AGENCYSCRIPT
CoursesEnterpriseBlog
đź‘‘FoundersSign inJoin Waitlist
AGENCYSCRIPT

Governed Certification Framework

The operating system for AI-enabled agency building. Certify judgment under constraint. Standards over scale. Governance over shortcuts.

Stay informed

Governance updates, certification insights, and industry standards.

Products

  • Platform
  • Certification
  • Launch Program
  • Vault
  • The Book

Certification

  • Foundation (AS-F)
  • Operator (AS-O)
  • Architect (AS-A)
  • Principal (AS-P)

Resources

  • Blog
  • Verify Credential
  • Enterprise
  • Partners
  • Pricing

Company

  • About
  • Contact
  • Careers
  • Press
© 2026 Agency Script, Inc.·
Privacy PolicyTerms of ServiceCertification AgreementSecurity

Standards over scale. Judgment over volume. Governance over shortcuts.

On This Page

Before You GenerateDefine the Brief PreciselyConfirm the Tool Fits the JobDuring GenerationGenerate Options, Then SelectWatch for the Usual DefectsBefore You PublishRights and ProvenanceFidelity and AccuracyAccessibility and LayoutFinal Human ReviewContext and PlacementCommon Items People Wish They Had CheckedThe Likeness and Style TrapThe Documentary MisreadThe Silent Brand DriftHow to Use This ListAdapt It to Your Risk LevelMake It a Shared StandardAutomate What You Can, Gate What You CannotFrequently Asked QuestionsWhy check the brief before touching the tool?What is the most commonly skipped item?Do I really need to add text separately?Why does a second reviewer matter so much?How do I scale the checklist to deadline pressure?Should I disclose that an image is AI-generated?Key Takeaways
Home/Blog/Before You Ship a Generated Image, Confirm These Things
General

Before You Ship a Generated Image, Confirm These Things

A

Agency Script Editorial

Editorial Team

·April 7, 2019·8 min read
AI image generatorsAI image generators checklistAI image generators guideai tools

A checklist is only worth keeping if it catches the mistakes you actually make. The failures with generated images are remarkably predictable: a garbled word slips into a poster, a likeness too close to a real person sneaks through, a license question goes unasked until a client asks it for you. Each of those is preventable with a thirty-second check.

This is a working list you can run before any generated image leaves your hands. It is organized by phase, from defining the brief to the final review before publish. Every item includes a short reason, because a checklist whose logic you understand is one you will actually use rather than skip.

Treat it as a starting template. Cut items that do not apply to your context and add ones specific to your brand, but resist the urge to trim it down to nothing on a deadline. The deadline is exactly when the skipped check comes back to bite you.

One framing makes the whole list easier to use. Generated images fail in two broad ways: they fail to do their job, which a brief-fit check catches, or they create a liability, which a rights and accuracy check catches. Almost every item below maps to one of those two concerns. If you remember nothing else, remember to ask of every image, does this do its job, and does this expose us to anything. The rest is detail that helps you answer those two questions reliably.

Before You Generate

Define the Brief Precisely

  • Name the style, palette, and composition explicitly. Vague briefs average toward generic output.
  • State any negative space needed for text or layout. Retrofitting space is harder than reserving it.
  • Decide whether the brief tolerates a plausible approximation or demands literal fidelity. This single question determines whether the tool fits at all.

Confirm the Tool Fits the Job

  • Verify the subject is not a real product, real person, or exact brand asset. Those drift and are unreliable.
  • Check whether you need consistency across multiple images, which requires deliberate setup rather than default generation.

During Generation

Generate Options, Then Select

  • Produce a batch, not a single image. The value comes from cheap variety and hard selection.
  • Save the prompt and seed for anything promising. You will want to reproduce or iterate later.

Watch for the Usual Defects

  • Inspect hands, faces, and any small repeated elements for distortion.
  • Avoid generating embedded text; plan to add typography in a design tool instead.

Before You Publish

Rights and Provenance

  • Confirm the tool's license permits commercial use for your context. Terms vary by platform and plan.
  • Check that the output does not closely mimic a living person, a trademarked character, or a copyrighted style tied to a named artist. Proximity invites disputes.

Fidelity and Accuracy

  • Verify nothing factual is implied that is false: wrong product details, impossible scenes presented as real, misleading depictions.
  • For any image that could be read as documentary, disclose that it is generated where honesty or regulation requires it.

Accessibility and Layout

  • Write meaningful alt text describing the image's content and purpose.
  • Confirm the image works at the sizes it will actually appear, not just at full resolution.

Final Human Review

  • Have a second person look at it. The generator's most convincing errors are the ones you stop seeing after staring too long.
  • Confirm it matches the brand voice, not just the brief. On-brief and off-brand is a common failure.
  • Check the image at a glance the way a real viewer will see it, not under careful study. Many defects that survive close inspection jump out in a quick, distracted look, which is how your audience actually encounters it.

