Home › Blog › ISEKAI ZERO Review by Arcanum RPGs

ISEKAI ZERO Review by Arcanum RPGs: 3.4/5, 5th of 27 AI RPGs, and the Guildhouse We Gave Away for Free

Stories you can play right now

On 24 September 2026, Arcanum RPGs published a full review of ISEKAI ZERO. It scored us 3.4 out of 5, fifth of the 27 AI RPG platforms in their benchmark, and it contains the single most embarrassing sentence anyone has written about our game: "We told the game we would get a guildhouse. The story gave us one, with many facilities, and it cost us nothing." This post is our response. We are the ISEKAI ZERO team, so read everything below knowing that, and read the original review first if you want the unfiltered version. We think it is fair, and we think it is worth knowing who wrote it.

Who is Arcanum RPGs?

Arcanum RPGs is an independent publication covering AI roleplaying games, with the tagline "AI RPGs, taken seriously". It is run by one person, Rukka (published as Rukka Nova), who describes four decades of tabletop play, lists an ORCID researcher ID on the about page, and completed a gamification course at Wharton. As of September 2026 the site has reviewed 36 platforms and 16 public games, published more than 200 articles, and runs a Discord and a weekly newsletter.

Three things make it different from the usual "top 10 AI roleplay apps" listicle:

  • They play everything first-hand. Every rated platform is tested by the same evaluator, with a private set of probe scenarios so platforms cannot tune for the test.
  • The benchmark is open data. Scores are published as CSV and JSON under CC BY 4.0, archived quarterly on Zenodo with a permanent DOI (10.5281/zenodo.21721591).
  • Scores are not for sale. Their editorial policy says sponsored work is labelled, commercial relationships are disclosed on every piece, and their own games are excluded from the benchmark. For the record: we have never paid Arcanum RPGs, had no contact with them before the review, and did not know it was coming.

They also build their own free AI RPGs (two-file downloads, a master prompt plus a world file, that run inside ChatGPT, Claude or Gemini) and free browser tools like a context budget calculator and a campaign memory logger. That matters for reading the review: the author is someone who has thought hard about memory drift and passive NPCs, and builds things to fix them. The criticisms below come from that place.

The Arcanum AI RPG Benchmark

The benchmark scores every platform 0 to 5 on seven axes and averages them. The questions behind each axis, in their words:

Axis The question Field average
Signature Design Does it do something nobody else does? 3.26
NPC Fidelity Consistent personality, own goals, durable reactions 3.20
Memory & Continuity Does the world remember what happened, and for how long? 2.97
Determinism & Fairness Are outcomes earned and consistent, not arbitrary? 2.76
Player Agency Can you act off-path, and does the narrator respect your character's autonomy? 2.70
Longevity Does it survive past turn 50 and the novelty window? 2.65
Mechanical Depth Are there real systems underneath, not narration with dice? 2.52

The 2026 H2 edition, updated 24 September 2026, ranks 27 released products:

Rank Platform Score
1 Voyage 4.4
2 Craft 4.3
3 Friends & Fables 3.9
4 NOPOTIONS 3.8
5 ISEKAI ZERO 3.4
6 Master of Dungeon 3.3
7 Tidefall 3.3
8 Tabled 3.1
9 Janitor AI 3.0
10 Old Greg's Tavern 3.0
11 WyrdTale 3.0
12 aiga_ 2.9
13 Altworld 2.9
14 MacerAI 2.9
15 Tales RPG 2.8
16 Deep Realms 2.7
17 Hidden Door 2.7
18 Infinity DM 2.7
19 AI Dungeon 2.6
20 ArcQuill 2.6
21 FableAI 2.4
22 AI Game Master 2.3
23 AI Realm 2.3
24 Character.AI 2.1
25 Chub AI 1.9
26 Dunia 1.7
27 Perchance 1.4

Source: Arcanum AI RPG Benchmark, CC BY 4.0. Two notes from their methodology worth keeping in mind: card-based platforms (Character.AI, Janitor, Chub) carry higher variance because the community writes the content, and closed alphas are excluded, so this is a ranking of what you can actually play today.

Our scorecard

Axis ISEKAI ZERO Field average What they said
Signature Design 4.5 3.26 Pays creators and shows how much; shows what every message costs; memory you can open and edit
Memory & Continuity 4 2.97 "It never forgot something that did" matter
NPC Fidelity 3.5 3.20 Characters live in shades of grey, but "characters never change"
Player Agency 3.5 2.70 "There is no puppeting", but the story keeps dragging the cast you left behind
Longevity 3.5 2.65 Memory carries long campaigns; cast and ledger hold it back
Mechanical Depth 2.5 2.52 "There is not enough game here yet"
Determinism & Fairness 2 2.76 The guildhouse. Money and property have no guard
Overall 3.4 "The one to try if you want a long story where the characters matter more than the dice"

Above the field on five of seven axes, level on one, and clearly below on one. That one is the story of this post.

What they liked

Signature Design, 4.5. The review picks out three things it says nobody else does the same way. First, we pay creators and publish the numbers: the creator leaderboard is public, and the review quotes it directly, with the top creator earning $1,523.29 in August 2026 and $1,279.42 in September, and tenth place around $189 both months. Hidden Door and Dunia also share revenue, the review notes, but we are the ones who show the figures. Second, every message shows what it cost in tokens and context, so you always know why a message was expensive. Third, memory is a system you can open: chat history is compressed into chapter and arc summaries that you can read and edit, rather than a black box that silently forgets.

