How to Evaluate an AI Vertical Drama Production Partner: What to Ask

The AI vertical drama market is not a regulated industry with published quality standards, licensed operators, or a credentialing body that distinguishes experienced production companies from people who bought a Kling subscription last month. Anyone with access to AI generation tools can describe themselves as an AI vertical drama production studio. The gap between that description and the operational reality can be $80,000 of your commissioning budget.

The commissioning organization that does not have a due diligence framework for evaluating AI production partners is making decisions based on whoever presents most confidently. That is not a commissioning strategy. It is a luck strategy.

This is the complete due diligence guide for anyone commissioning AI-native vertical drama: what to ask, what the answers should look like, what answers reveal a partner who cannot deliver, and what questions most commissioners forget to ask until after the contract is signed.

Why AI Production Partner Evaluation Is Harder Than Conventional Production

Evaluating a conventional live-action production company is a reasonably established discipline. You can look at their produced content on streaming platforms, review their delivery track record, speak with platforms they have delivered to, and assess their crew roster against industry standards. The evaluation criteria are knowable and the evidence is verifiable.

Evaluating an AI-native production company is harder for three specific reasons.

First, the entry barrier is low. A generation tool subscription, a laptop, and a willingness to describe one's output as professional quality is all that is required to present as an AI vertical drama production company. The tools exist. The operational discipline to produce consistent, platform-standard output at series scale does not automatically come with the tools.

Second, sample quality does not predict series quality. An AI-native production company can generate an impressive five-episode sample by spending disproportionate time and resources on those five episodes, curating only the best generated frames, and applying more manual post-production work than the per-episode economics of a 70-episode series will support. A sample that looks impressive does not mean the production company can deliver that quality across 70 episodes on a contracted timeline at the quoted price.

Third, character consistency across episodes is the defining technical challenge of AI-native production and is invisible in a five-episode sample. Character drift that begins to appear at episode 15 or 20 cannot be detected by reviewing a sample. The evaluation has to look for the production infrastructure that prevents character drift rather than for evidence of character drift in the sample.

Evaluation Area 1: Delivered Productions

The single most important evaluation question for any AI production partner is simple: name every series you have produced and delivered to a platform, and tell me where I can watch them.

The answer to this question immediately separates production companies with real delivery track records from those without one.

What a strong answer looks like: A specific list of series titles, the platforms they are distributed on, and the episode counts and dates of delivery. Series that are live on ReelShort, DramaBox, FlexTV, GoodShort, NetShort, or any established vertical drama platform and that the commissioner can verify by searching for them on the platform.

What a weak answer looks like: References to productions "in development," productions "in post-production," productions "delivered to a client who is not ready to announce," or productions distributed on platforms the commissioner has not heard of and cannot locate. A production company that has produced 10 series but none of them are verifiable on any known platform has either delivered to non-serious buyers or has not delivered to buyers at all.

What to do with the answer: Search for the named series on the named platforms. Watch episodes 1, 5, 10, and 30 if the series has reached that episode count. You are looking for character consistency across those episodes, audio quality on your phone speaker, correct episode-end mechanics, and whether the production quality is consistent or whether it degrades across the series.

The red flag to walk away from: A production company that cannot name a single delivered series on a verifiable established platform is a production company that has not demonstrated delivery capability. An impressive demo reel is not a delivery track record.

Evaluation Area 2: Platform Relationships

An AI production company that produces series no platform will acquire has a quality or infrastructure problem. An AI production company that has established acquisition relationships with one or more recognized vertical drama platforms has demonstrated that its output meets at least one platform's acquisition standard.

Ask specifically: which platforms have you delivered to and who is your acquisition contact at each platform?

What a strong answer looks like: Named platforms with named acquisition contacts whose roles can be verified on LinkedIn or through direct platform inquiry. A production company with established relationships at ReelShort, DramaBox, GoodShort, or any other recognized platform has been through the acquisition review process and passed it.

What a weak answer looks like: Platform names without contact names, vague references to "platform partners," or contacts at platforms that do not appear in any industry directory. A production company that claims platform relationships but cannot provide the name of a specific acquisition contact they have worked with is either misrepresenting the relationship or has not had a serious acquisition conversation.

