By Rjay Zuriaga Casto

Iloilo City Mayor Raisa Treñas-Chu denied it, but the data raises serious questions.

As the city government of Iloilo intensified enforcement against transient vendors or bolanteros, a pattern also emerged online: suspicious accounts repeatedly appeared in the comment sections of news outlets to defend the city’s policy.

The controversy stems from the stricter enforcement of the Iloilo City Market and Slaughterhouse Code of 2009, which prohibits the sale of fruits, vegetables, eggs, meat, and native delicacies outside public markets or within 100 meters of market premises.

Although the ordinance had long existed, it was not always strictly enforced. That changed after the reopening of the redeveloped Iloilo City Terminal (Super) Public Market and Iloilo City Central (Tienda Mayor) Public Market in November 2025, when the Iloilo City Local Economic Enterprise Office (LEEO) stepped up enforcement.

Treñas-Chu has rejected allegations that the city government maintains a troll network to defend the city’s controversial enforcement.

In a press briefing on April 23, she said the supportive comments were organic and insisted the city government was not behind the suspicious accounts being flagged.

“We do not have a troll network. If that will be flagged, you have to review who these people are, but I am confident to say that these are not from us,” she stressed.

She said the city government had nothing to do with the accounts and suggested that those posting supportive comments may simply have been market vendors helping defend the policy online or people directly affected by the issue.

“Everything is organic […] [Those that comment positively, they are the vendors themselves. You have to go inside the markets, they said, ‘mayor it will be us who will defend in the pages’,” she said.

“It is not really us. It is the vendors that said that they are helping in the comment section,” she continued.

But the data reviewed by Daily Guardian (DG) complicates that claim.

Across multiple news outlet pages, suspicious accounts did not merely express support and did not simply appear at random. They arrived in bursts, repeated the same defenses, crossed between political, government, and media spaces, and were tied by overlapping following and friendship networks.

Those patterns do not disprove the possibility of genuine support. They do, however, raise questions about whether all of the visible support was spontaneous and independent.

For an account to be flagged as suspicious, it generally had to show at least two observable indicators rather than just one. Greater weight was placed on profiles that appeared new or empty, lacked visible personal content, and showed limited identity signals — markers that more directly suggest an account may not belong to a genuine, active user.

Low friend or follower count was treated only as a secondary indicator. No fixed numerical cutoff was applied, because manual review found that some accounts displaying suspicious activity patterns still carried friend or follower counts of up to 1,000 — making raw numbers alone an unreliable basis for flagging.

DG collated and monitored Facebook posts from print and radio news outlets across Iloilo City related to the bolantero controversy.

The review covered posts published from April 10—during the controversial nighttime crackdown along Fuentes Street—to April 15, when reports emerged about former Senate President Franklin Drilon’s visit to the Iloilo Central and Terminal markets.

Posts were capped at 10 per outlet, although some organizations had fewer than five relevant stories within the period.

Daily Guardian treated and analyzed three datasets:

burst windows of comments across news outlets;
accounts engaging across the pages of Treñas-Chu, the Iloilo City Government, and news outlets; and
accounts engaging only on news outlet pages.

The more than 1,500 accounts — 1,598 in total — were drawn from comment sections of news outlet posts from like Daily Guardian, Alston, and other publications, scraped using Apify. Because the dataset was too large for immediate full manual review, an AI-assisted classification was used as a first-pass filter. Each comment was sorted by stance toward the Iloilo City government: supportive or positive, critical or negative, or neutral and mixed.

Comments classified as critical of the city government were excluded from the working dataset at this stage. The reasoning was directional: the investigation’s central premise was that accounts may have been deployed to boost or defend the city government’s image online, making the critical-comment accounts outside the scope of that inquiry. AI classification proved reliable in identifying negative-leaning comments, allowing the team to set them aside with confidence.

This first filter reduced the dataset to 846 accounts whose comments were either supportive or neutral and mixed. These were then subjected to a second manual review, this time based on actual comment content. After that pass, more than 200 accounts remained as those that had repeatedly posted comments supportive of or defensive toward the city government. Only at that stage were the accounts assessed for profile-level indicators — whether a profile appeared new or empty, lacked visible personal content, showed limited identity signals, or carried unusually sparse public activity.

Burst windows

Burst windows refer to short periods when comments suddenly cluster within narrow time intervals.

Across 63 posts from news outlets, these bursts suggest that some Facebook discussions surrounding the transient vendor controversy were not driven solely by spontaneous public reaction.

In ordinary public discussion, support, criticism, jokes, arguments, and unrelated remarks tend to appear unevenly and unpredictably.

In analyzed datasets, however, many of the bursts were both compressed in timing and narrow in narrative direction.