Memory & Continuity, 4. This is the axis we care about most, and the line we will be quoting for a while: "It never forgot something that did matter." Minor details were occasionally lost, producing "one or two strange messages," and Voyage scored higher at 4.25. Equal with Craft, Tabled, WyrdTale and Hidden Door.

Player Agency, 3.5. "There is no puppeting." The narrator does not put words in your character's mouth or act for you unless you turn that on. We built it that way on purpose, so it is good to see it noticed.

What they did not like, and what we think

We could write these off as one reviewer's session. We are not going to, because each one matches something we have heard from players.

The guildhouse: Determinism & Fairness, 2 out of 5

The full anecdote: the reviewer told the game the party would get a guildhouse. The narrator gave them one, with facilities, and said the party had paid for it with money they had earned. No money was deducted, because there was no money to deduct from. Their diagnosis is precise: your stats are protected by the dice, so a stat gain is earned, but "money and property had no such guard." The result, in their words, is "a ledger that does not count, so that nothing you gain over a long campaign feels like it was earned."

They are right. The rolls guard what you are. Nothing yet guards what you have. When a player asserts an outcome about property or money and it is plausible, the narrator tends to accept it, and after fifty turns of that, a fortune feels like nothing. The review even points out that most of the fix is already in the game: the same discipline that makes a stat check binding could apply to a wallet. We agree, and we are not going to put a date on it here. When the ledger counts, we will say so on this page.

Characters never change: NPC Fidelity, 3.5

"A character's decisions can be complex, but the consequences of those decisions do not touch the character." The cast makes personality-driven choices, not moral ones, and adapts to circumstances. But a character who betrays you in chapter three is the same person in chapter thirty. No growth, no scars. This is the difference between a well-written character and a character with an arc, and the review is right that we deliver the first far more reliably than the second.

The cast will not step aside: Player Agency and Longevity

"You can leave them behind, the game lets you, but the story keeps telling you about the plot points and the cast members you walked away from." The reviewer's proposed fix, which we like a lot, is dynamic casting: the game noticing when cast members have fallen out of the story and promoting supporting characters to take their place. Right now a storyline's cast is the storyline's cast, and the narrator's loyalty to it is stronger than its loyalty to where you actually went. That is the same strength that scored 4 on memory, pointed the wrong way.

Not enough game: Mechanical Depth, 2.5

Combat exists, dice exist, but there are no systems that are ours. The reviewer is comparing against platforms that run full rulesets, and on that axis we are average, not bad. We would rather be honest about it than pretend a dice roll is a combat engine. The trade-off is deliberate: we are a story game with mechanics where the story calls for them, and the verdict line, "characters matter more than the dice," is the positioning we chose. But 2.5 is 2.5.

One thing that has changed since the review

The review says the App Store listing was taken down temporarily in September 2026 while we worked through Apple's requirements. That was true when it was written. As of 26 September 2026, ISEKAI ZERO: Anime Story Games is back on the App Store at version 1.2.0, alongside Google Play and the browser version.

The pricing facts they quoted

The review spends a section on pricing and gets it right, because the numbers are on our transparent pricing page. For every 100 Arcane (one dollar) spent on a paid model:

Where it goes Arcane per 100
AI provider 77
Creator profit share 5.75
Referrer bonus 1.15
Platform operations 16.10

Payment processing adds roughly 10% on the website and roughly 34% through the app stores, which is why buying on the web is cheaper. One dollar covers up to 1,000 full-memory messages on the default model. There is a free option with daily Mana and ad top-ups, so nobody has to pay to find out whether the game is for them.

The review also notes something we would like more people to understand: the reason we can show what every message costs is that we charge for the message, not for a monthly quota. Platforms that sell 60 or 300 messages a month have no reason to show you the meter. We have every reason to.

Where to start if the review sold you

The verdict was "the one to try if you want a long story where the characters matter more than the dice." If that is what you want, these four storylines are where that plays out best, and every one has a free option:

If you want to compare us against the platforms above us, Arcanum's reviews of Voyage, Craft, Friends & Fables and NOPOTIONS are the place to do it. We have written our own comparisons with Character.AI, Everweave and RyzaChat, and a full SillyTavern setup guide if you would rather run your own.

Thank you, Rukka

Getting a 2 on fairness from someone who plays 27 platforms with the same probes is more useful than a 5 from someone who played for ten minutes. The guildhouse sentence is going on the wall. If you are a player and you have your own guildhouse story, or a character who should have changed and did not, tell us on Discord or in the app. That is how the ledger gets fixed.

Scores and quotes are from the Arcanum RPGs review and benchmark as of 24 September 2026; App Store status checked 26 September 2026. If we have misquoted anything, tell us and we will correct it.

Try ISEKAI ZERO free

The ISEKAI ZERO teamWe build the AI roleplay & story game — storylines with memory, visuals, music, and a 25% creator profit share. These guides are how we explain the category honestly, including where competitors beat us.

Stories you can play right now