Reading 303 Entries to an AI Creator Award: Where the Work Actually Clusters

The International AI Creator Award 2026, run by AICU Inc. and operated by AICU Japan, opened for entries on 20 September. By 23:08 Japan time on 17 October it had 303 entries from 171 creators across 17 countries and regions, and every one of its ten categories had something in it. On the opening day the organisers logged 18 entries. Between 13 and 16 October they were logging 26 to 37 a day as the second-period deadline approached.

Award entry counts are not a census of anything. But the way those 303 entries distribute across categories is a decent proxy for where AI-assisted creative production has enough volume to be judged as a field rather than a demo reel. The organisers published the per-category breakdown, and the shape of it is more interesting than the total.

The distribution

Music, music video and dance led with 62 entries. Independent video followed at 59, then character, comic and webtoon work at 52. Publishing came in at 38, and games, tools and apps at 34. The long tail: advertising, promo and broadcast at 16, commercial film at 10, services at 7, events and XR at 7, and 18 in the open wildcard category.

Two things stand out. The first is that the top three buckets are the formats where generative tooling is most mature and the audience is already enormous. Video and music have had fast, cheap, iterable tools for a while, and both have distribution channels that do not require permission from anyone. Volume follows capability.

The second is that publishing and games rank fourth and fifth. Those are not prompt-to-output formats. A finished book or a playable build takes editing, structure, testing and completion, and it is the kind of artifact that a judge can actually consume end to end. Their position in the table is a quiet argument that the awards, and probably the wider audience, are rewarding finished things rather than impressive fragments. You cannot play a render.

The caveat matters as much as the numbers: 256 of the 303 entries came from Japan, followed by Korea at 9, Canada at 7, the United States at 6 and Hong Kong at 5. This is a picture of one very active market plus a thin international fringe, not a global ranking. Read it as "here is what creators in the most engaged market are shipping," and it gets more useful. The snapshot also moves fast: AICU's own notice page, published the same day, gives 302 works from 172 creators across 18 countries as more entries arrived, so treat any single total as a timestamp rather than a result. Second-period entries close on 19 October at 21:00 Japan time, and a third round runs to 26 October, with recommended entries free and late entries at $30 per work.

One primary category, cross-category judging

A rule worth stealing from the award's structure: each entry picks exactly one primary category, and nominations and voting happen in that category. Judges may still evaluate a work from the perspective of other categories, but that does not move the entry. If you picked wrong, you contact the office and they reassign the primary category rather than asking you to resubmit from scratch.

You can see the reasoning. A music video shot as a narrative short could reasonably sit in either bucket, and the honest answer to "which category is this" is often "whichever comparison you would rather be judged against." That is a useful question to ask before submitting anywhere, and it applies outside awards too. When you publish work, you are choosing the shelf it sits on, and the shelf decides what it gets compared to. One consequence worth knowing: the work number is reissued under the new category, so the entry's identifier changes along with the reassignment.

The work-number scheme is the quietly useful part

The organisers are moving from numeric category codes to letters. An entry like A26M-57 breaks into three parts: A26 for the 2026 award, M for music, and 57 for the sequence number within that category. The old format, A2605-57, said the same thing with digits, and the two formats map one to one, so existing numbers keep working.

The category letters are S for service, F for commercial film, V for independent video, C for character and comic, M for music, B for broadcast, P for publishing, G for games, tools and apps, E for events and XR, and W for the open wildcard.

The detail that shows someone thought about it: the letters I, L, O and U are never used. I and L are too close to the digit 1, O to 0, and U to V, so those combinations get misread when a number is typed into a form or read aloud over a call. The letters A and X are held back for categories that do not exist yet.

That is a small, cheap decision with a long payoff, and it is worth copying anywhere you name things that have to survive being copied by hand. Character banks, shot lists, asset libraries and client-facing file names all end up transcribed between systems and people, and the difference between an identifier that survives that trip and one that does not is a good chunk of your day. Reserving a few characters for the future costs nothing and saves a migration later.

What to take from an entry table

If you are choosing what to build next, the distribution suggests looking past the formats with the loudest output. Video and music are crowded because the tools are easy and the reach is instant. The categories with fewer entries and higher completion cost, publishing and playable builds, are where a finished artifact stands out from a sea of clips.

If you are entering anything, pick the category where your strongest work survives the comparison, not the one that matches the tool you happened to use. The award's own design assumes work can honestly belong in more than one place.

And if you are building a system to keep track of what you make, take the naming lesson seriously. Short, unambiguous, human-readable identifiers that leave room to grow will outlast whatever app you are storing them in. That is the same lesson that keeps showing up in how prizes for AI creative work are being written: the process and the structure around the output are what hold value, while a single nice render is just a render. The WFAIA contest in Tokyo went as far as requiring a workflow file with every video entry, and the reason is the same one visible in AICU's category codes. The field is old enough now to care about keeping records.