What to Look for in an AI Vertical Drama Production Partner: The Due Diligence Checklist
Most AI video generators can produce a single impressive clip. The YouTube Short that goes viral. The TikTok that stops the scroll. The concept demonstration that a production company shows in a pitch meeting.
Very few can produce 70 consistent episodes that pass platform acquisition review.
The gap between those two capabilities is the most commercially important thing to understand before signing a commission agreement with an AI-native vertical drama production partner. The production company that can generate a compelling five-second clip from a sophisticated prompt is demonstrating model access and basic prompt competency. The production company that can generate episode forty-seven with the same character visual identity, the same visual register, the same audio standard, and the same structural discipline as episode one is demonstrating production infrastructure.
The right partner needs to connect story, production, testing, localisation and platform economics in one workflow. That connection is what the due diligence checklist below evaluates. Not whether the production company has access to AI tools. Every production company in 2026 has access to AI tools. Whether the production company has built the production infrastructure that makes those tools commercially viable at series scale.
What the Checklist Is Evaluating
The due diligence checklist covers six capability domains. Each domain represents a specific component of the AI-native production pipeline. A production company that excels at four domains and fails at two produces a series that will either fail acquisition review or require expensive revision before distribution.
The domains are:
Character consistency infrastructure
Arc and script quality discipline
Audio post-production capability
Visual register management
Delivery and documentation standards
Platform relationship and commercial track record
Each domain includes specific questions, specific evidence to request, and specific red flags that indicate infrastructure gaps rather than capability claims.
Domain 1: Character Consistency Infrastructure
Character consistency is the most commercially critical production capability for AI-native vertical drama and the capability that most clearly separates production-grade from consumer-grade AI production.
Questions to ask:
How do you maintain character visual identity across 70 episodes produced over six to eight weeks by operators who may be working in different sessions?
What is your character reference pack build process and how long does it take to build a production-ready reference pack for a new series?
How do you handle character consistency failures when they appear in a generation batch?
Evidence to request:
A sample of three episodes from a completed series, chosen at random by you rather than by the production company. Ask for episodes one, twenty-five, and sixty-five. Compare the controlled alpha's facial structure, eye colour, and skin tone across all three. If the character is visually identical across the three episodes, the production company has consistent character reference infrastructure. If there is visible variation, the infrastructure is insufficient.
Ask to see the character reference pack document for a prior production. A production-grade reference pack contains approved master images for each character, arc position variant images for wardrobe progression, and the ControlNet geometry reference for the standard camera positions. A reference pack that is a folder of unorganised reference images is not a production-grade reference pack.
Red flags:
The production company describes character consistency as a prompt engineering challenge rather than an infrastructure challenge. Character consistency through prompt description alone is consumer-grade production. Production-grade character consistency requires Soul ID training or equivalent character model locking.
The sample episodes show visible character variation between episodes one and sixty-five. Any production company whose sample shows this variation cannot produce a series that passes tier-1 or tier-2 platform acquisition review.
Domain 2: Arc and Script Quality Discipline
The visual production is the delivery mechanism. The arc and script quality is the commercial substance. A production company that generates visually consistent content from a poorly designed arc map is producing content that looks correct but fails at the paywall.
Questions to ask:
What does your arc map look like and at what stage of pre-production is it completed?
Who writes the scripts for a commissioned series and what quality review process do the scripts go through before generation begins?
How do you specify the paywall episode's button cut and how is that specification translated into the script brief?
Evidence to request:
An example arc map from a prior production, with client-identifying information removed. A production-grade arc map specifies the paywall episode position, the midpoint reversal episode, the three to four tension axes and their primary episode blocks, and the specific emotional debt type driving paywall conversion. A document that describes the story's general direction is not an arc map.
An example writer brief for a vertical drama episode. A production-grade writer brief specifies the timestamp skeleton, the hook detonation requirement, the paywall brief, and the dialogue constraints. A brief that says write a 90-second episode about X is not a production-grade writer brief.
Red flags:
The production company generates from scripts that have not been reviewed against a locked arc map. Arc-map-free scripting produces episodes that advance the story but not the arc's commercial mechanics.