Context and Placement

  • Verify the image reads correctly in its final surroundings, beside the headline, within the layout, against the background it will sit on. An image that works in isolation can fail in context.
  • Confirm the crop and aspect ratio match every placement, since a composition that works wide may lose its subject when cropped square for a different channel.
  • Check that nothing in the frame conflicts with overlaid elements like buttons, captions, or logos that will sit on top of it after publish.

Common Items People Wish They Had Checked

The Likeness and Style Trap

The most expensive omissions cluster around two things: an output that resembles a real, identifiable person, and an output that closely imitates the signature style of a named living artist. Both invite disputes that no amount of image quality justifies. Add an explicit check that asks whether anyone could plausibly claim the image depicts them or copies their distinctive look, and route anything ambiguous to a second opinion before it ships.

The Documentary Misread

An image that is obviously illustrative needs no disclaimer, but an image that could be read as a photograph of a real event does. The check is simple: would an average viewer assume this actually happened or actually exists? If yes, and it does not, you have a truthfulness problem to solve through disclosure or a different image. This matters most for news-adjacent, testimonial, and before-and-after contexts.

The Silent Brand Drift

Generators have house styles, and over many images a brand's visual identity can quietly drift toward the tool's default aesthetic rather than the brand's own. No single image triggers it; the check is periodic, looking at a body of recent work and asking whether it still looks like the brand or like the generator. Catching drift early is far cheaper than rebuilding a diluted identity later.

How to Use This List

Adapt It to Your Risk Level

A throwaway internal slide needs three checks; a paid client campaign needs all of them. Scale the rigor to the stakes. The point is a conscious decision about which checks to skip, not skipping them by accident.

Make It a Shared Standard

The list works best as a team norm, embedded in your asset workflow, so quality does not depend on whether one designer happened to remember. Pair it with the framework and metrics pieces below to turn ad-hoc generation into a repeatable practice.

Automate What You Can, Gate What You Cannot

Some items lend themselves to automation: license terms can be documented once per tool, alt-text reminders can live in your publishing template, and a generated-asset tag can be required before publish. Others resist automation and need human judgment, the likeness check, the brand-fit read, the documentary misread. Build the routine items into your tooling so attention is freed for the judgment items. The goal is not to slow everyone down with bureaucracy but to make the unskippable checks genuinely unskippable while keeping the cheap ones invisible. A checklist that lives in the workflow rather than in someone's memory is the one that actually prevents the failures it was written for.

Frequently Asked Questions

Why check the brief before touching the tool?

Because vague briefs produce generic images, and the most expensive mistakes are decided before you generate. Defining style, composition, and fidelity needs up front determines whether the tool even fits.

What is the most commonly skipped item?

Rights and provenance. People generate something they love and forget to confirm the license permits commercial use or that it does not mimic a real person or protected style. That gap surfaces at the worst possible moment.

Do I really need to add text separately?

For anything longer than a couple of words, yes. Even capable models drift into garbled letterforms, and embedded-text failures are among the most public and embarrassing. Add typography in a design tool.

Why does a second reviewer matter so much?

The model's best errors are subtle and you go blind to them after repeated viewing. A fresh pair of eyes catches the extra finger, the off-brand tone, or the implausible detail you stopped seeing.

How do I scale the checklist to deadline pressure?

Match rigor to stakes. Internal throwaways need a few checks; paid campaigns need all of them. Decide consciously which to skip rather than abandoning the list because time is short.

Should I disclose that an image is AI-generated?

Where honesty or regulation requires it, yes, especially for anything that could be read as documentary or factual. Disclosure protects trust and increasingly satisfies platform and legal expectations.

Key Takeaways

  • Most generated-image failures are predictable and catchable with quick checks.
  • Decide before generating whether the brief tolerates approximation or needs literal fidelity.
  • Confirm license rights and avoid close mimicry of real people or protected styles.
  • Add text and verify accessibility separately; never trust embedded text or skip alt text.
  • A second human review catches the convincing errors you have gone blind to.

Search Articles

Categories

OperationsSalesDeliveryGovernance

Popular Tags

prompt engineeringai fundamentalsai toolsthe difference between AIMLagency operationsagency growthenterprise sales

Share Article

A

Agency Script Editorial

Editorial Team

The Agency Script editorial team delivers operational insights on AI delivery, certification, and governance for modern agency operators.

Related Articles

General

Prompt Quality Decides Whether AI Earns Its Keep

Prompt quality is the single biggest variable in whether AI delivers real work or expensive noise. The model matters, the platform matters — but the prompt you write determines whether you get a first

A
Agency Script Editorial
June 1, 2026·10 min read
General

Counting the Real Cost of Every Token You Send

Tokens and context windows sit at the intersection of AI capability and operational cost—yet most business cases treat them as technical footnotes. That's a mistake that costs real money. Every time y

A
Agency Script Editorial
June 1, 2026·10 min read
General

Rolling Out AI Hallucinations Across a Team

Most teams discover AI hallucinations the hard way — a confident-sounding wrong answer makes it into a client deliverable, a legal brief, or a published report. The damage isn't just to the output; it

A
Agency Script Editorial
June 1, 2026·11 min read

Ready to certify your AI capability?

Join the professionals building governed, repeatable AI delivery systems.

Explore Certification