What to verify: Contact the acquisition contact at the named platform independently. Ask whether they have acquired content from the production company and whether they would describe the relationship as active. Most acquisition teams will confirm a working relationship without breaching confidentiality.

Evaluation Area 3: The Character Reference Infrastructure

This is the evaluation area that most commissioners never ask about and that most directly predicts whether a 70-episode series will hold character consistency or drift visibly.

Ask specifically: how do you build and maintain character reference packs for a series, and can you show me the reference documentation for a previously produced series?

What a strong answer looks like: A detailed explanation of the pre-production character reference pack construction process: primary reference image selection, emotional register references, lighting variant references, wardrobe configuration documentation, and negative reference sets. A production company with established infrastructure will have a documented workflow for this that they can describe specifically and illustrate with examples from previous productions.

What a weak answer looks like: A description of the character reference process that amounts to "we describe the character consistently in our prompts." Consistent prompt descriptions produce consistent character types. They do not produce consistent specific characters. A production company that does not have a visual reference infrastructure for character consistency will produce a series that drifts, regardless of how carefully their prompts describe the character.

What to test: Ask to see the character reference documentation for one of their previously produced series. A production company with genuine infrastructure has this material. A production company that built its previous series on prompt descriptions alone does not have reference documentation to show.

Evaluation Area 4: Delivery Track Record Against Timeline

The timeline commitment in a vertical drama production contract is not a wish. It is a delivery obligation with commercial consequences for the commissioning organization's platform relationships. A production company that cannot deliver on time is a production company that damages your platform relationship on your behalf.

Ask specifically: for every series you have previously delivered, what was the contracted delivery date and what was the actual delivery date?

What a strong answer looks like: A track record of on-time delivery across multiple productions. A production company with an established pipeline that has delivered five series on time has demonstrated that their timeline estimates are reliable. Some variation is acceptable. A pattern of late delivery is not.

What a weak answer looks like: Vague responses about delivery being "approximately on schedule," explanations of why specific productions ran late that frame the delay as exceptional rather than structural, or an inability to name specific contracted delivery dates because the previous productions did not have formal delivery commitments.

What to verify: Request copies of previous delivery confirmation emails or platform acceptance notices with dates. A production company with a genuine delivery track record has this documentation. A production company that has not delivered formally to any platform cannot produce it.

Evaluation Area 5: The Phone Test on Delivered Work

No due diligence framework for AI production partners is complete without watching delivered work on a phone. Not on a laptop. Not on a monitor. On a phone, in a lit room, at arm's length, without headphones.

This is the single most reliable quality test because it replicates the actual delivery environment. A production that looks acceptable on a laptop screen and fails on a phone is a production that fails the platform's acquisition review and the viewer's experience.

Specifically watch for:

Audio on the phone speaker. Does the dialogue hold its intelligibility on the phone speaker in the ambient noise of the room? Does the music overwhelm the dialogue? Does the mix sound flat or compressed in a way that diminishes the scene's emotional charge? Audio failure is the most common single reason for platform acquisition rejection, and it is audible in the first 30 seconds.

Character consistency at episode 10, 20, and 30. Does the lead character look like the same person at episode 30 as they did at episode 1? Skin tone consistency, facial proportion consistency, and wardrobe continuity across episodes are the visible markers of a production company's character reference infrastructure. Drift at episode 10 means the infrastructure was not adequate for a full series.

Episode-end mechanics. Does each episode end before the tension releases? An episode that resolves its central conflict before the cut does not convert at the paywall. Watch the last 10 seconds of three consecutive episodes. If any of them feel complete rather than suspended, the production company does not understand the format's conversion mechanics.

Production consistency across the series. Does episode 25 look like it was produced by the same team as episode 5? A production company that runs out of quality consistency across a 70-episode series delivers a product that the platform's recommendation algorithm will not favor and that the audience's second-session completion rate will not support.

Evaluation Area 6: The Production Pipeline Documentation

Ask the production company to describe their production pipeline in specific operational terms. Not conceptually. Specifically: what happens in each week of a production, who does what, how are quality gates managed, and what happens when generated output fails the quality standard?

