Olesia Buriak


Emoji reactions as an influence tool: the Krakow case

Emoji reactions as an influence tool: the Krakow case

Emoji reactions added in bulk to a few posts on Telegram channels targeting Ukrainian audiences on 7 November sat almost unnoticed for two days, then were cited as evidence in a screenshot post that reached 752K views. The anomaly was visible in the data on the first day, before anything had been built on it.

Key judgments

  • Emoji reactions were added in batches to single posts on large Telegram news channels targeting Ukrainian audiences on 7 November. Laughing or gloating reactions spike on one Krakow post on each of three channels, 3,940 to 5,218 per post, and sit at the channel's usual level on the other 40 Krakow posts. Between the first snapshot and three to four days later, these reactions added about 1 to 3% of their final count, while views added 15 to 23% and other reactions on the same posts about 10 to 20%. Confidence: high for @times_ukraina and @insiderUKR, where a control post on the same channel kept growing; moderate for @voynareal, which has no control post of its own.

  • Heart reactions on first-half-hour posts carry the same signature on all six channels. Posts about the attack and its victims published between 15:01 and 15:28 UTC carry 5 to 28 hearts per 1,000 views, later posts 0.1 to 1.3, and the hearts stopped growing within hours. We assess it likely that a batch sits on top of some organic hearts. Confidence: moderate, because hearts can express sympathy and the organic share cannot be separated.

  • The behaviour is wider than the Krakow story. The same frozen-growth signature appears on a post about Russian strikes on Kyiv, published 27 minutes before the first Krakow report, and on thumbs-down reactions under President Zelenskyy's condolences to Poland. We assess it likely that reaction batches are a recurring practice on at least some of these channels. Confidence: low to moderate, resting on two posts outside the main set.

  • The batch was in place two days before the claim built on it. The counts were present on 7 November at 19:17 UTC. On 9 November a Polish-language X account published a screenshot of them (7572K views). The asset alleged that the counts showed Ukrainians celebrating the attack. Russian Telegram channels picked up the reactions angle from 9 November. Confidence: high on the sequence. The data does not show whether the same actor was behind more than one stage.

  • Reaction counts are weak evidence of public mood, in either direction.On the channels where 😁 is available and in use, every Krakow post carries some laughing reactions at the channel’s usual level, 35 to 429 per post. . This report makes no assessment of what Ukrainian audiences feel about the attack. Its finding concerns the specific counts that were presented as evidence.

  • This report attributes the batches to no actor. Reactions, views and followers for any public Telegram channel can be bought from commercial services by anyone, with no access to the channel. A batch therefore says nothing about the role of the channel it sits on. The channel's own administrators, a competitor and a hostile actor are all consistent with the data.


The case in brief

On 7 November 2026 a shooting in the centre of Krakow killed several people. Large Telegram news channels targeting Ukrainian audiences reported it from 15:01 UTC. All times in this report are UTC.

On 9 November, Krengiel (@krulbijedame), an X account labelled as a parody account, posted a screenshot of the first report by Times of Ukraine (@times_ukraina). The asset, as this report calls any account, channel or outlet, alleged that the "5.6K hearts, 3.9K smiley emojis" under the report showed Ukrainians celebrating the deaths of Poles. It drew 7572K views and 2K reposts, and the claim spread further on Polish-language X (example, example, example), on the Polish site Magna Polonia, and in Russian Telegram channels.

The asset's claim rests on one kind of evidence: reaction counts under a post. This report examines how those counts behave.

Figure 1. The two screenshots posted by Krengiel (@krulbijedame) on 9 November.


The pattern

A reaction count that grows with its audience has a recognisable shape. It rises while new readers arrive, it appears across posts on the same story, and it moves together with the other reactions on the post. The counts examined here break that shape in three ways.

Two terms carry the analysis. A spike is a reaction count far above the channel's baseline on a single post. A batch is our assessment of how a spike arrived: many reactions added in a short window, where organic reactions accumulate one by one as readers see the post.

  1. Single-post concentration. The reaction is high on one post and at baseline on the posts before and after it, on the same channel and the same story.

  2. An added layer. With the spiking reaction removed, the post looks like its neighbours: similar views, similar comments, similar mix of other reactions.

  3. Early freeze. The spiking reaction is almost complete within hours and then stops, while views and other reactions on the same post keep rising for days.

