Backseat
analyzed 2026-10-13 · complete count of 72 English reviews
Across all 72 reviews, atmosphere and story are the most consistently praised aspects, each mentioned in 31.9% of reviews. Atmosphere draws only positive mentions, with players describing how the game "creates an incredibly tense and genuinely frightening atmosphere". Story is nearly as unified, at 18 positive against 3 negative, and one player notes "The story managed to draw me in from the very beginning". Gameplay appears in 33.3% of reviews but splits more evenly: 14 positive, 7 negative, 3 mixed. Player choice, cited in 15.3% of reviews, leans positive, with players pointing to how "the multiple endings make the choices you make feel meaningful". Replayability comes up less often and divides, though one player writes "very replayable game with multiple endings". Criticism clusters around length and technical problems. Game length, mentioned in 11.1% of reviews, draws no positive mentions — 4 negative and 4 neutral — with complaints that "the games is a little long" alongside the opposite view that "playing through the game once feels like a demo". Bugs appear in 9.7% of reviews, all negative, including a softlock report. Performance, also 9.7%, is mixed at 3 positive against 4 negative.
numbers checked against this report's stored data · quotes copied verbatim from reviews · plain text is the language model's own wording
Aspects
share of sample, axis to 40% · exact counts, no margin of error
- positive
- mixed / neutral
- negative
-
gameplay
14 positive · 7 negative · 3 mixed · 0 neutral
-
Trying to solve mysteries while under pressure, all while driving a car, creates an incredibly tense and genuinely frightening atmosphere.
-
The game is incredibly tense; you encounter a wide variety of criminals, and hanging between life and death until you identify all of them makes the game insanely thrilling.
-
amazing concept and great execution!
-
-
atmosphere
23 positive · 0 negative · 0 mixed · 0 neutral
-
Man, the concept and atmosphere are so cool.
-
Trying to solve mysteries while under pressure, all while driving a car, creates an incredibly tense and genuinely frightening atmosphere.
-
Simple idea, great atmosphere, and surprisingly tense.
-
-
story
18 positive · 3 negative · 1 mixed · 1 neutral
-
I started out trying to figure out the criminals in the car, and suddenly found myself right in the middle of a whole story.
-
The story managed to draw me in from the very beginning, and I think it offers a different experience for horror game fans.
-
got 2 of the endings and had a blast doing it!
-
-
player choice
9 positive · 1 negative · 1 mixed · 0 neutral
-
The story is really interesting, and the multiple endings make the choices you make feel meaningful.
-
It’s not a very long game, but the choices and different outcomes made it fun to replay.
-
Many possible outcomes, and I have to say it does have a good story.
-
-
game length
0 positive · 4 negative · 0 mixed · 4 neutral
-
this is a good game didn't get board the only thing is the games is a little long
-
… a positive review as made by a smaller developer but its too stale and more story focused for me and not what i expected playing through the game once feels like a demo tbh u can replay for different ending but theres nothing to make it not feel tedious to do this this game is basically a …
-
An absolute hidden gem with incredible storytelling and atmosphere, held back only by its runtime. At just 2 hours long, it ends right as it's hitting its stride.
-
-
bugs
0 positive · 7 negative · 0 mixed · 0 neutral
-
It has some rough edges, but the core idea is genuinely fun and different.
-
Parked at objective waiting for one of the criminals to get in the car and nothing happened, and i was in a softlocked state.
-
This game is so buggy and there are some things that dont make sense.
-
-
performance
3 positive · 4 negative · 0 mixed · 0 neutral
-
Well, general experience was not bad but I have some performance issues
-
It's horribly optimized, and honestly it seems like a cash grab scam.
-
and it need a litte bit more optimized it goes so bad and i have a RTX 5060
-
-
replayability
3 positive · 2 negative · 0 mixed · 0 neutral
-
It kept me hooked and curious about how it would end, and now I’m really interested to see what the other endings are like.
-
very replayable game with multiple endings even in this price range.
-
Thankfully, the branching paths and alternate storylines add much-needed replay value.
-
Review timeline
weekly peak 81 reviews · all languages
weekly share rated positive · 50% reference · weeks with <5 reviews faded
the timeline as a table
| week | reviews | up | down | positive | note |
|---|---|---|---|---|---|
| week of 09 Sep 2026 | 81 | 56 | 25 | 69% | |
| week of 16 Sep 2026 | 25 | 16 | 9 | 64% | |
| week of 23 Sep 2026 | 6 | 4 | 2 | 67% | |
| week of 30 Sep 2026 | 4 | 3 | 1 | 75% | small sample |
| week of 07 Oct 2026 | 2 | 2 | 0 | 100% | small sample |
How this report was made
- Population
- all Steam reviews of the game (every language, purchase type, and verdict), walked newest-first per window; only English-language reviews are analyzed
- Sample
- complete count: every usable English review (72), fetched whole-life
- Aspect yield
- 61 of 72 analyzed reviews carry at least one aspect (84.7%); every share on this page divides by all 72
- Interval regime
- exact counts, no margins of error to calibrate
- Fetch paths
- 1 time window fetched directly (windowed)
- Language mix
- 124 fetched: english 72 · turkish 12 · russian 10 · 12 more
- Steam-marked windows
- none flagged by Steam (Valve's off-topic flagging began March 2019)
- Marked share
- 0.0% of sample inside marked windows, under the 2% floor
- Narrative
- passed the grounding gate first try
- Evidence floor
- aspects and emergent themes appear only with 5+ supporting reviews; smaller counts sit at the classifier's false-positive floor
- Instrument readings
- measured once on the study corpus for this model and prompt: frozen calibration, not properties of this run
- · aspect tagging vs. human labels
- 0.801 [0.752–0.844] (F1 against a human-labeled test set; 1.0 is best)
- · quotes filed under the wrong aspect
- 11.6% [6.6–19.6] of quotes, measured by hand on a sample; lower is better — measured on the prior model (2026-08-05); re-measurement pending
- · independent model vs. production labels
- 0.791 [0.772–0.810] (F1 of the production labeler against an independent judge model) — measured on the prior model (2026-07-25); re-measurement pending
- Versions
- deepseek-flash · prompt classify-v1 · ontology v2
- Run
- serve-20261013T155349Z-4d0f69ec · code 798e06c
full methodology in the milestone write-ups