Between August 1 and August 31, 2026, an audience intelligence study profiled 104,439 people inside the conversation around Grand Theft Auto VI. The month contained a dated announcement, an eight day leak, eight days of silence from the publisher and a premiere on August 27. Across 209 posts on four channels, that conversation produced 19,358,195 likes and 361,950,063 views.
Volume of that size is easy to report and hard to read. It says nothing about who arrived, what they felt on any given day or whether any of that attention turned into the deliberate actions a launch depends on. This study answers those questions with audience intelligence, using Felton Audiences to profile the active audience day by day through a month the brand did not control.
The study was produced independently, without the involvement of Rockstar Games or Take-Two. It measures effect and takes no position on whether the leak was a security failure or a deliberate act. The full report is available as a free download.
The dataset at a glance
The study covered 31 days and four channels: three owned by the publisher and the largest third-party leak account in the category. Felton Audiences profiles the people who comment, a population distinct from the people who view or follow. 209 posts were in scope and ten posts covering adjacent titles were excluded, accounting for 1.4% of the month’s likes.
Trust and emotional composition were measured where comment-level data was available. The remaining channels were measured in reach and in economic value. The study rests on a sample of the conversation rather than every comment published in August.
August 6: a dated announcement sets the month up
On August 6 the publisher dated its Extended Look for August 27, opening a 21 day runway. That single day produced 2,715,343 likes (110.7 times a normal day in the same dataset) and $1,999,284 in True Audience Value (TAV), 177 times a normal day.
It also produced 11,384 newly profiled people (16.8 times the daily norm) and pushed the Community Trust Score (CTS) to 6.0, a little over half a point above the month’s quiet baseline of 5.44. Every phase that followed is measured against those numbers.
Eight days of leaks, measured against a normal day
On August 18, day 12 of the 21, a third-party account published gameplay footage and a complete map. Take-Two began removals the same day, which confirmed the material as genuine. The leak account’s posting activity moved from 58 to 1,284 in 48 hours. The publisher said nothing for eight days.
Across that window, August 18 to August 25, the daily averages against a normal day looked like this:
| Metric | Normal day | Leak window (per day) | Multiple |
|---|---|---|---|
| TAV | $11,293 | $19,466 | 1.7x |
| Likes | 24,533 | 59,060 | 2.4x |
| Comments | 879 | 1,773 | 2.0x |
| Newly profiled people | 676 | 1,020 | 1.5x |
| High intent actions | 282 | 118 | 0.4x |
Volume rose on every line in that table. High intent actions were the single exception, running at 0.4 times a normal day while everything around them climbed.
The leak window produced 0.3% of the month’s high intent actions
High intent actions are the deliberate ones: saving a post, resharing it, quoting it. August produced 330,690 of them.
The eight leak days produced 948, which is 0.3% of the month. The six days of official content that followed, August 26 to August 31, produced 243,214, which is 73.5%.
0.3%
Eight days of leaks
August 18 to 25, the most visible phase of the month
73.5%
Six days of official content
August 26 to 31, after the publisher spoke
Trust reached its floor five days into the silence
Trust sat at 5.44 through the quiet build and 6.0 on announcement day. When the leak broke on August 18 it was still holding at 5.3. It kept falling while nothing was said, reaching 4.5 on August 23, five days into the silence. Negative sentiment moved from 17.3% before the leak to 39.5% on the same day. Toxicity ran at 5.71% across the month, roughly ten times a normal brand baseline.
The recovery divides cleanly between the two things the publisher did. The statement on August 26 bought 0.3 of a point. The content that followed bought 1.9. Emotion AI tracks which of 25+ emotional states an audience expresses each day, which is what makes that split measurable at all.
Community Trust Score, August 2026
1 to 31 August 2026, nine readings
The day the silence broke
On August 26, 24 hours before the premiere, the publisher spoke. That day carried 45,676 comments (26.9% of the entire month’s conversation) and 36,546 newly profiled people, 52 times the day before.
| Aug 23 (floor) | Aug 26 (statement) | Aug 28 (peak) | |
|---|---|---|---|
| CTS | 4.5 | 5.1 | 7.0 |
| Negative sentiment | 39.5% | 32.0% | 13.7% |
| Newly profiled people | 1,065 | 36,546 | 10,930 |
After the premiere, trust reached 6.8 and then 7.0, the highest reading of the month and a full point above anything recorded before the leak. August closed 0.51 of a point above the 5.44 baseline it started from. An independent market estimate put pre-orders on reveal day at 436% above the previous day, a figure neither the publisher nor Take-Two confirmed and one Felton did not produce.
The audience grew fastest on a channel the brand does not own
Inside the leak window the third-party account grew its active audience 12.6 times. The publisher’s own channel grew 1.1 times. In absolute terms that is 4,886 newly profiled people on the account the brand has no relationship with, against 1,692 on the one it controls.
Owned-channel analytics report the owned channel. The audience forming around the title during those eight days was largely assembling somewhere the brand cannot post, cannot moderate and cannot measure from inside its own dashboard.
Two audiences inside one launch
Audience Segmentation profiles each channel separately. The two channels did not hold the same people.
The audience on the official channel, 66,546 people, is 82.9% male, 43.0% Millennial and concentrated in the United States (32.1%), Brazil (15.5%) and Spain (5.9%). Its dominant emotional state is Excitement at 46.4% and its sentiment is 53.2% positive. Roughly 30% work in arts, entertainment, sports and multimedia.
The audience on the leak account, 8,215 people, is 91.6% male, 49.8% Gen Z, with 18 to 24 as its largest age bracket and the United States at 43.4%. Its dominant emotional state is Neutral at 49.4%, followed by Amusement at 23.1%. Its sentiment is 31.8% positive. Under 5% work in the same creative industries.
| Official channel | Leak account | |
|---|---|---|
| Dominant generation | Millennials 43.0% | Gen Z 49.8% |
| Aged 18 to 24 | 38.2% | 47.5% |
| Male | 82.9% | 91.6% |
| Dominant emotion | Excitement 46.4% | Neutral 49.4% |
| Positive sentiment | 53.2% | 31.8% |
| Brazil | 15.5% | 5.4% |
| Growth in the leak window | 1.1x | 12.6x |
What the audience carries into the conversation
Brand Detection profiles the brands an active audience already engages with in its own public activity.
Apple leads at 33.0% of the profiled segment, followed by Nike at 28.7%, BMW at 6.8%, Samsung at 5.8%, Ray-Ban at 5.3%, PlayStation at 5.0% and New Era at 4.6%. The most frequently detected items across profiles are t-shirts (4,256), caps (2,439), cars (2,263) and iPhones (2,252).
Brand affinity inside the profiled audience Share of the segment that already engages with each brand
What audience intelligence adds to a crisis
None of the findings above are visible from reach. Impressions across August would have produced one enormous flat line with the leak window near its top, ranking the least converting phase of the month as its most successful.
Audience intelligence describes who is inside the crowd, what they feel and what that attention is worth. Across this month it surfaced three things a volume dashboard could not: the leak converted almost nothing, the silence cost more in trust than the incident itself and the audience was growing fastest on a channel the brand does not own.
The full report is available as a free download on felton.ai.