The three markers appear together on laughing and gloating reactions (section 1), on hearts (section 2) and on posts outside the Krakow story (section 4). Section 3 covers the timing evidence, which carries the most weight. None of the markers depends on what a post says, so the same test applies to any channel in any language.


1. Laughing and gloating reactions spike on single posts

What was observed. Laughing or gloating reactions spike on one Krakow post on each of three channels and on none on the other three. Counts are given per 1,000 views so posts with different reach compare fairly. Every other Krakow post sits at the tragedy baseline: the same channel's level of the same reaction on other tragic news that day, mainly the Russian strikes on Kyiv. That baseline is about 0.1 to 1.0 per 1,000 views (up to 1.23 on @voynareal, and up to 1.9 for 🔥 on @insiderUKR).

Channel

Views per post

Krakow posts

Reaction read as laughing or gloating

Spike post

Spike, per 1,000 views

Other Krakow posts, per 1,000 views

Times of Ukraine (@times_ukraina)

~450k

5

😁

15:01, first report: 3,940 😁

8.7

under 0.3 (35 to 106 laughs each)

INSIDER UA (@insiderUKR)

~420k

2

🔥 (😁 effectively unused that day)

15:01, first report: 3,946 🔥

8.9

0.27

Реальна Війна (@voynareal)

~320k

8

😁

17:02, sixth post: 5,218 😁

14.4

0.2 to 1.3 (63 to 429 laughs each)

УС (@u_now)

~150k

13

😁

none


0.1 to 0.7

Лачен пише (@lachentyt)

~545k

3

😁

none


0.2

Труха⚡️Україна (@truexanewsua)

~610k

12

🔥 (😁 disabled on these posts)

none


under 0.3 at baseline


Figure 2. Laughing or gloating reactions, and grief and anger reactions, on all 43 Krakow posts.
Figure 2. Laughing or gloating reactions, and grief and anger reactions, on all 43 Krakow posts.


How it behaved.

  • Concentration within a channel. On @times_ukraina, the first report carries 3,940 laughs and the next four Krakow posts 35 to 106 each, including a post at 21:21 with 88. This is 8.7 gloating reactions per 1,000 views on a spike post against 0.08–0.23 on other posts about Krakow. On @insiderUKR, the first report has 8.9 🔥 per 1,000 views and the second, 17 minutes later, 0.27. On @voynareal, the spike sits on the sixth of eight posts: 14.4 gloating reactions against 0.2–1.3 

  • No match across channels. Four channels opened with a short breaking report, three of them in almost the same words. Two of those first reports carry a spike (@times_ukraina, @insiderUKR, both 15:01). The other two, on @voynareal and @truexanewsua at 15:04, carry none.

  • The reaction was available and in use. 😁 stayed in use on later Krakow posts on @times_ukraina and @voynareal. The @lachentyt and @voynareal audiences use it heavily on political news, up to 31 and 18 per 1,000 views. Therefore, the near-absence of laughs on the other Krakow posts is not down to a switched-off reaction or an audience that doesn’t use it. 

  • The spike posts are otherwise ordinary. Their views match the channel's other Krakow posts (0.9 to 1.2 times the daily median), comments sit close to the channel median where available, and grief and anger reactions fall within the range of the channel's other Krakow posts.

  • Grief and anger are distributed differently. They are spread across all Krakow posts on every channel. Laughing and gloating sit near zero everywhere except one post per each of the three channels; three other channels have no gloating spikes at all.

Reading. A reaction that reflects how a channel's readers respond to a story appears across that channel's posts on the story, as grief and anger do here. A count that appears once and nowhere else on the same story is consistent with a batch placed on that post.


2. Heart spikes in the first half hour

Analyst investigation, using reaction and view counts collected by Vantage.

What was observed. On all six channels, Krakow posts about the attack and its victims published between 15:01 and 15:28 carry 5 to 28 hearts per 1,000 views. Later posts carry 0.1 to 1.3, even when they are about the victims. The one exception is President Zelenskyy's condolences on @u_now, at 7.6.

Figure 3. Hearts per 1,000 views on all 43 Krakow posts, by time of publication.


How it behaved.

  • On or off, with no gradient. After about 15:28 the heart spike disappears whatever a post is about. Before that, posts describing the shooting carry the spike, while posts mainly about the shooter's arrest or photos of him stay near 0.5. On @u_now, the 15:07 post ("5 killed") has 25.1 per 1,000 and the 15:20 post ("shooter detained") has 0.5. On @truexanewsua, a 15:06 post about police neutralising the shooter has 0.8, between two posts with 7 to 9.

