I Am a Local: The 'Fellow Townsman' Fan Page's Backend is an Neihu Company and an AI Instruction That Forgot to Delete

30-Second Overview: You have likely followed at least one local fan page like "I Am a Kaohsiung Person" or "I Am a Taichung Person," sharing night market news and weather, feeling like "us locals." In February 2026, someone caught an uncleaned AI instruction in a post on the "I Am a Taipei Person" fan page. Following this thread back leads to a building on Jihu Road in Neihu, Taipei, and a group of people surnamed Lin running an enterprise network. They use AI to rewrite news from pro-China media into a "fellow townsman" tone, then distribute it in bulk on fan pages disguised as locals from various counties and cities. The most ironic part is that it hardly fabricates news, so the ruler of "verifying truth" cannot measure it.

In February 2026, a fan page calling itself "I Am a Taipei Person" posted what looked like an ordinary local image-text post. The real problem was a line of text remaining below the post: "Remove sensitive words, strengthen local Taiwanese colloquialisms" ¹.

This is an instruction written for AI. It should have been deleted after generation, just as an author erases pencil marks on a draft before submitting. But this time, someone slipped up and posted the instruction along with the finished product. The fan page "Kaohsiung Lives Well" (@takaogoodday) captured this image on February 19 and made a summary; two days later, Liberty Times, Newtalk, and Formosa TV followed up on the same day ². A forgotten line of text exposed the entire backend operation.

Kaohsiung Lives Well's February 2026 exposure summary: Lists the operating structure of LIFE Life Network, content mostly from CTWANT, the residual AI instruction "Remove sensitive words, strengthen local Taiwanese colloquialisms," and marks the office location on Jihu Road in Neihu. (Source: Kaohsiung Lives Well Threads, reported and reviewed citations)

📝 Curator's Note
Pay attention to what this instruction says. "Remove sensitive words" assumes the original draft contains words that Taiwanese readers should not see. "Strengthen local Taiwanese colloquialisms" assumes the finished product should pretend to be written by Taiwanese. In other words, this is not a fan page editor polishing text; it is a production line doing two things: erasing the origin label, then attaching a "local" label. The entire technique usually hides behind the finished product; this time, the manufacturing process leaked the filling itself.

The Backend is on Jihu Road 10 Alley, Neihu

Following the clue from "Kaohsiung Lives Well" to check company registrations, the backends of these county and city fan pages converge to the same address: 5th Floor, No. 46, Jihu Road 10 Alley, Neihu District, Taipei City ¹.

Under this address operates a content business outwardly using the names "LIFE Life Network" (life.tw) and "US-based All Access Group." However, breaking down the commercial registration reveals that it hangs one signboard outwardly, but behind it is a group of companies. Jishuo Technology Co., Ltd., Unified Business Number 89977832, established in 1995, Chairman Lin Xianming; Bosite Technology Marketing Co., Ltd., established in 2005, in charge of Lin Xiuzi; Nianjie Technology Co., Ltd., established in 2011, in charge of Xu Yunting ³. The three persons in charge are different, but they rely on Lin Xianming, Zhang Yujia, Lin Xiuzi, and Lin Zhaofeng, four individuals, rotating names on each other's board of directors and supervisors list, plus the same address, weaving a cross-shareholding group of the Lin family. It is not a single person named Lin Xianming as a figurehead; it is a family net.

Jishuo Technology (Unified Business Number 89977832) commercial registration: Representative Lin Xianming, registered address 5th Floor, No. 46, Jihu Road 10 Alley, Neihu, board of directors and supervisors list consists of Lin Xianming, Zhang Yujia, Lin Zhaofeng, and Lin Xiuzi, four Lin family members rotating names. (Source: opengovtw, data from Ministry of Economic Affairs Commercial and Industrial Registration Public Disclosure)

The capability combination of this net is more noteworthy than its low profile. Bosite hangs the brand name "Online Polling Network POLLSTER" outwardly, specializing in public opinion polling ⁴. That is to say, this group produces "public opinion data" with one hand. The group's recruitment page simultaneously admits it is "the only unit in the country simultaneously operating social media websites, professional market research, news portal websites, blog broadcasting networks, SEO website optimization, and Facebook platform marketing" ¹, holding nearly a hundred fan groups as "distribution channels" in the other hand. Nianjie Technology's recruitment post writes even more directly: "Use AI text, AI images, AI video tools to produce interesting content, assisting over 100 fan pages in posting and material creation" ¹. Producing data, mass-producing content, and bulk distribution: three things in the hands of the same people at the same address.

