Home›Guides›Which apps admit they use AI

Which AI Drama Apps Admit They Use AI? All 23 Checked

Updated October 3, 2026 · Compiled by the AI Drama Reviews editorial team · 6 of 23 apps disclose AI in production · Every statement read on the company's own property on October 3, 2026

DramaBoxEditor's pick — biggest short-drama library, lowest price to go unlimited 9.8Excellent

• Thousands of series, roughly 200 new titles every month
• Free episodes daily, plus ad-unlocks — no card needed to start
• Unlimited plans from about $5.99 a week, the lowest entry price we found
• Runs on Android, iOS and in a desktop browser

Go to DramaBoxAffiliate link · free to download · in-app purchases available

Affiliate link. We may earn a commission if you install through it, at no extra cost to you — it never changes our scores or placements. How we test · Disclaimer

What this page records, and what it refuses to record

This page records one thing: what each company has itself published about AI being used to produce the shows in its catalogue. It is a register of statements, not a verdict on technology. A company that has published nothing appears here as a company that has published nothing, and nowhere on this page does that become a finding that the company does not use AI.

That rule is not pedantry. Over 95% of new Chinese microdramas were AI-generated in the first quarter of 2026, according to the China Netcasting Services Association, and most of what a US viewer watches on these apps is stocked from that pipeline. Silence in a storefront and absence of AI in a catalogue are, on the current evidence, almost unrelated things.

The two questions we keep apart

Every page on this site answers two separate questions in two separate columns, and never merges them. The first is disclosure: what the company has put in writing. The second is our own catalogue judgement, which is dated, revisable, and clearly ours. The first is checkable by anyone with a browser. The second is an editorial opinion formed by watching the catalogue, and it carries exactly as much weight as an editorial opinion should.

Most coverage of this category collapses the two. An app is called "an AI drama app" because its catalogue looks synthetic, or declared human because its marketing says Hollywood. Both moves substitute an impression for a record. Keeping the columns apart is the only way a reader can tell which of the two they are being given.

What counts as a statement here

A statement counts if the company published it on a property it controls: its own website, its own terms or privacy pages, a store listing it wrote, or its corporate newsroom. Executive interviews are recorded separately and labelled as interviews, because a quote given to a journalist is not something a company has chosen to publish and keep published.

That boundary explains two entries that would otherwise look inconsistent. FlareFlow counts as a disclosure because the statement sits in COL Group's own newsroom. ReelShort does not, even though its chief executive has been clearer about AI than any other executive in the category, because every one of those words was spoken to a reporter and none of them appears on ReelShort's site, in its terms, or in either store listing.

All twenty-three apps, in their own words

Here is the complete register, checked on October 3, 2026. Six companies disclose AI in production, two of them only in the product's name; one has an executive on the record in an interview and nothing on a page it controls; two claim human production; two mention AI only when recommending titles to you; and twelve have published nothing at all.

The entire public record, in six sentences

Strip out the categories and the caveats and the whole first-party record of AI in production across twenty-three companies is eight short statements, two of which are product names. Three were published in the last six weeks, which is the clearest sign that this is a live question rather than a settled one.

Which apps say their shows are AI-made?

Four, and two of them say it only in the product's name. SagaBox and ShortPeak say it in words; MoboDrama ships as "AI Comic Dramas" and VibeShort as "AI Short Dramas", and neither explains anywhere what the AI actually does. The two that use words:

SagaBox and ShortPeak. Both are browser-only products with small catalogues, both sit in the bottom third of our ranking, and between them they account for the only two unambiguous statements in the category that the content itself is generated.

SagaBox: one sentence on a homepage

SagaBox describes its own product, on its homepage, as "AI-generated vertical short films and comics". That is the whole disclosure. There is no policy page, no explanation of which tools, and no per-title marking. The About page does not mention AI at all, and the terms and privacy pages return 404.

It is worth being precise about what this is. One line of marketing copy describing the product is a genuine first-party statement and we count it as one, but it is not a disclosure regime. A reader who wants to know whether a particular SagaBox series was generated or filmed has nothing on the site to consult. The company that publishes it does not name itself anywhere we could find.

ShortPeak: the only disclosure that lives in a terms page

ShortPeak's terms of service carry a section headed "AI-Generated Content" which states that content "is produced using artificial intelligence generation tools, either in-house or through our production partners". The page was last updated on September 2, 2026. Nothing on the homepage, the about page or the support page says the same thing.