Across multiple posts, government-aligned comments repeatedly arrived in narrow, and repeatable time clusters.

Rather than showing a wide mix of viewpoints, these bursts often revolved around the same talking points:

the law should be followed,
fairness requires discipline,
order must be maintained,
vendors have legal alternatives such as Market Day
and the city government is helping rather than oppressing.

The comments also did not appear randomly throughout the day. Instead, they clustered in recurring time pockets: late morning to noon and mid-afternoon.

One of the clearest patterns appeared from morning to noon. In several burst windows around 8:30 a.m. to 12:00 noon, comments defending the city government arrived in short succession and emphasized the same message: rules must be followed, order is necessary, and sympathy for vendors should not override the law.

A second pattern appeared in the mid-afternoon, particularly around 3:55 p.m. to 4:24 p.m. Here again, supportive comments came in compressed intervals and repeated familiar lines about public order, regulation, and legal alternatives for vendors.

In contrast, the government-critical comments tend to arrive when a post is still emotionally hot or freshly circulating, while many government-aligned comments more often appear in follow-on defense windows that try to reframe the discussion.

DG’s analysis of the time burst does not by itself prove that the supportive accounts were inauthentic. It does not prove payment, command-and-control, or a formal political operation.

Janina Santos, a data analyst of DoubleThink Lab, said timing clusters become more suspicious when paired with other warning signs.

“The clustering of timing becomes suspicious when coupled with other indications of inauthenticity and coordination. For instance, close or clustered posting time of copy pasta narratives or even similarly-themed messages by possible fake accounts,” she stressed.

Prof. Sharlene Gotico, teacher of communications and currently affiliated with the Central Philippine University, said the compressed timing and narrow narrative direction of the government-supportive comments raises some red flags.

In highly charged public issues, she said, genuine public reaction is usually uneven, dispersed, and varied in tone and language.

“While this in itself does not absolutely prove orchestrated activity, it indicates a degree of message consistency and timing that may reflect more than just ordinary and organic public reaction,” she added.

Page crossovers

Out of 273 accounts showing crossover activity across the pages of Iloilo City Raisa Treñas-Chu, the city government, and news outlets, 41 accounts were manually reviewed and flagged as suspicious based on indicators such as:

new or nearly empty profiles
low friend/follower counts
no visible personal content
limited identity signals

Based on their comment content, these 42 accounts consistently promoted narratives centered on legality, discipline, and order.

To avoid presuming that suspected accounts were directly linked to or operating on behalf of the city government, DG separately filtered accounts whose visible activity was limited to direct engagement with news outlet posts alone.

From more than 1,500 scraped accounts, DG also manually identified 50 suspicious-looking accounts within this group that carried similar talking points and likewise defended the government’s position.

Agenda-setting and narrative control

Santos said repeated activity by the same accounts across political pages, government pages, and independent news outlet threads may, on one hand, reflect genuine public interest in local political developments.

But when such activity is coupled with other indicators of inauthenticity and coordination, she added that it can also point to influence operations meant to protect reputations, shape public opinion, or encourage compliance.

“At the end of the day, their aim or objective is secondary. If they are using deception and coordinated manipulation to realize their aim, this is still something we ought to question and expose,” she said.

Gotico said the repeated use of the same frames can be read as a strategic effort to simplify a complex controversy into a more manageable public narrative.

The effect, she said, is not only persuasive but interpretive: it narrows the range of meanings available to audiences and privileges one frame over others.

“It positions one frame over the other frames and may potentially provide legitimacy to that frame,” she said.

Gotico said the pattern in which pro-government responses often appear after criticism is especially important.

She pointed out that the sequence suggests “reactive communication and partly crisis communication as a response to a reputational threat.”

“In communication theory, this can be viewed as likely agenda-setting, as it elevates certain themes. My personal opinion is that it is strongly like narrative control because apparently, the effect is to shape how the controversy is interpreted rather than just adding another opinion into the public discussion,” she emphasized.

She added that support expressed on political or government pages is expected. But support appearing in third-party environments such as independent media pages carries greater persuasive value because it creates the impression that the message is not merely official government messaging, but something affirmed by the wider public.

In that sense, Gotico said repeated supportive comments in news outlet threads can generate an appearance of consensus around a particular frame.

The friends-following network

The friendship and following patterns add another layer to the comment analysis. Beyond the timing bursts and repeated talking points, the account-level network data suggests that many of the suspicious profiles were not moving independently across Facebook.

Among the accounts that engaged across the pages of Treñas-Chu, city government, and news outlets, several showed overlap in both the pages they followed and the friends they were connected to.