As for the signboard "US-based All Access Group," no company named "All Access" can be found registered in Taiwan, and the "ALL ACCESS HOLDING GROUP LTD." hanging at the website footer cannot find a corresponding entity in US state registration records ¹. "US-based" is more like a sticker pasted on the door, essentially the same thing as the "local" characters on those county and city fan pages.

LIFE Life Network service page under Jishuo Technology, platform advantages clearly list "Million Monthly Traffic," "Precise Targeted Marketing," and "Native Advertising Placement, Natural and Not Obtrusive"—meaning you can stuff content into your feed for the sponsor. (Source: aams.tw)

✦ It produces public opinion with one hand making polls, mass-produces content with another hand using AI, and the third hand hangs the content on nearly a hundred fan pages disguised as fellow townsmen for bulk distribution.

CTWANT News, Dressed in a Fellow Townsman's Tone

When you click on the "news" reposted by "I Am a Kaohsiung Person," you think you are reading local Kaohsiung media. In fact, most of this content comes from the CTWANT Group and CTWANT ¹.

CTWANT's issuing unit is Wangdao Wangtai Media Co., Ltd., belonging to the Wang Wan Zhongshi Media Group along with CTWANT Television ⁵. This group's political color is not a secret. CTWANT News Channel was found in violation 25 times by the NCC between 2014 and 2020, with cumulative fines exceeding 11.53 million NTD, the highest fine amount among TV stations; in November 2020, seven NCC commissioners unanimously resolved not to renew the license, and CTWANT stopped broadcasting in December, switching to YouTube and the internet ⁶. Earlier, a 2019 report by the UK's Financial Times quoted sources from Wang Zhong's reporters, stating that the Chinese State Council Taiwan Affairs Office once made phone calls to issue editorial instructions, requesting support for specific candidates ⁵.

📝 Curator's Note
There is a clever time lag hidden here. When CTWANT was a TV station, it was regulated by the NCC; violations would result in fines, and eventually, the license was even revoked. But when its content is rewritten by LIFE into "local image-text" using AI and distributed on "I Am a Kaohsiung Person," readers see fellow townsmen, not CTWANT; the source is washed away. After CTWANT went offline and switched to the internet in 2020, it had already left the jurisdiction of the NCC. A piece of content originally under regulatory supervision, bypassing the AI process, finally lands on a fan page where no competent authority can manage it. Every step of this path makes it further from "accountability."

CTWANT News Channel was fined over 11.53 million NTD by the NCC for violations between 2014 and 2020; after the license was not renewed and it went offline in 2020, it switched to the internet, leaving the jurisdiction of the communications regulatory authority. Photo: NCC Yanping South Road Office. (Photography: Solomon203 / CC BY-SA 3.0)

Why use AI for this process instead of manual clipping? Scholar Wang Hongen gave a key criterion. When studying similar techniques, he pointed out that deliberately "using AI to rewrite articles" is "purposefully using AI to avoid investigation and detection" ⁷. Because identical content repeatedly copied and pasted will be caught by platforms as homogeneous, leading to reduced reach. AI rewriting makes every piece look different, smoothing out the fingerprint of the origin. In this usage, AI is a weapon against detection: it automates and scales the "washing of origin," saving even the cost of manual copying.

Before Exposure, After Exposure

"Every once in a while, the controversy of 'I Am OO Person' fan pages is mentioned," this is the opening sentence of "Kaohsiung Lives Well's" exposure article ². This sentence is not rhetoric.

