The change is visible in the volume, the language, and the response. By 2025, posts classified as hostile made up one in five of Rowling’s trans-related tweets. The subject appeared on most days when the archive records her posting.
These findings describe the collected tweets and their corrected model classifications. They are not estimates of a complete account history. Definitions and calculations.
01 · Change over time
Hostile posts rose 56% in 2025
The archive contains 1,194 trans-related posts in 2024 and 1,011 in 2025, a fall of 15.3%. Over the same period, posts coded hostile rose from 134 to 209. Their share nearly doubled, from 11.2% to 20.7%.
The increase appears in both original posts and replies. Including mocking posts brings the 2025 total to 41.5% of trans-related tweets. The partial 2026 sample has a lower hostile share, 13.1%; the record does not show an uninterrupted rise.
Select a coloured segment to read the tweets. The counts beside each year are the denominators. 2026 is partial; 2020–21 have small samples.
| Post type | 2024 | 2025 |
|---|---|---|
| All trans-related posts | 11.2% (134/1194) | 20.7% (209/1011) |
| Originals | 12.9% (61/474) | 22.9% (100/436) |
| Replies | 10.1% (73/720) | 19.0% (109/575) |
02 · The language used
Misgendering labels rose to 42.0%
In 2024, 341 of 1,194 trans-related posts were marked for misgendering (28.6%). In 2025, that rose to 425 of 1,011 (42.0%). Both the count and the share increased.
Sexualized language is especially concentrated in replies: 81 of the 111 posts with that descriptor are replies. It is marked in 4.7% of trans-related replies, compared with 2.4% of originals—about twice the rate.
“Can we talk about the fetishists yet?” is one example of the sexualized framing. Read that post in full, or search every post using “fetish” and related words.
Misgendering figures by year
These are descriptor annotations, which may reflect quoted or reply context as well as the author’s text. They are not literal word counts. Read the replies marked for sexualization.
03 · A recurring accusation
33 posts call trans people fetishists
Rowling has used the word “fetishist”, or another word beginning “fetish”, in 41 posts in this archive. In 33 of them the word is applied to trans people, to trans women as a class, or to an individual she describes as a man dressing as a woman. Those posts have been liked 505,330 times.
The accusation is recent. The earliest post that applies the word to trans people is dated 1 July 2024; nothing collected before it does, over an archive that starts in 2009. There were 9 in 2024, 19 in 2025 and 5 so far in 2026. It appears far more often in replies than in her own posts: 24 of the 33 are replies to other accounts.
As a share of everything she writes about trans people it stays small, moving from 0.8% of trans-related posts in 2024 to 1.9% in 2025. The significance is not the frequency. It is that a specific sexual allegation, which no evidence in the record supports, is made repeatedly about a group of people and endorsed half a million times.
| Year | Posts | Trans-related posts | Share | Likes |
|---|---|---|---|---|
| 2024 | 9 | 1,194 | 0.8% | 70,192 |
| 2025 | 19 | 1,011 | 1.9% | 410,178 |
| 2026* | 5 | 358 | 1.4% | 24,960 |
Counted from her own text only, never from a quoted or reply context. Whether a use is aimed at trans people is an editorial reading, recorded post by post; the 8 excluded uses include 2 where she repeats the phrase “genital fetishist” in order to reject it as a smear against lesbians, and 3 non-sexual uses such as “fetishise victimhood”. 4 of the 33 quote or parody someone else, which she endorses. How this was counted. Read every post using the word.
04 · Where hostility is concentrated
Sport has the highest hostile share
Across the full collection, 74 of 208 sport-related posts are coded hostile (35.6%), the highest share among topic categories. Single-sex spaces has the largest hostile count: 275 posts, out of 1,302 in that category (21.1%).
Those are different findings: a topic can produce the largest number of hostile posts because it is discussed much more often. The bars below compare the share within each topic.
Hostile share by topic
| Topic | Trans-related posts | Hostile | Share |
|---|---|---|---|
| Sport | 208 | 74 | 35.6% |
| Child safeguarding | 810 | 198 | 24.4% |
| Single-sex spaces | 1,302 | 275 | 21.1% |
| Specific individuals | 1,150 | 211 | 18.3% |
| Language & definitions | 1,203 | 174 | 14.5% |
| Youth healthcare | 354 | 42 | 11.9% |
| Other | 193 | 21 | 10.9% |
| Media & institutions | 644 | 67 | 10.4% |
| Free speech & cancellation | 721 | 73 | 10.1% |
| Law & gender recognition | 326 | 33 | 10.1% |
| Adult healthcare | 87 | 5 | 5.7% |
Topics overlap; a single tweet can appear in several rows. Comparisons do not hold year, post type, or specific events constant. A trans-related label does not establish the identity of a person discussed. Select a row to read its hostile tweets.
05 · The response
Hostile originals drew more likes in every quarter of 2024–25
The median hostile original received more likes than the median of other trans-related originals in all 8 quarters. Comparing within quarters reduces the distortion from comparing older and newer posts. Separating originals from replies avoids mixing two very different engagement scales.
Across 2025, the medians were 35,247 likes for 100 hostile originals and 22,360 for 336 other trans-related originals, a ratio of 1.58 to one. Medians limit the influence of a few exceptionally viral posts.
Quarterly medians and sample sizes
Each comparison requires at least five posts with recorded likes in both groups. “Other” means all other stance labels among trans-related originals. This is an association: the archive cannot establish whether language, topic, timing, reach, or another factor caused the difference. Likes do not measure agreement.
All figures
Stance and topic totals, and the original engagement study.
Stance
One label per trans-related tweet.
Stance definitions. These are model interpretations, not individual human judgments.
Topics
A tweet can have more than one topic.
Terms used for trans people
Tweets containing at least one descriptor in each category.
Rhetorical devices
Model annotations; multiple devices can occur in a tweet.
Likes over time · original study
Median likes per original tweet. Replies and retweets are excluded. This older series ends in August 2026 and retains its original collection-time counts.
Median likes rose from 2,009 in the year before 27 October 2022 to 8,799 in the year after, a 4.38-fold increase. This does not establish whether changes in platform ownership, follower count, or the content of her tweets caused the increase.
Quarterly figures
06 · A recurring subject
Trans-related posts on 234 days in 2025
There are collected posts on 273 different days in 2025. 85.7% of those days include at least one trans-related post. The ten busiest trans-related days account for 16.7% of the year’s total; most posts are spread across the rest of the year.
This is a sustained pattern across the collected timeline. Empty dates in the calendar mean no matching tweets were collected, not proof that nothing was posted.
Select a date to open its tweets. The comparison uses observed posting days, not all calendar days.