  • An added layer. With hearts removed, a spike post matches its neighbours. On @u_now, the 15:07 post has 10.4 other reactions per 1,000 views and the 15:29 post also has 10.4.

  • Early freeze. Hearts on the spike posts stopped growing within hours, while hearts on neighbouring posts without a spike kept growing with views (section 3).

  • Top of the day. On four channels the most-hearted post of the whole day is a Krakow post, above good news such as new US aid.

Reading. Hearts can express compassion, and solidarity with Poland in the first minutes is a plausible explanation for part of these counts. We assess it likely that a batch sits on top of some organic hearts: the on-or-off pattern and the early freeze match the laughing spikes and differ from ordinary hearts on neighbouring posts. On its own this cannot rule out a separate group of real readers adding hearts early. Views cannot settle the question either, since a few thousand extra visitors would barely change a count of around 450,000.

In a screenshot a heart reads as sympathy as easily as approval, so the hearts matter less for the claim itself. They matter for the pattern: the same behaviour appears on a reaction with a different meaning.


3. The spiking reactions stopped while readers kept arriving

Analyst investigation, using two Vantage snapshots per post: 3 to 10 hours after publication, and 3 to 4 days later.

What was observed. Between the two snapshots, views added 15 to 33% of their final count and most reactions added about 10 to 20%. The spiking reactions added about 1 to 3% (3.5% in one case). The comparison covers 14 spiking reactions and 8 control reactions on posts without a spike.

Figure 4. Share of each final count added after the first snapshot: posts with a spike (top) and control posts without one (bottom).


How it behaved. The clearest evidence is the same reaction on two consecutive posts of one channel.

Channel and reaction

Post with the spike

Next Krakow post, no spike

@times_ukraina, 😁

15:01 post: 0.8% added (30 laughs, from 3,910 to 3,940)

15:18 post: 8.6% added (32 to 35 laughs)

@insiderUKR, 🔥

15:01 post: 1.3% added

15:18 post: 12.8% added (102 to 117)

Hearts, several channels

Spike posts: about 1 to 3.5% added

Posts without a heart spike: 11 to 37% added, roughly in step with views

Views kept coming. 22% of the @times_ukraina's 15:01 post’s final views came after the first snapshot, but only 30 laughs (0.8%)  Had laughs kept pace with views, the post would have gained about 1,080. 

  • Only the spiking reaction froze. The 15:18 posts on @times_ukraina and @insiderUKR serve as controls for 😁 and 🔥, and they also carry a heart spike of about 7 per 1,000 views. On those two posts the hearts froze (2.4% and 3.5% added) while 😁, 🔥 and the other reactions kept growing.

  • @voynareal has no control. Its 17:02 spike shows the same freeze, and it has no comparison post from its own channel.

Reading. Readers keep reacting for as long as new readers see a post. These reactions stopped while views kept rising, which fits reactions added in one early window. The snapshots show that the window closed within 3 to 10 hours of publication (80 minutes on one post). They do not show the moment the reactions arrived, or who added them.


4. The same signature outside the Krakow story

Analyst investigation, using reaction and view counts collected by Vantage.

What was observed. Two posts outside the main comparison carry the same markers.

  • A post published before the attack. @insiderUKR's 14:34 post, about Russian strikes on Kyiv, has 18.6 hearts per 1,000 views. Its hearts were 94.5% complete after 80 minutes, while 🔥 on the same post behaved as an ordinary reaction and kept growing (47% added after the first snapshot).

  • A different reaction on a different kind of post. President Zelenskyy's condolences to Poland on @u_now (16:56) carry 3,958 👎 (25.2 per 1,000 views). The channel's level elsewhere is at most 2.4 per 1,000 views, and 0.1 to 0.3 on his other posts that day. The 👎 count shows the same early freeze. 👎 is not a hostile reaction to Poles as such, so it stays out of the laughing and gloating comparison.

Pattern observation: Three of the spikes were almost the same size at the first snapshot, on three channels and in three different reactions: 3,910 😁 on @times_ukraina, 3,896 🔥 on @insiderUKR and 3,898 👎 on @u_now, all within 14 of each other. Similar sizes would be consistent with batches of a standard size. The pattern does not cover every spike: @voynareal’s laughing spike (5,102 at the first snapshot) is larger, and the heart spikes vary widely. With three matching cases we note the pattern but draw no judgment from it. 