At the end of December 2025, the military issue fan page "New · 27th Brigade" had already named "I Am a Taipei Person" and "I Am a Kaohsiung Person," saying their "content is almost entirely news written by AI based on current events, paired with AI-generated images," and "all openings and formats are almost identical," suspecting behind them "organized action for upcoming election public opinion operations" ⁸. This was about two months before the systematic exposure in February 2026; at that time, no one had pulled the thread to that company in Neihu.

And after being exposed in February, this production line did not stop. In March 2026, the same series of fan pages issued AI-generated fake images, spreading rumors that Taiwanese fans, after watching a game at Tokyo Dome, left "areas piled with bento boxes, beverage cups, plastic bags, and other garbage," and "female restrooms were a mess," which was exposed as coming from the same source ¹. A content production line already named by national media continued to operate normally within a month after exposure; this itself illustrates one thing: no mechanism can stop it.

Another "She" in the Same Ecosystem

In June 2026, another wave of seemingly completely different things emerged on Taiwan's social media: a large number of dating accounts, with the self-introduction "In Taiwan, only marry Taiwanese."

A citizen account ctchart.lab using AI tools for data analysis compiled a set of numbers: 252 accounts, 458 posts, of which 99.6% were concentrated in just three days from June 2 to June 4; at 10:37:53 AM on June 4, nine accounts posted in the same second; there were 90 waves of "same-second simultaneous posting" within three days; 32 sets of self-introductions were identical; the accounts' "factory specifications" were highly consistent: 100% single females, 98% marked no children, age concentrated between 30 and 49 years old, 99.6% of BMI not exceeding 20.5, 90 had a height of 165 cm ⁹.

The most critical column is geographic location. Of the 252 accounts, only 25 had "Account Location" marked: 21 China, 3 Pakistan, 1 USA—Taiwan, 0. The remaining 227 were all hidden ⁹.

📝 Curator's Note
Seeing the elements "China IP + Pretending to be Taiwanese + Only Marry Taiwanese," the most natural reaction is to blurt out "Cognitive Warfare." But the evidence actually points to another direction. What these accounts are truly doing is guiding people to add LINE "Sincere Dating," and directing targets off the platform is the fingerprint of "Pig Butchering Scam" fraud. Political trendsetting wants to keep people on the platform to wash comments and boost volume; it would never take people off-platform to LINE. Looking at geography: Pakistan is not a typical distribution for political cyber armies, but it is a known location for transnational fraud parks: in 2026, Pakistani police raided a fraud center in Faisalabad, arresting 149 people, of whom 48 were Chinese nationals ¹⁰. The same shell does not necessarily contain the same organism.

It needs to be clarified: The number of these 252 accounts was currently only produced by one citizen using AI via ctchart.lab, without third-party verification. It is a clue worth pursuing, not yet a conclusion.

Placing the dating account case in this article is not because it is the same batch of people as "I Am OO Person"; they are likely two different things. It is placed here because it shares the same outer shell logic with "I Am OO Person": a pleasant, harmless, "local" labeled identity, followed by an invisible backend. The difference lies only in what the backend wants to take from you. Seeing a China IP and shouting "Cognitive Warfare," or seeing a fellow townsman's tone and believing it is a fellow townsman, is actually the same kind of laziness.

"Coordinated Behavior" Does Not Equal "Cognitive Warfare"

The most fitting academic term for "I Am OO Person" is "Coordinated Behavior."

The Taiwan Information Environment Research Center (IORG) defines it as: "The behavior of repeatedly publishing the same or highly similar content, or the same links, on news media, social media, and instant messaging platforms within a short period" ¹¹. Note the restraint in this definition: it only describes the "behavior," not assuming it is overseas forces behind it, nor assuming malice. Nearly a hundred format-consistent fan pages synchronously reposting the same content fits the characteristics of "Coordinated Behavior," but "Coordinated Behavior" itself is only a neutral behavior marker.