The production company cannot provide an example arc map or writer brief and describes their story development process as primarily creative rather than structural.
Domain 3: Audio Post-Production Capability
Audio quality failure is the second most common reason AI-native vertical drama fails platform acquisition review, and the most common reason it fails the phone display test that every quality assessment should include.
Questions to ask:
What is your LUFS target for delivery and which standard does it align to?
Do you conduct a phone speaker playback test on every episode before approval? On which devices?
Are you delivering audio stems separately or as a combined stereo master?
Evidence to request:
Watch the sample episodes on a consumer phone at arm's length with phone speaker audio in ambient room noise. Ask the production company to specify the LUFS target they produced the episodes to before you conduct the test. If the dialogue is intelligible without straining on the phone speaker in ambient noise, the audio post-production is calibrated correctly. If the dialogue is muddy, compressed, or competing with the music bed in ambient noise, the audio was not calibrated for the delivery device.
Ask for the audio specification document from a prior delivery. A production-grade audio specification includes LUFS integrated target, true peak ceiling, the dialogue notching process for phone frequency response, and the stem delivery format.
Red flags:
The production company evaluates audio quality on studio monitors or desktop speakers rather than on consumer phones. This is the most common audio quality gap because it produces audio that sounds good in evaluation and sounds wrong in distribution.
The production company delivers combined stereo masters without stems. Combined stereo delivery prevents localisation at quality and prevents platform-level audio remediation when needed.
Domain 4: Visual Register Management
Visual register management is the production discipline that maintains the series' colour palette, lighting character, and environment visual standard across all 70 episodes and across multiple generation sessions.
Questions to ask:
Do you produce a visual style guide for each commissioned series and at what stage of pre-production is it completed?
How do you communicate the approved visual register to generation operators working in different sessions?
What is your approval process for generation batches before they are incorporated into the production?
Evidence to request:
Request the visual style guide from a prior production. A production-grade style guide specifies the colour palette in specific hex codes or LUT references, the lighting setup for each primary scene type with technical parameters, and the phone display validation criteria applied at each batch approval stage.
Review the sample episodes on a phone in ambient light specifically for visual register consistency. Does the colour temperature of the controlled alpha's office in episode one match episode twenty-five? Does the confrontation lighting's shadow depth match across episodes? Visual register drift that is subtle on a monitor is visible on a phone.
Red flags:
The production company describes their visual register management as a prompt-based process without a documented style guide. Prompt-based visual register management produces output that is approximately consistent within a single session but drifts between sessions as different operators interpret the same prompts differently.
The sample episodes show visible colour temperature or lighting character variation between episode blocks. Block-to-block visual drift indicates session-to-session inconsistency in the generation workflow.
Domain 5: Delivery and Documentation Standards
The delivery package is the commercial instrument that closes the acquisition conversation. A production that delivers content without a complete data room is a production that cannot close a platform deal requiring E&O coverage and chain of title documentation.
Questions to ask:
What documents do you include in your standard delivery package?
Do you carry E&O insurance coverage for AI-generated content and can you provide a certificate?
What AI tool usage documentation do you include in the chain of title package?
Evidence to request:
Request a delivery package index from a prior production. A production-grade delivery index includes: video master files at platform-specified codec and resolution, audio stems, subtitle files, chain of title documentation including work-for-hire agreements for all writers, AI tool usage documentation, performer releases, and E&O certificate.
Ask the production company to confirm the specific E&O carrier and whether the policy explicitly covers AI-generated content claims. The E&O post covers why this distinction matters. A production company that cannot confirm AI-specific E&O coverage is leaving the commissioning party exposed to uninsured chain of title risk.
Red flags:
The production company's delivery package does not include chain of title documentation for AI tool usage. In 2026, AI tool usage documentation is a standard delivery requirement for platform distribution.
The production company does not carry E&O insurance or carries a policy with an AI exclusion endorsement without disclosure.
Domain 6: Platform Relationship and Commercial Track Record
The production company's prior platform relationships are the most commercially valuable reference for a commissioning party evaluating whether the production company can deliver content that platforms actually acquire.
Questions to ask:
Which platforms have acquired series you have produced and can you name them?
What paywall conversion rates have your series achieved in primary distribution and can you provide platform dashboard documentation?