Reading. On at least one channel, heart spikes are not limited to the Krakow story and predate it. We assess it likely that reaction batches are a recurring practice on some of these channels. This supports the judgment that the counts were added. It also limits what can be said about purpose: the data does not show that the batches were aimed at Poland, or at this story.


5. From reaction count to screenshot to amplification

Analyst reconstruction, using Vantage snapshots and Russian-language Telegram data. 

What was observed. The counts were in place on the day of the attack. The screenshot built on them appeared two days later and then moved into Russian-language channels.

When (UTC)

What happened

Source

7 Nov, 15:01

@times_ukraina publishes its first Krakow report.

Telegram

7 Nov, 19:17

First snapshot of the post, about four hours after publication: ❤ 5,526 and 😁 3,910 are already there. Over the next three days they add under 2%, while the post's other reactions add about 10%.

Vantage snapshot

9 Nov, 15:37

Krengiel (@krulbijedame) posts the screenshot with arrows: "5.6 thousand hearts, 3.9 thousand smiles".

Krengiel on X

9 Nov, 22:34

Bezsonov, former "DPR" deputy information minister, republishes both screenshots with a dehumanising caption (21.1K views).

@NeoficialniyBeZsonoV, Andrzej_HS thread

10 Nov, 11:03

Коза кричала (@kozakrichala): "A Pole is surprised… in a large Ukrainian channel, nothing but hearts and smileys". The post spreads to at least 10 Russian Telegram channels and reaches 143K views.

@kozakrichala

10 Nov, 14:00

Polish site Magna Polonia publishes the reaction counts from the screenshot with hostile comments quoted from X, and calls on the security services to look into it. It notes that bots cannot be ruled out.

Magna Polonia

11 Nov

Соловьёв (@SolovievLive) runs a "Polish blogger compared Crocus and Krakow" post.

@SolovievLive

How it behaved.

  • The count came first. The first snapshot shows the counts on 7 November at 19:17. The screenshot followed 44 hours later.

  • Russian channels came to the reactions angle late. In the first two days they pushed a different line, a "Ukrainian trace" in the attack: RT (@rt_russian) on a patch reportedly worn by the shooter (7 November, 21:03), @SolovievLive the same evening, and Русский тарантасъ (@DmitriySteshin) calling the attack Poland's "payback" for supporting Ukraine (8 November). The reactions angle appears in this data only after Krengiel's screenshot.

  • The screenshot outran the post. The original Telegram post has about 450K views in Ukrainian. The screenshot drew 757K views on X in Polish, and the Russian-language repackaging a further 143K.

Reading. The sequence is count, then screenshot, then amplification. The data shows the order. It does not show that one actor ran more than one stage, or that the count was placed with the screenshot in mind.


Figure 5. Magna Polonia article of 10 November repeating the reaction counts from the screenshot.


Why it matters

A reaction batch is close to invisible when it lands. It adds no post, no account and no text. There is nothing to read, translate or fact-check. On a post with 450,000 views, four thousand extra reactions change one small number under the text. An analyst reading the channel that afternoon would have seen an ordinary breaking-news post.

Anyone can place one. Reactions, views and followers for any public Telegram channel are sold by commercial services (e.g., Telegram Reactions). The buyer needs a link to the post and no access to the channel. A batch can therefore be placed on someone else's post, without the knowledge of the channel's administrators, for very little money. It works twice. The first effect is immediate: everyone who opens the post sees what looks like the reaction of thousands of other readers. The second is deferred, and larger. The count stays on the post as a durable, public number that anyone can photograph later and present as evidence of what a population feels. A reaction count needs no translation and fits in one image, so it crosses languages and platforms more easily than any text.

In this case the deferred effect was the incident. The batch sat on the post for two days. Then one screenshot turned it into a claim about Ukrainians, in Polish, on another platform, with more views than the original post. Within hours it was repackaged for Russian-language audiences. The reaction count was the building block, and the screenshot was what was built from it.

The build stage is where the advantage is. The data does not show that the batch was placed for this purpose. It shows that the material for the claim existed two days before the claim. Seen on 7 November, the anomaly already says which posts carry screenshot-ready counts, which claim those counts would support, and which audiences that claim would be aimed at. That is a two-day head start on a narrative that reached 757K views in a single post. Seen on 9 November, the same anomaly is a rebuttal, written after the claim has travelled.