📝 Curator's Note
The words most easily misused in Taiwan in recent years are "Cognitive Warfare" and "Cyber Army." IORG researcher Wang Xi said a heavy word: "The element constituting a 'Cyber Army' is the existence of a 'monetary consideration' relationship"; if "whenever you dislike a discourse you put on the 'Cyber Army' hat... it is very dangerous for Taiwan's public discussion," because "casually accusing cognitive warfare is exactly what information manipulators want to see," the more you stick labels randomly, the better the real cyber army can hide ¹². Wasting the word is equivalent to helping the enemy clear a larger camouflage.

Even the person with the least reason to speak for this technique refuses simple attribution. In his 2021 TSSCI paper on Chinese cognitive warfare, Shen Bo-yang drew a "Diamond Model," laying the initiators of information manipulation on a coordinate. In the "Economic Motivation" box, he clearly listed "commercial content farms, general YouTubers/streamers, fan page administrators"; in the "Political Motivation" box, it is the National Security Bureau, PLA, State Council Taiwan Affairs Office, Publicity Department, and CAC ¹³. He wrote clearly in the paper: "Initiators confusing the public's cognitive system are not limited to overseas forces" ¹³. Motivation is a continuous spectrum from commercial to political, with various "helpless" roles standing in the middle.

Even Meta thinks so. In its 2020 official statement, it wrote that when handling "coordinated inauthentic behavior," the platform "looks at behavior, not content, regardless of who is behind it, what was posted, whether overseas or domestic" ¹⁴, and admitted that coordination and authenticity are two spectra, not black-and-white switches. From Taiwan's researchers to the world's largest social platform, the same advice is derived: Look clearly first, then define.

Not Fake News Does Not Mean No Problem

If you use the ruler of "verifying truth" to measure "I Am OO Person," you will measure an awkward result: it basically measures nothing.

Because it mostly reposts real news. The content reported by CTWANT and CTWANT is not necessarily false; after AI rewriting, the factual level often withstands verification. Zhongzheng University's Huang Junru distinguished two easily confused terms: disinformation is "malicious information," intentionally causing harm; misinformation is "erroneous information," unintentional error, the biggest difference lies in intent ¹⁵. By this standard, "I Am OO Person" likely counts as neither, because it does not operate on "content being false."

Its problem is at another level. IORG splits information manipulation into three types; besides "Fact Manipulation," there is another called "Source Manipulation" [^6]: the source is hidden, disguised. "I Am OO Person" does exactly this: washing away the Wang Zhong series origin, attaching the identity of fellow townsmen. The content is true, but your judgment on "who this is, why you are shown it" is tampered with. IORG has an even colder metaphor: even if no one reads a single post, these contents are "accumulating local keyword inventory," which is "fertilizer for future fake news" ¹¹.

And in all reports, the same reservation appears repeatedly: "Summarizing currently public information, it is impossible to understand where the clients and cash flow of this series of companies come from" ¹. This sentence is the safety valve this entire article must hold. No evidence shows Lin Xianming is Chinese capital or a CCP agent; all investigated are "cannot be found," not "found." Existing evidence only supports a low-profile local marketing group in Neihu, a secondary distribution channel for pro-China media, an AI localization process, and an unclear cash flow. It claims it can "insert specific content" for sponsors ², but who that sponsor is, no one can find.

A Forgotten Prompt Editor, and a National-Level Content Farm

To see the position of "I Am OO Person" clearly, the best method is to place it together with a true national-level operation.

In February 2026, Wang Hongen exposed another case. A content farm called "Borderless Media," with a website hosted in Qinhuangdao, Hebei, "has direct contact with the Propaganda Department and the CAC." Its residual AI instruction was written like this: "For Taiwan users, use Traditional Chinese, re-edit this article, word count controlled at 500, retain the original opening part, do not change the original historical truth" ⁷. The "For Taiwan users" and "do not change the original historical truth" in the instruction are a top-down, politically motivated production line speaking.

