Home / Review
Detective Sherlock: Shadow Stalker Review
A living Holmes cast, buried under bugs and rough edges

Detective Sherlock: Shadow Stalker: 88% positive from 476 Steam reviews, 24-hour peak of 2 players, priced at $19.99, SteamGame.ai potential score 56/100 (data as of 2026-11-19). Source: SteamGame.ai (Find the next hit before it hits.)
Pros
- Typed questioning makes characters feel like they think and remember
- The mental lock and evidence loop delivers real interrogation pressure
- Faithful reworking of classic Holmes material
Cons
- Crashes, lost saves and quest triggers that break progression
- Free-form input still gates progress on exact phrasing and synthesized clues
- Token purchases on top of the game price
Detective Sherlock: Shadow Stalker hands you a 3D Baker Street where nobody offers a dialogue wheel. You type your questions, press E on props, and the cast answers back in its own words. Players who stick with it describe a genuine shift in how a detective game can feel: the AI remembers earlier remarks and stays in character, and a few reviewers even report developing a rapport with Watson. The problems sit alongside that hook. Crash bugs and dead-end quests show up across several post-launch reviews, and a separate token system for the conversation model keeps surfacing in negative ones. The store labels the game Early Access, and the ending is still being built.
The typing hook holds up
Where the game earns its keep is in the interrogation chair. One 10-hour reviewer walks through the loop plainly: move with WASD, crouch, jump, collect evidence, then open the clue menu, question yourself, and present the right item to crack what the game calls a mental lock. Multiple players say the mental lock plus evidence presentation creates real pressure, and that characters scheme, deflect and lie rather than reciting lines. A 20-hour player posted a one-line celebration about ending up engaged to Watson. Another reviewer notes that the model can be swapped in the settings between providers, and that the same NPC reacts differently depending on which one runs. The freedom is narrower than the store text implies, though. Reviewers with hours in the double digits describe typing variation after variation of the same question and getting nothing until the exact phrasing lands, which may make the tool feel far more scripted than it looks.
The moment Stapleton starts in on biology, my head kind of spins — but you can chat insects and plants with him like it's nothing. Your knowledge really runs deep.
The clue system compounds that friction. Four post-launch reviewers complain that clues pile up faster than you can use them, that a second tier has to be synthesized from the first, and that the game rarely signals which one belongs in a given confrontation. One player describes brute-forcing the deduction step out of frustration. Others report the opposite failure: they work out the culprit early and the game refuses to notice. Because the model generates responses, it can also invent details that contradict what another character said, and two reviewers flag that inconsistency as a direct hit on case logic. A detective story lives on the reader trusting the testimony, and this one asks you to cross-check it.

Technical trouble dominates the complaint list. Seven reviews after launch describe being sealed inside a room with no exit, quest triggers firing before they should, broken quest givers, lost saves and crashes that wipe progress; the newest of those is from October. One player stuck in a mine says the community's own deduction help did not resolve it. Another refunded within minutes after the server stopped responding four reloads in a row. There is a real fix on record too. The developers added a map in an update, and a player who had been getting lost hunting for NPCs says navigation improved noticeably.
Visuals run hot and cold. The store trailer sets an expectation that five reviewers say the build does not meet; the harshest, from November, calls the models and facial animation unnaturally stiff and asks for a refund. One player, by contrast, says the character modeling is fine for them. The generated voice work gets two complaints for flat delivery and odd noises. On payment, the token packs sit outside the $19.99 asking price and appear in four negative reviews, including a player who says the allowance never comes back. The developer's reply argues that main-story play may need no extra tokens and that heavy free-form chatting is what burns them, which is a fair distinction, though the meter is still there and still visible.
Very buggy and constantly soft locks you if you phrase a question wrong. I just don't understand why devs can't add a skip dialogue button to progress the story or a cheat button to automatically unlock hints and locks?
The adaptation underneath the tech has fans. Three players praise how closely the Baskerville material follows the original, and one recalls spotting cobwebs during a housekeeper's story about cleaning, then using it against her. As for length, the sample runs from people quitting at 20 minutes to one at 31 hours, so the honest answer is that the ceiling depends on how much you enjoy talking.

This review is based on 73 Steam player reviews and store data. See all reviews on Steam →
Quick answers
Do I need to buy tokens to finish the story?
The developer says a player who mainly investigates and pushes the main story may get through available content without buying extra tokens, and that heavy free-form chatting is what drains them. Four negative reviews raise the cost anyway, including one player who says spent tokens never return. That is enough for several players reporting it, not a full picture of how the meter behaves.
Can I get stuck and have no way forward?
Yes, and several players have. Reviews after launch describe a character sealed in a cellar, a cursor that will not register a visible clue, and a mine sequence one player could not clear even with community deduction help. A few players ask for a hint or skip option as a fix.
Does the early access build include the ending?
The store page states the ending is still in development, and one post-launch reviewer describes the story simply stopping. The same complaint does not appear in the later reviews sampled, so the state of the finale at the time of writing is not something players here confirm.
By the NumbersUpdated 2026-11-19. Data updates on a daily schedule.
- Potential score
- 56 list metric, not a quality score
- Positive (all languages)
- 88% Very Positive
- Total reviews (all languages)
- 476
- New reviews/day (7d)
- 0
- Peak players (24h)
- 2
- Price
- $19.99
- Release date
- Aug 21, 2026
- Developer
- Gamercury
- Publisher
- Gamercury
- Est. sales
- 14.3K~28.6K rough estimate from review count
Source: public Steam store pages and player reviews. Review counts, positive share and daily new reviews cover all languages; the player peak is the highest count in the last 24 hours; estimated sales are a rough estimate based on review count, for reference only. Updated 2026-11-19; data updates on a daily schedule. Methodology
Daily New Reviews (all languages)
Is interest rising or fading?
Developer View
- Developer
- Gamercury
- Publisher
- Gamercury
- Developer's games on Steam
- 1 incl. this one; matched by developer name, may be off
- Early Access
- In Early Access
The developer discloses on the Steam store page that this game uses AI-generated content.
Read the developer's disclosure
The in-game real-time dialogue text and voice are generated by AI. All code and 3D art assets are handcrafted by the development team. The AI model has been fine-tuned with multiple safety measures — including prompt restrictions, keyword filtering, and compliance safeguards — to ensure all generated content remains appropriate.
If you encounter any inappropriate or non-compliant content during gameplay, please report it to our support team, and we will promptly review and make the necessary adjustments.
Thank you for your understanding!
The disclosure is the developer's own wording from the Steam store page. More new games with an AI disclosure →