Anomaly is what makes the build stage detectable. None of the three markers in this report depends on what a post says. They are departures from a channel's own behaviour: a reaction far above its baseline, on one post, that stops growing while the audience keeps arriving. Detection built on behaviour finds this while it is still a small number under a post. Detection built on content finds it once someone has turned it into a claim.

For teams responsible for national or organisational resilience, the practical points are these:

  • Treat engagement numbers in a screenshot as an unverified claim until their behaviour over time has been checked.

  • Treat a reaction anomaly on a high-reach post about a sensitive event as an early indicator, even when nothing has been said about it yet.

  • Ask of any anomaly what it could later be used to prove, and to whom. The answer points to where the next stage would appear.

Outlook and what to watch

Reaction batches are cheap, fast to place, and already present on these channels beyond a single story. We assess it likely that reaction counts under posts about sensitive cross-border events will be used again as screenshot evidence of public mood, in this pairing of countries and in others. Confidence: moderate.

Signs that would sharpen this assessment or signal escalation:

  • a reaction batch on a post about a new sensitive event involving a neighbouring country or an ally;

  • a shorter gap between a batch and the first screenshot of it, which would point toward the two stages being linked;

  • the same accounts repeatedly screenshotting posts that carry a batch;

  • batches of near-identical size across unrelated channels, which would point toward a shared service or operator;

  • batches arriving on older posts shortly before those posts are screenshotted;

  • the reactions angle appearing in Russian-language channels before it appears in the target country's own language.

Method and caveats

This report covers every post published on 7 November 2026 by six large Telegram news channels targeting Ukrainian audiences, including 43 posts about the Krakow attack. Two comparisons carry the analysis: Krakow posts against each other within a channel, and the same kind of post across channels. Reaction and view counts were collected by Vantage. The comparisons, judgments and confidence levels are the analysts'. Quoted posts and claims are reproduced as material circulating online, and this report endorses none of them.

  • Timing gaps. Each post's reactions are seen at only a few points: first 3 to 10 hours after publication (80 minutes for one post), then 3 to 4 days later. The exact moment the reactions arrived is not visible. Four posts have a single snapshot.

  • Reaction settings. Channel admins can switch reactions on and off. 😁 is off on @truexanewsua's Krakow posts and may have been off on @insiderUKR for part of the day (1 to 23 per post). A missing reaction may reflect a setting and say nothing about readers.

  • 🔥 as a stand-in. Where 😁 is unavailable, 🔥 is read as gloating, because these audiences use it to celebrate strikes on Russia: 19 to 29 per 1,000 views on posts about destroyed Russian missile launchers that day, against about 0.2 to 0.6 on a typical post. It is an imperfect substitute.

  • Thumbs excluded. 👍 and 👎 are ambiguous: 👍 under an arrest post approves the arrest. On Krakow posts 👍 stayed at or below each channel's usual level, apart from small rises on arrest posts. 👎 stayed near zero, except on the post covered in section 4.

  • Comparable audiences. The between-channel comparison assumes that the audiences of these large general-news channels are broadly comparable.

  • Comments. Comments under the posts were collected and reviewed. They did not bear on the behaviour of the reaction counts and are outside the scope of this report. 

  • Public mood. The report assesses how specific reaction counts behaved. It makes no assessment of what Ukrainian audiences think or feel about the attack.

  • No attribution. The spiking reactions do not behave like reactions that accumulate with readership. The data does not show who added them or why. A batch on a post does not implicate the channel that published it, since reactions can be bought for any public post.

Every Vantage detection reports the originating post, the network amplifying it, the techniques in use, and how far it has spread, with the evidence behind each claim attached.

Every Vantage detection reports the originating post, the network amplifying it, the techniques in use, and how far it has spread, with the evidence behind each claim attached.

Every Vantage detection reports the originating post, the network amplifying it, the techniques in use, and how far it has spread, with the evidence behind each claim attached.

Get the LetsData newsletter.

Short reads from our team on narratives, incidents and what they mean for security teams.

© LetsData 2026

Get the LetsData newsletter.

Short reads from our team on narratives, incidents and what they mean for security teams.

© LetsData 2026

Get the LetsData newsletter.

Short reads from our team on narratives, incidents and what they mean for security teams.

© LetsData 2026