Placing this instruction side by side with "I Am OO Person's" "Remove sensitive words, strengthen local Taiwanese colloquialisms," the difference is obvious at a glance. Wang Hongen himself separates the two very cleanly; he specifically reminds that some cases "are not just a domestic part-time editor forgetting to delete the prompt" ⁷, implying that some cases are just that. One is a local business in a Neihu company greedy for traffic; one is a national-level operation connected to the CAC; one's prompt handles "local tone," one's prompt handles "historical truth." They can be compared on the same gray spectrum, but absolutely cannot be confused.

Control Group: AI instruction residual on Qinhuangdao "Borderless Group" post exposed by Wang Hongen—"For Taiwan users, use Traditional Chinese, re-edit this article, word count controlled at 500, retain the original opening part, do not change the original historical truth." This is a national-level operation connected to the CAC, completely different from "I Am OO Person's" local profit-making prompt. (Source: Voice Tank / Wang Hongen, reported and reviewed citations)

Do not connect the thread too quickly; this is the lesson most worth remembering in recent years. During the 2022 pandemic, after artist Guo Hanyun relayed "many children just left," IORG found at least 25 fan pages posting the identical "Many Children Left!" within 10 seconds, with the first comment all linking to the same content farm ¹⁶; that was a textbook-level coordinated operation. But even in that case, Storm Media's verification reminded that mainstream media reports were actually earlier than the clustering push of those fan pages; "many fan pages posting simultaneously" does not necessarily equal an information war ¹⁷. Moreover, those named fan pages (Ai Experience, BuzzHand, that string) were not equated with "I Am OO Person" by any public data. Kaohsiung Lives Well said this series also posted similar content during the pandemic, but did not provide screenshots or direct links. Looking similar does not mean it is the same batch of people.

It is worth mentioning that the matter of "fake local identity" has a purer political version in Taiwan earlier. Taiwan Democracy Lab recorded during the 2024 election, an overseas group used hundreds of fake accounts to operate life fan pages like cameras, cooking, and art; fan page names and categories were all life types, but post content only related to Taiwan politics and elections, and "all administrator locations were located overseas" ¹⁸. That was "overseas fake accounts disguising as local." "I Am OO Person's" disguise layer is actually one level deeper: it is "real local entity + local media content + AI-enhanced local tone," using regional identity like "I Am a Kaohsiung Person" as turf, harder to dismantle than a fake cooking fan page.

Everything Is Legal, That Is the Problem

The most unsettling thing about "I Am OO Person" is that almost every part of it is legal.

Registering a company is legal. Reposting news is legal. Using AI to generate image-text is legal. Accepting a sponsor's commission to do native advertising placement; as long as the conflict of interest is disclosed, the Fair Trade Commission's testimonial advertising regulations cannot touch it ¹⁹. Assembling these legal parts together yields a machine impersonating fellow townsmen, with unknown origin, that can be inserted by sponsors, but you cannot find which screw is illegal.

In the past, discussing this situation, the habitual saying was "governance vacuum." But in 2026, this saying needs correction, otherwise it will be caught by readers who know the law. The vacuum has already been filled by one piece. The Regulations on the Prevention and Control of Fraud Crime Hazards, passed by the Legislative Yuan in 2024, commonly known as the Anti-Fraud Special Law, Article 31 requires online advertising platforms to disclose "information of the commissioned broadcaster/sponsor," and mark whether AI-generated images are used; Article 32 requires platforms to take down fraudulent advertisements within a time limit, otherwise they bear joint liability for compensation with the sponsor ²⁰. The Ministry of Digital Affairs has already fined Meta three times according to this law, cumulatively 18.5 million NTD ²⁰. The "fraudulent advertising" vacuum has already been patched by a toothed law.

The problem is, this law only binds "paid advertising." It can manage a paid fraudulent advertisement, but cannot manage a free-operating, fellow-township-impersonating "I Am a Kaohsiung Person," because that is free-operating "content," not within the jurisdiction of "advertising." And the horizontal law that can truly manage "platform accountability, operator disclosure," namely the Digital Intermediary Service Act of 2022, was shelved due to concerns about "internet censorship" in that year ²¹, and has not revived as of 2026.

