading this way helps avoid two extremes. It prevents traditional experience from being dismissed as superstition, while also preventing a similar finding in modern research from being taken to mean that ancestors already knew every scientific principle. Mature two-way knowledge allows questions to be asked of different methods and leaves room for not yet having an answer.