This is the most legally considered sentence any of the twenty-three has written, and it is also the one almost nobody will read. Putting a production disclosure in the terms rather than in the catalogue is a defensible choice for a lawyer and a poor one for a viewer. Note also the phrase "either in-house or through our production partners", which keeps the claim broad without committing to a share of the catalogue.

Which app labels individual AI titles?

One: StardustTV. A small badge reading "AI" sits above the thumbnail of individual titles in its catalogue, across the trending, new-release and exclusive rows. It is the only app of the twenty-three where a viewer can see, before pressing play, that this particular series was produced with AI and that one was not.

What StardustTV's badge tells you

The badge tells you that StardustTV has classified that title as AI-produced. Alongside it, the company describes the method in its own store listings: Apple's listing says "StardustTV uses AI-assisted production to create stories in new ways", and Google Play goes further, announcing "the brand-new AI-generated short drama section" built with "full-process creation and polishing".

Two different framings from one company in the same month is itself informative. Apple's version describes a technique applied across the service; Google's describes a distinct section of the catalogue. The badge on the site is consistent with the second reading: some titles carry it, most do not.

What it does not tell you

The badge is StardustTV's own classification, published to no standard and audited by nobody. There is no definition of what qualifies a title for it, no stated threshold, and no way to verify that an unbadged title was filmed. We verified the badges on StardustTV's website; we have not independently confirmed that they appear identically inside the iOS and Android apps.

Taken at face value it is still the most useful piece of consumer information in this category, and it costs StardustTV something to publish. That is worth saying plainly, and it is why StardustTV ranks fourth of twenty-three while the two apps that disclose most bluntly rank twentieth and fourteenth. Disclosure is one axis among six, not a substitute for catalogue, price or quality.

Which company discloses AI in the writing?

One: COL Group, the parent of FlareFlow. The disclosure concerns scripts rather than footage, and it appears in corporate rather than consumer communications, which makes it the most revealing entry in the register.

FlareFlow and COL Group's Xiaoyao model

COL Group's newsroom states that FlareFlow "leverages COL Group's proprietary large language model, Xiaoyao, to streamline script development and enhance creative quality". The Chinese original of the same release is more specific: the model, which it names 中文逍遥, is used for script outlines, character biographies and dialogue polishing at the storyboard stage. The release is dated August 24, 2025.

Read it carefully and it is narrower than it is often reported. COL says a language model assists script development. It does not say scripts are written by the model, and it says nothing at all about how the video is produced.

Why the parent discloses and the app stays silent

FlareFlow itself discloses nothing. Its site, its Apple listing and its Google Play listing contain no reference to AI; the Apple listing mentions only a "smart recommendation engine". The AI statement exists because a listed parent company was telling investors and the trade press what its technology does.

That split runs through the whole category. AI is an asset when the audience is an investor and a liability when the audience is a viewer, and the same company will happily say in one room what it omits in the other. Nothing obliges it to be consistent, which is the subject of the regulatory section below.

Has any app said it does not use AI?

One executive has, once, about scripts. No company in the twenty-three has published a denial on its own properties, and the distinction between an executive's answer in an interview and a company's published position matters more here than anywhere else on this page.

ReelShort: a denial in 2024, a different tone in 2026

Asked by TIME in November 2024 whether ReelShort's writing sounded AI-generated, founder and chief executive Joey Jia answered: "Oh, no. It's our in-house editors." In the same interview he said that if AI could write content and make money straight away, he would do everything with AI, and described an in-house editorial and screenwriting team managing the whole content process.

Nineteen months later, at the Vertical Media Summit in June 2026, the same executive said: "If I'm 100% a businessman, I would stop live-action today, because AI response for market is way better than live-action." The two statements are not strictly contradictory — one is about who writes the scripts, the other about which format performs — but anyone citing the 2024 denial as ReelShort's current position should know the 2026 remark exists. Neither appears anywhere on ReelShort's own site, which remains silent on AI in its terms, its privacy policy and both store listings.

Playlet markets human crews without denying AI

Playlet's store listings say its shows are "produced by Hollywood star teams" and its corporate site says "all content is handled exclusively by Playlet's in-house teams", naming filming crews in seven countries. It is the strongest human-production marketing in the category.

It is also not a denial. Nowhere in anything Playlet publishes does the company say it does not use AI, and a claim that filming is done in-house is compatible with AI in dubbing, in post, or in writing. We record Playlet as claiming human production and as having published nothing about AI, because both are true and they are different facts.