The Digital Intermediary Service Act, which manages "platform accountability, operator disclosure," was shelved in the Legislative Yuan in 2022 due to controversy, and has not revived as of 2026; the Anti-Fraud Special Law, which only binds "paid advertising," patches the vacuum but cannot manage free-operating fellow-township-impersonating fan pages. Photo: Legislative Yuan Chamber. (Photography: Lin Gaozhi / CC BY-SA 4.0, Wikimedia Commons)

📝 Curator's Note
So the governance vacuum has not disappeared; it has just shrunk—from "comprehensive" to "non-paid political and emotional operation fan pages." A fan page impersonating Taiwanese, not disclosing operators, not charging you money, only wanting to influence you, falls exactly in this shrunk hole: it is not advertising, so the Anti-Fraud Special Law cannot manage it; the law that manages it lies in a place no one dares to touch. Reporters' Liu Zhixin used a saying: This ecology turns "citizens in a society" into "mercenaries of the hidden hand behind the scenes," called "recruiting mercenaries" ²². Those recruited may not know they are recruited; you, liking and reposting, may also not know who you are working for.

So, Do You Still Know Who You Are Reading?

Back to that "I Am a Taipei Person" post, and that forgotten instruction: "Remove sensitive words, strengthen local Taiwanese colloquialisms."

What this sentence truly leaked is not that a certain article fabricated a lie, but that an entire technique of manufacturing "identity" was exposed to the sunlight: first erase the origin, then attach the local. After AI makes this process almost zero-cost and able to automatically evade detection, recognizing the truth of a news item is no longer the hardest thing; it mostly reposts real news. The hardest thing is recognizing its origin and cash flow: Who wrote this? Why do you want to see it? Did someone pay for it?

And what truly protects you is never a stance—whether it is the CCP? Whether it is cognitive warfare?—because even Taiwan's most capable experts in cognitive warfare are calling for caution against hasty labeling. What protects you is the ability to trace the source: Ask one more question "Who is this? Where is the origin? Who paid for it?"

That forgotten instruction was caught this time. Next time, it will not forget to delete.

Further Reading

• Cognitive Warfare — The framework, boundaries, and "complexity over accuracy" reading principle of cognitive warfare
• Poisonous Potato Cognitive Warfare — How a narrative war "steps on the foundation of truth," belonging to the same paradigm of rejecting binaries as this article
• Taiwan Artificial Intelligence Lab — Local technology using AI to detect coordinated cognitive operations
• Taiwan Media and Press Freedom — The larger context of Wang Zhong controversies and media group structures

References

Image Sources

Most images in this article are screenshots of public pages, cited under Article 52 of the Copyright Law "Fair Use for Reporting and Commentary," with sources marked:

• Cover Image/Fan Page Side-by-Side: "I Am a Kaohsiung Person" "I Am a Taipei Person" Facebook Fan Pages (@Kaohsiung.Info, @Taipei.Info) screenshots.
• Exposure Summary: Fan Page "Kaohsiung Lives Well" (@takaogoodday) Threads post screenshot.
• Jishuo Technology Commercial Registration: opengovtw, data from Ministry of Economic Affairs Commercial and Industrial Registration Public Disclosure.
• LIFE Life Network Service Page: aams.tw (Jishuo Technology) screenshot.
• Qinhuangdao Content Farm AI Instruction: Screenshot from Voice Tank / Wang Hongen article.
• NCC Yanping South Road Office: Photography Solomon203, CC BY-SA 3.0 (Wikimedia Commons).
• Legislative Yuan Chamber: Photography Lin Gaozhi, CC BY-SA 4.0 (Wikimedia Commons).
• Dating Account Screenshots involve real photos that may have been stolen; this article does not reproduce them based on privacy, only presenting data and text descriptions.