Have you had a series rejected at platform acquisition review and what was the reason?
Evidence to request:
Platform acquisition confirmation for at least one prior series. A production company that has never had a series acquired by a named platform is not a production company with a commercial track record. It is a production company with production capability and no distribution validation.
Platform dashboard data from a prior series showing episode completion rate, paywall conversion rate, and day-7 retention. These numbers are the commercial evidence that the production company's output performs in live distribution. A production company that refuses to share any platform performance data from prior series is withholding the most commercially relevant evidence available.
Red flags:
The production company can demonstrate impressive visual quality but cannot name a platform that has acquired their content. Visual quality and distribution viability are different capabilities.
The production company quotes only concept test performance data and has no live distribution performance data. Concept test data predicts performance. Live distribution data confirms it. A production company without any live distribution performance data across any prior series has not validated its production at commercial distribution scale.
The Scoring Framework
Score the production company across all six domains on a simple three-point scale:
3 points: Production-grade evidence provided. All questions answered specifically. No red flags.
2 points: Some evidence provided with gaps. Questions answered generally. One minor red flag.
1 point: No meaningful evidence provided. Questions answered vaguely. Multiple red flags or one major red flag.
0 points: Domain capability not present.
Score interpretation:
15 to 18 points: Commission with standard contract terms. The production company has demonstrated production-grade capability across all domains.
10 to 14 points: Commission a concept test series before committing to full production. The gaps identified in the lower-scoring domains should be specifically addressed in the concept test brief.
6 to 9 points: Do not commission until the production company can demonstrate the missing capabilities. The gap between their claim and their evidence is too large to accept at full production commission investment.
Below 6 points: Do not commission. The production company does not have the infrastructure to deliver a series that passes platform acquisition review.
Axis AI Studios Perspective
The due diligence checklist above is the checklist we would want a commissioning party to apply to Axis AI Studios before signing a commission agreement. A production partner that cannot answer the questions above with specific evidence rather than general claims is a production partner whose commercial claims should be treated with caution.
At Axis AI Studios, the character consistency infrastructure, the arc and script quality discipline, the audio post-production capability, the visual register management, the delivery documentation standards, and the platform relationship track record are all documented and available for commissioning party review before any production agreement is signed.
For commissioning parties who want to conduct a formal due diligence review before entering a production agreement, reach out at business@axisaistudios.com.
FAQ
How Long Should a Due Diligence Process Take Before Commissioning?
A thorough due diligence review using the checklist above takes two to three weeks from initial evidence request to commission decision. The timeline is dominated by the evidence review rather than the question-and-answer process. Reviewing three sample episodes on a phone, reading through a visual style guide and arc map, and confirming an E&O certificate with a broker typically takes three to five working days per domain when the evidence is complete and organised.
Should a Commissioning Party Request References From Prior Clients?
Yes, specifically from the platform acquisition teams that have acquired the production company's content rather than from the brands or IP holders that commissioned the production. The platform acquisition team's assessment of the production company's delivery quality and revision rate is the most commercially relevant reference available. A brand client who commissioned a branded microdrama may not have the format expertise to evaluate the production company's vertical drama specific capabilities.
Is It Reasonable to Commission a Concept Test Series as a Due Diligence Step?
Yes, and it is the most commercially reliable due diligence step for a commissioning party that has not previously worked with the production company. A concept test series at $15,000 to $20,000 produces three episodes that can be evaluated against all six domains of the checklist simultaneously. The concept test result is more reliable evidence than any prior production example because it is produced specifically under the commissioning party's brief rather than under a prior client's brief.
Further Reading
For the brief that the production company executes when commissioned, the guide to how to brief an AI-native production partner covers what the brief must specify and what happens when it does not.
For what platform acquisition teams look for when reviewing AI-native content at delivery, the quality assessment guide for platform buyers covers the five quality markers and the phone display test that the due diligence sample review uses.
For the data room and delivery documentation that Domain 5 of this checklist evaluates, the guide to what vertical drama data rooms look like covers the complete documentation package that every AI-native production delivery should include.
Let's set
the new standard together.
If you're working on something, we'd like to hear about it.