What a strong answer looks like: A week-by-week production calendar with specific activities, responsible roles, and quality gate criteria at each stage. A description of the batch generation and batch review workflow that catches character drift and technical failures before they multiply across the series. A clear answer to what happens when a generation batch fails the quality review: the specific process for identifying failing shots, regenerating them, and integrating the corrected output before the batch is approved.

What a weak answer looks like: A general description of the production process that does not include quality gate criteria, does not describe how failures are caught, and does not address what happens when output does not meet the required standard. A production company that has not built quality gates into its pipeline has not built a production pipeline. It has built a generation workflow that produces whatever the tools produce and delivers it.

Evaluation Area 7: The Delivery Package

Ask specifically: what does your standard delivery package include, and can you provide a sample delivery manifest from a previous production?

The delivery package for an established vertical drama platform requires video files in the correct codec and resolution, audio mixed to mobile loudness standards, separated stems for localization pipelines, subtitle masters in the correct format, music cue sheets, and series metadata. A production company that delivers a combined mix without stems to a platform that requires stems creates a problem the commissioner has to solve.

What a strong answer looks like: A complete list of delivery components that matches or exceeds the platform's standard delivery specification, evidence that the production company confirms the delivery specification directly with each platform before production begins, and a sample manifest from a previous delivery.

What a weak answer looks like: A delivery package description that does not include stems, does not include music cue sheets, or that describes the technical specifications in approximate rather than precise terms. "We deliver in the standard format" is not a delivery specification. A production company that does not know the difference between -14 LUFS and -23 LUFS has not calibrated its audio pipeline to mobile delivery standards.

Evaluation Area 8: Rights and Compliance Documentation

Ask specifically: how do you ensure that AI-generated character imagery does not constitute a digital replica of a real performer, and what documentation do you provide to confirm this for US distribution?

Under the SAG-AFTRA Verticals Agreement, AI digital replicas of real performers must be deleted within 90 days. Generated fictional characters without real performer likeness components are not covered by these provisions. The production company needs to be able to confirm specifically that their AI-generated characters are fictional rather than based on real performer imagery.

What a strong answer looks like: A clear explanation of the distinction between generated fictional characters and digital replicas, a documented workflow that confirms no real performer imagery is used as generation reference material without explicit consent documentation, and a willingness to provide a written representation on this point in the delivery agreement.

What a weak answer looks like: Uncertainty about the legal distinction, a vague assurance that "we follow all regulations," or an inability to describe the specific steps in their workflow that prevent real performer imagery from entering the generation reference pipeline without consent.

Also ask about music clearance. Every piece of music in the series requires documentation confirming that the license covers commercial distribution on paid subscription streaming platforms. A production company that uses a library music subscription without confirming that the subscription tier covers this distribution type is creating a rights problem the commissioner inherits at delivery.

Evaluation Area 9: Pricing Transparency

Ask for a detailed line-item budget for the proposed production, not a lump sum.

A production company that can provide a specific line-item breakdown of what the quoted price covers, AI generation credits per episode, audio post-production per episode, color grading per episode, delivery preparation, and project management, is a production company that has built and managed budgets at series scale. A production company that provides a lump sum without breakdown cannot account for how the budget is allocated across the production or demonstrate that the budget is adequate to complete the series at the quoted quality standard.

Also ask: what are the specific conditions under which the budget would change? A production that was quoted as fixed-price but that contains unlimited revision allowances, scope creep provisions, or undefined quality standard targets is not a fixed-price production. The price changes when the commissioner asks for any adjustment the production company considers outside the original scope.

The Question Most Commissioners Never Ask

After all the structural due diligence, the question that most consistently reveals the difference between a production company that understands the format and one that does not is this:

Walk me through your paywall episode approach. How do you determine where the paywall lands in the arc, and what specifically does the final free episode cut on?

A production company that understands vertical drama's commercial mechanics answers this question immediately and specifically: the paywall is identified in script development as a structural milestone, it is placed at the first genuine peak of the central tension arc, and the final free episode cuts at the moment of maximum unresolved tension before any form of response to the tension peak.

A production company that does not understand the format answers this question by talking about where the platform places the paywall, or by describing the paywall as a platform decision rather than a production decision.

The paywall moment is a production decision. It is the most commercially important production decision in the series. A production company that treats it as a platform decision is telling you, in answer to one question, that its series will convert at 2% rather than 12%.

