The dataset viewer is not available for this split.
Error code: FeaturesError
Exception: ArrowInvalid
Message: Schema at index 1 was different:
date: string
displayed_score: double
total_ratings: int64
developer: string
note: string
vs
year: int64
mean_star_english: double
n_english: double
mean_star_all_storefronts: double
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/split/first_rows.py", line 249, in compute_first_rows_from_streaming_response
iterable_dataset = iterable_dataset._resolve_features()
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 4379, in _resolve_features
features = _infer_features_from_batch(self.with_format(None)._head())
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2661, in _head
return next(iter(self.iter(batch_size=n)))
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2839, in iter
for key, pa_table in ex_iterable.iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2377, in _iter_arrow
yield from self.ex_iterable._iter_arrow()
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 580, in _iter_arrow
yield new_key, pa.Table.from_batches(chunks_buffer)
~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^
File "pyarrow/table.pxi", line 5039, in pyarrow.lib.Table.from_batches
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
return check_status(status)
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: Schema at index 1 was different:
date: string
displayed_score: double
total_ratings: int64
developer: string
note: string
vs
year: int64
mean_star_english: double
n_english: double
mean_star_all_storefronts: doubleNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
X (Twitter) app reviews — before and after the Musk acquisition
Aggregates over 92,088 reviews of the X (formerly Twitter) app — 49,723 from the iOS App Store (fanned across 93 country storefronts, reaching back to 2009) and 42,365 from Google Play — theme-coded and split around Elon Musk's acquisition (deal closed 2022-10-27, rebrand to X 2023-07-23). Snapshot 2026-07-16.
Companion to the study X's App Store Rating Hasn't Moved Since Musk Bought It. The Reviews Have.
Headline findings
- The displayed App Store rating never moved: 4.6 stars throughout — 6.6M ratings the week the deal closed (Oct 2022), 8.6M by mid-2024, 11M today. The all-time aggregate is designed not to move.
- Written-review sentiment fell sharply. Across 10,861 English App Store reviews, the mean star of people who wrote text dropped from 3.81 before the acquisition to 2.95 in the X era; the one-star share of written reviews rose from 17.8% to 43.6%. The break lands in 2023, the first full year under new ownership.
- The topics shifted more than the scores. Mentions of Elon Musk went from 0.8% to 16.4% of reviews — the biggest mover. Subscription/pricing complaints tripled (3.2% → 9.2%), a Grok/AI theme rose from near-zero to 4.6%, and free-speech praise more than doubled (3.2% → 7.4%).
- The narrative the data does NOT support: censorship and bias complaints did not rise — 7.7% before, 6.6% after. Moderation grievances predate Musk.
Files
| File | Contents |
|---|---|
data/summary.json |
Full aggregate: corpus sizes, theme before/after, star-by-year, rating snapshots |
data/theme-before-after.csv |
Share of English reviews mentioning each of 14 themes, before vs. after, with counts |
data/star-by-year.csv |
Mean written-review star by year (English subset + all 78 storefronts) |
data/rating-snapshots.csv |
Displayed App Store rating / total ratings / developer at three points in time |
Methodology
Reviews were pulled via the Crawlora web-data API on 2026-07-16. Apple caps its public review feed at 10 pages × 2 sort orders per country storefront (~1,000 max), but each storefront is an independent feed reaching back years, so fanning across 93 storefronts yields a large, historically deep corpus with genuine pre-acquisition reviews. Google Play was collected via continuation-token deep pagination. Each English-storefront review was theme-coded with a documented keyword classifier across 14 categories; a review may carry more than one theme. Rating snapshots were read from the Internet Archive and the live listing.
Caveats
- Written reviews are a self-selected, opinion-heavy minority — not users. The App Store's "most helpful" sort over-samples heavily-upvoted (often critical) reviews. Read the shifts in what this vocal group writes, not the absolute levels as a poll of all users. This is why the displayed aggregate stays 4.6 while written-review sentiment drops.
- Theme coding is keyword-based, counts a mention, not a verified fact. "Reviewers mention X more" is not "X is more true."
- The corpus spans ~78 storefronts and many languages. Star ratings are language-independent (used for the trajectory); theme coding runs only on the 10,861 English-majority-storefront reviews.
- Google Play cannot cover the pre-2022 period at scale, so the before/after rests on the App Store.
- This is a composition-shift measurement, not a verdict on whether X improved or declined.
License & citation
Released under CC BY 4.0 — use it for anything, just attribute.
Crawlora (2026). X (Twitter) app reviews before and after the Musk acquisition. https://github.com/Crawlora-org/x-app-reviews-data (CC BY 4.0)
See CITATION.cff for machine-readable citation metadata.
- Downloads last month
- 25