The overlap was not limited to one type of page. Many shared following pages were food-related, while others were directly tied to local politics, government, and media.

On the friendship side, several of the suspicious accounts are not just active in the same issue space; they are also embedded in overlapping Facebook friend networks. Some accounts share the same friends, while others appear as direct friends of one another.

A similar pattern also appeared among the suspicious accounts whose visible activity was limited to news outlet comment sections.

Even when the visible activity of these accounts is limited to media pages, they still converge in the pages they follow, including media organizations, the Iloilo City Government, Mayor Raisa Treñas-Chu, Councilor Miguel Treñas, and former Mayor Jerry Treñas.

They are also tied through overlapping friendships, suggesting that even the accounts that do not visibly engage official government or political pages are still situated within the same broader monitored environment.

Santos described these patterns as “relatively strong” evidence of networked behavior.

“Often, network pages and accounts are single-purpose, meaning largely political. It must be noted however that many inauthentic assets also post/repost content that is non-political, in this case food, to disguise the real purpose of the platform and to have a captured audience of those interested in their other content,” she said.

“The [following] train exhibited by the [accounts] shows strong networked behavior,” she emphasized.

Gotico likewise said the direct friendship links among several of the accounts point to engagement occurring within a connected account environment.

While this does not automatically mean the network is centrally managed or fake, she said it weakens the argument that the accounts are merely unrelated users arriving by coincidence at the same time, in the same spaces, while using similar tones and frames.

She added that if many of the same profiles are following similar pages and then appearing in the same comment spaces, that suggests they are embedded in the same digital informational ecosystem.

“This is important because coordinated, and/or semi-coordinated communication behavior, usually depends not just on shared messaging but also on shared sources of triggers and cues,” she explained.

Structured, strategic operation

Santos said the similarities in the suspicious accounts’ comment themes are difficult to dismiss as incidental.

“The findings are more consistent with a more structured influence operation. The patterns of behaviors indicate coordination, specifically the network of accounts involved in coordinated commenting. The similarities in the themes of the comments is also highly suspect,” she said.

She also pointed out that the combination of indicators allows for a reasonable assumption that some level of coordination may be present.

“The presence of inauthentic accounts on the government pages as well as the local media platforms, the network graph, and the reiteration of the same themes and messages are indicators of this,” she emphasized.

Similarly, Gotico said the findings point toward a messaging pattern that is “repetitive, structured, and strategic, rather than something that is organic.”

She highlighted that the data reasonably supports the view that discourse around the ambulant vendors issue may have been shaped not only by genuine public opinion, but also by cross-platform amplification, message reinforcement, and possible narrative management.

“The observed patterns appear less consistent with purely organic and spontaneous public support and more consistent with some degree of organized supporter mobilization and strategic communication management, though not in a way that outrightly allows us to make absolute claims about intent and authorship, including direct coordination,” she added.

Illusory consensus

Beyond timing patterns and network overlap, the findings also point to a deeper psychological effect: the manufacture of apparent public agreement on the issue concerning the bolanteros.

Gotico said repetition matters not only because of what is being said, but because of what repeated messaging signals to everyone else reading the thread.

“Repetition, especially across multiple spaces and accounts, can create a perception of social agreement, as people generally do not only read the comments section for additional information, but also for cues about what others seem to think,” she said.

That repetition matters because visible repetition can silence disagreement, according to Gotico.

She pointed to Elisabeth Noelle-Neumann’s Spiral of Silence theory, which holds that people are less likely to speak up when they think their opinion is in the minority, because they fear social isolation.

As a result, opinions that seem to have majority support become even more visible, while opposing views become quieter and seem rarer than they really are.

“When repetition may be strategically coordinated, this can create an information/discussion vacuum in which those who hold dissenting/opposite opinions would rather stay quiet and not participate in the discussion. When this happens, social media visibility rests on the message that is repeated,” she stressed.

Santos, meanwhile, said manipulated engagement by inauthentic or coordinated networks can also spill over into real support through the “bandwagon effect.”

“‘If more people support it, then maybe it’s right.’ This mindset is human, and is taken advantage of by actors behind fake amplification,” she said.

That, Santos emphasized, is what makes fake amplification powerful, as it does not only flood a thread with favorable comments, but to shape how ordinary readers interpret the climate of opinion around an issue.

She warned that even authentic users may start echoing the dominant line once they perceive it as the majority view.

Santos said that in this way, coordinated repetition can generate real downstream effects, blurring the line between manufactured consent and genuine support.

“Even if the aim of the actor is good, if they go about it in a deceptive and manipulative manner, then that is still problematic. The end does not justify the means,” she said. — With research by Rei Ebenezer Duhina, Daily Guardian Intern