The AI claims that are not about production

Three apps use the word AI prominently while saying nothing about how their shows are made. GoodShort's Play listing promises that "our powerful AI engine learns your taste"; Sereal+ advertises "AI recommendation algorithms"; StardustTV, separately from its production statement, says "our AI knows what you like". All three describe recommendation, not creation.

Sereal+ supplies the sharpest illustration of how slippery the word has become. The same company that advertises AI recommendations uses its terms of use to forbid anyone crawling its catalogue to train "large language models, other artificial intelligence, algorithm models". AI is a feature when it serves you a title and a threat when it reads one.

The twelve that have published nothing

DramaBox, StoReel, ShortMax, DramaWave, FlickReels, MyMuse, NetShort, Kalos TV, TopShort, StarShort, TouchShort and YourChannel have published nothing about AI in production, in any form we could find. Between them these twelve include the largest app in the category and the smallest, so this is not a story about scale. StoReel and MyMuse are in this group despite catalogues we judge AI-native, which is exactly the gap this page exists to record.

What "nothing published" is evidence of

It is evidence that nobody has made a statement, and nothing more. DramaBox is the clearest case: no AI language anywhere on its own properties, while third parties report it seeking AI production partners and running hybrid AI dubbing. We do not publish those third-party claims as facts about DramaBox, and we do not publish the silence as a denial either. Both would be dishonest in opposite directions.

We disclose a commercial interest here. DramaBox is the app that pays this site, it sits first in our ranking, and it is in the group that tells you nothing about AI. Saying so is the minimum price of the rest of this page being worth reading.

Several of them barely publish anything at all

For five of the twelve, the absence of an AI statement is part of a wider absence of readable corporate publishing. StarShort's terms and privacy pages render only through JavaScript and returned no text to any tool we have. TopShort serves a TLS certificate that does not match its own hostname. TouchShort publishes three different company names across one product and a Gmail address for contact. YourChannel has roughly 900 installs and no iOS listing.

This matters for how much weight to put on the silence. A company with an About page, a newsroom and a legal team that has chosen to say nothing about AI has made a decision. A company with no readable About page at all has not necessarily decided anything.

Why a US viewer sees no labels

Because no US rule requires one, while the two jurisdictions that do require labels do not reach a US storefront. The same company can be legally obliged to stamp a label on every episode at home and legally free to ship the identical catalogue to an American phone with nothing on it at all.

China: a label on every episode since September 2026

China requires two labels and has done so at two levels. The Measures for Labeling AI-Generated Synthetic Content, issued jointly by the Cyberspace Administration, the Ministry of Industry and Information Technology, the Ministry of Public Security and the National Radio and Television Administration, took effect on September 1, 2025. They require an explicit label a viewer can see on a video's opening frame and an implicit label written into the file's metadata.

A year later the sector got its own rule. NRTA Order No. 16, the Measures for the Administration of Microdrama Development, was signed on July 30, 2026 and took effect on September 1, 2026. Article 34 requires that a microdrama generated with AI carry a prompt label in a prominent position in every episode, binding both the production company and the broadcasting platform. Registration thresholds for AI microdramas were set lower than for live action from July 1, 2026 — tier one from 800,000 yuan rather than three million — on the stated reasoning that AI production costs less.

The European Union: Article 50 since August 2026

The EU AI Act's transparency obligations became applicable on August 2, 2026. Article 50(2) requires providers of systems generating synthetic video to mark outputs in a machine-readable format. Article 50(4) requires deployers to disclose artificially generated or manipulated content.

For drama specifically there is a carve-out worth knowing: where the content is part of an evidently artistic, creative, satirical or fictional work, the duty is limited to disclosing the existence of generated content in a way that does not hamper enjoyment of the work. In practice a credit-style notice satisfies it. A separate point of confusion is worth clearing up: the Digital Omnibus regulation that entered into force on July 27, 2026 delayed the high-risk regime, not Article 50.

The United States: no federal labelling duty

As of October 3, 2026 there is no US federal law requiring that video entertainment be labelled as AI-generated. The AI Labeling Act of 2026, S.4915, was introduced on June 24, 2026 and referred to the Senate Commerce Committee, where it remains. The NO FAKES Act, advanced out of Senate Judiciary in June 2026, is a likeness and voice right rather than a labelling duty.