Axis AI Studios Perspective

The due diligence framework in this post exists because the questions it asks are the questions Axis AI Studios can answer. Delivered series on verifiable platforms. Named acquisition contacts. Character reference documentation. Delivery track records with dates. Line-item budgets. Paywall episode architecture designed at the script stage.

The production company that is uncomfortable with this level of scrutiny is not a production company that should receive a commissioning budget. The production company that answers every question with specific, verifiable evidence and offers to provide documentation at each step is a production company that has been through the delivery process enough times to have built the infrastructure the questions are asking about.

If you want to run this due diligence framework against us, we welcome it. The answers exist, the documentation is available, and the delivered series are on the platforms. For commissioning conversations, reach out at business@axisaistudios.com.

Due Diligence Checklist: Questions to Ask Every AI Production Partner

Delivered productions:

• Name every series you have produced and delivered. Where can I watch them?

• What platform distributed each series?

• What were the episode counts and delivery dates?

Platform relationships:

• Which platforms have you delivered to and who is your acquisition contact at each?

• Would that contact describe your relationship as active?

Character consistency infrastructure:

• How do you build and maintain character reference packs?

• Can you show me the reference documentation for a previously produced series?

• What is your process when generated output fails the character consistency check?

Delivery timeline:

• For every previous delivery, what was the contracted date and the actual date?

• Can you provide delivery confirmation documentation with dates?

Phone test:

• Can I watch episodes 1, 10, 20, and 30 of a previously produced series on my phone right now?

Production pipeline:

• Describe your week-by-week production calendar with quality gate criteria.

• What specifically happens when a generation batch fails the quality review?

Delivery package:

• What does your standard delivery package include?

• Do you provide separated stems? Music cue sheets? Subtitle masters?

• Can you provide a sample delivery manifest from a previous production?

Rights and compliance:

• How do you confirm AI-generated characters are not digital replicas of real performers?

• What documentation do you provide on this for US distribution?

• How do you confirm music clearance covers commercial distribution on paid platforms?

Pricing:

• Provide a line-item budget breakdown for this production.

• Under what conditions would the price change?

The format question:

• Walk me through your paywall episode approach. How do you determine where the paywall lands and what does the final free episode cut on?

FAQ

What Is the Single Most Important Question to Ask an AI Production Partner?

Name every series you have produced and delivered to a platform, and where can I watch them. Everything else in the due diligence process builds from the answer to this question. A production company with a verifiable delivery track record on established platforms has demonstrated the minimum requirement for serious commissioning consideration. A production company that cannot answer this question with specific, verifiable examples is not ready for a serious commissioning relationship regardless of how impressive their demo material is.

How Do I Evaluate Character Consistency Without Watching a Full Series?

Watch episodes 1, 10, and 30 of the same series on a phone. The character consistency problem in AI-native production manifests gradually across episodes, not within single episodes or in the first few episodes. A character who looks consistent in episode 1 may have drifted by episode 20. Episode 30 is the minimum episode count at which character drift becomes visible if the production infrastructure was inadequate. A production company that can only show you sample content rather than a full series cannot demonstrate that they have solved the consistency problem the full series requires.

Should I Commission a Small Test Production Before a Full Series?

Yes, if the production company does not have a verifiable delivery track record on established platforms. A 5 to 10 episode concept test at $15,000 to $30,000 generates performance data and demonstrates the production company's delivery capability before the full series budget is committed. The test is not a substitute for the due diligence questions in this guide. It is additional evidence that answers the questions the due diligence process raises. A production company that delivers a strong concept test with correct character consistency, mobile-calibrated audio, and proper episode-end mechanics has demonstrated more about their delivery capability than any portfolio presentation can.

Further Reading

For how the commissioning conversation following this due diligence process actually works, the buyer's guide to commissioning AI-produced vertical drama covers the full process from brief to delivery.

For the pre-production checklist that a quality AI production partner should be running before any generation begins, the pre-production checklist for AI vertical micro dramas covers every decision that distinguishes a production company with genuine infrastructure from one without it.

For the repeatable pipeline infrastructure that an established AI production company should have in place, the guide to building a repeatable AI drama production pipeline covers the asset libraries, generation workflows, and quality controls that produce consistent output at series scale.