The Federal Trade Commission's rule on impersonation of government and businesses, in force since April 1, 2024, contains no AI disclosure requirement and does not reach fictional drama. The proposed extension to impersonation of individuals is still unfinished: an FTC advance notice published on October 1, 2026 confirms the Commission is "still evaluating its options". Section 5 deception liability remains the only backstop, and it bites only if an app affirmatively misleads.

California's AI Transparency Act binds the tool, not the app

California's SB 942, as amended by AB 853, took effect on August 2, 2026, and it does not reach a short-drama app. It binds a "covered provider" — a developer of a generative AI system with more than a million monthly users — requiring embedded provenance data in generated media and a free public detection tool. Large online platforms join from January 1, 2027 and capture devices from January 1, 2028.

An app that licenses finished footage is outside its operative scope today. It is also worth correcting a persistent error: AB 3211, the Californian provenance bill routinely cited as law, died in Senate Appropriations in August 2024 and was never enacted. The broader federal posture points the other way in any case — a December 11, 2025 executive order directs a Justice Department task force to challenge state AI laws and asks the FCC to adopt a preempting federal standard.

Apple requires nothing, Google Play labels the trailer

Neither app store requires a developer to disclose that the content inside the app is AI-generated. Apple's App Store Review Guidelines mention AI exactly once, in clause 5.1.2(i), and that clause is about sharing personal data with third-party AI, not about content provenance.

Google Play comes closer and then stops short. Its AI-Generated Content policy governs apps that generate content and explicitly excludes apps that merely host it, which is what a short-drama app does. Separately, Play Console does require developers to self-declare AI-generated marketing assets, and declared assets are labelled on the store. The net effect is that Google may label the trailer an app advertises with while the show inside it carries nothing.

The gap between how these shows are made and what you are told

Production has moved to AI far faster than disclosure has, and the rules that would close the gap are either new or absent. The China Netcasting Services Association reported that more than 95% of new Chinese microdramas in the first quarter of 2026 were AI-generated. Over the same period exactly one of the twenty-three apps serving US viewers labelled a single title.

Three of the four labelling regimes that exist anywhere were switched on in the last fourteen months, and the one that would reach an American viewer does not exist at all.

The volume is not matched by quality, which is part of why the question matters to a viewer and not only to a regulator. DataEye counted roughly 221,900 new AI titles on Douyin in the first half of 2026, of which about 0.47% passed a hundred million views. Most AI output in this category is unwatched, and the question of whether a particular series is worth two hours is not answered by knowing how it was made.

How to check a catalogue yourself

You can settle the question for a specific app in about two minutes, and the order matters. The cheap checks that produce a definitive answer come first; the visual tells come last, because they are the least reliable part of the exercise.

The order to do it in

1. Read the store listing in full, including the what's-new notes. This is the developer's own writing and it is where StardustTV's statement lives. It takes thirty seconds.
2. Open the terms of service and search for "AI". ShortPeak's entire disclosure is in section 3 of its terms and appears nowhere else.
3. Look at the catalogue for per-title badges. Only StardustTV has them today. If you find one somewhere else, it is new, and we would like to hear about it.
4. Check the parent company's newsroom, not just the app. FlareFlow's disclosure is on COL Group's corporate site, and the app itself says nothing.
5. Watch one episode past the paywall and note the shot lengths. Generated footage rarely holds a continuous shot beyond ten or twelve seconds, and faces drift between cuts.
6. Write down the date you checked. Store listings change without notice and carry no archive. Every claim on this page is stamped with the day it was verified for the same reason.

A caution on the last visual step: these tells are suggestive, not conclusive. Cheap live-action microdrama is also shot in very short takes, and good AI work in 2026 clears the ten-second ceiling with careful assembly. Treat a suspicion as a suspicion.

What this page cannot tell you

It cannot tell you what share of any catalogue is AI-produced, because no app publishes that figure and no independent dataset measures it. That is the single largest hole in public knowledge about this category and we would rather name it than paper over it.

• No share figures exist. Not for DramaBox, not for ReelShort, not for the two apps that say outright that their content is generated.
• Store listings are not archived. Every quotation here was captured on October 2, 2026, and a company can change its wording tomorrow without telling anyone.
• Five of the twenty-three sites render only through JavaScript. For StarShort we could not read the terms or the privacy policy at all, so its entry records what is absent from everything else it publishes, not a certified absence.
• C2PA cannot close the gap. Content Credentials is a provenance standard, not a detector, and a joint advisory from the NSA and its UK, Australian and Canadian counterparts states plainly that a lack of provenance information should not automatically make media less trustworthy. A re-encoded streaming episode has no credentials by default.
• Interviews are not policy. ReelShort's chief executive has said more about AI than any other figure in the category, and none of it is a commitment his company has published.

Strengths and weaknesses of the disclosure that exists

Six disclosures in twenty-three is a poor record, and two of the six are only a product name. Judged as consumer information rather than as compliance, they divide sharply into one that helps a viewer at the moment of choosing and three that do not.

Two companion pages come out of the same October 3 check: the series that carry a per-title AI label, and what can and cannot be counted when someone asks how many AI dramas there are.

Frequently asked questions

The bottom line

Six of twenty-three companies have said anything about AI in the shows they sell. Two of the six said it only in the product's name, and one said it to investors rather than to viewers. If you want to know whether what you are about to watch was generated, StardustTV is the only app of the twenty-three that will tell you before you press play, and it tells you title by title rather than in a policy you have to go looking for.

The right conclusion is not that the silent twelve are hiding something. It is that nobody is obliged to say, and in the absence of an obligation almost nobody does. China now requires a label in every episode and the EU requires disclosure; the United States requires nothing, Apple requires nothing, and Google requires a label on the advertisement but not on the product. A company following the rules in all three places will label at home and stay silent here, and that is the whole explanation.

For a viewer, the practical consequence is narrower than it sounds. Knowing a series was generated does not tell you whether it is watchable — DataEye's figure of 0.47% of AI titles passing a hundred million views says most of it is not, and the same is true of most live-action microdrama. Disclosure is worth asking for because you are entitled to know what you are paying for, not because the answer predicts quality.

We will re-check every entry on this page in 90 days and after any change we notice sooner. If you find a statement we have missed, send it with a link and we will verify it, change the entry and date the change.

Related reading

Sources

• Cyberspace Administration of China and three ministries, Measures for Labeling AI-Generated Synthetic Content (国信办通字〔2025〕2号), issued March 7, 2025, effective September 1, 2025 — www.cac.gov.cn
• National Radio and Television Administration, Order No. 16, Measures for the Administration of Microdrama Development, signed July 30, 2026, effective September 1, 2026 (article 34 on AI labelling) — www.nrta.gov.cn
• European Union, Regulation (EU) 2024/1689 (AI Act), Article 50, transparency obligations applicable from August 2, 2026 — eur-lex.europa.eu
• US Congress, AI Labeling Act of 2026, S.4915, introduced June 24, 2026, referred to the Committee on Commerce, Science, and Transportation — www.congress.gov
• Federal Trade Commission, Rule on Impersonation of Government and Businesses, 16 CFR Part 461, effective April 1, 2024, and the advance notice of proposed rulemaking of October 1, 2026 — www.federalregister.gov
• Apple, App Store Review Guidelines, clause 5.1.2(i), accessed October 3, 2026 — developer.apple.com
• Google, AI-Generated Content policy for Google Play, accessed October 3, 2026 — support.google.com
• Google, Declaring AI-generated content in Play Console, accessed October 3, 2026 — support.google.com
• NSA, ACSC, Canadian Centre for Cyber Security and NCSC-UK, Content Credentials: Strengthening Multimedia Integrity in the Generative AI Era, January 2025 — media.defense.gov
• Coalition for Content Provenance and Authenticity, specification 2.4, April 2026; ISO 22144 remains in development and is not a published standard — c2pa.org
• TIME, interview with Joey Jia, founder and chief executive of Crazy Maple Studio (ReelShort), November 10, 2024 — time.com
• TheWrap, Vertical Media Summit coverage quoting Joey Jia, June 5, 2026 — www.thewrap.com
• COL Group corporate newsroom, "COL Group Launches FlareFlow, a Global Micro-Drama Platform", August 24, 2025 — plain text, not linked: col.com/article/1959876057200820224
• SagaBox homepage, ShortPeak terms of service section 3, StardustTV Apple App Store and Google Play listings and catalogue badges, Playlet store listings and playletmedia.com, GoodShort and Sereal+ store listings — all read directly and quoted above, all accessed October 3, 2026. Not linked: this site carries no direct links to the apps it ranks.
• China Netcasting Services Association on the AI share of new Chinese microdramas, Q1 2026; DataEye on ~221,900 Douyin AI titles and the 0.47% hit rate, first half of 2026. Both also used in our statistics guide.