Where Synthetic Creators Fit—and Where Real People Still Matter

Devon Ariza 12 August 2026

AI UGC can make creator-style video easier to produce, revise, and localize. It can also create a category error: a synthetic actor may look like a customer, speak like a customer, and appear in a testimonial format without ever having used the product.

That distinction should shape every buying and production decision. AI UGC is best treated as a production and testing method—not as a substitute for genuine experience, audience trust, or creator relationships. Its value is primarily operational: more variants, fewer filming constraints, and faster iteration. Its limits become clear when the message depends on lived experience, physical proof, community context, or personal credibility.

What AI UGC Actually Means

AI UGC generally means synthetic or AI-assisted content designed to resemble creator-made social video. Typical formats include reviews, testimonials, demonstrations, unboxings, first impressions, comparisons, app walkthroughs, and direct-to-camera ads. Vendors use the term for videos generated from prompts, reference images, product assets, scripts, synthetic actors, and artificial voices. LTX Studio, for example, defines AI UGC as user-generated-style content created with AI rather than real creators.

The label is useful shorthand, but it can also mislead. Fully synthetic content is not literally generated by a user or customer. It borrows the visual language of UGC: handheld framing, conversational delivery, casual settings, captions, quick hooks, and an apparently personal point of view.

Four categories are often grouped together even though they carry different production, rights, and trust implications:

Category	Who or what appears	What it can legitimately represent
Genuine customer content	A real customer, usually speaking voluntarily	That customer’s actual experience, subject to permission and claim review
Paid human creator content	A creator hired and briefed by a brand	The creator’s performance and any genuine experience they are qualified to describe
Fully synthetic AI UGC	A generated actor, voice, scene, or combination	A scripted brand message or visualization—not lived product experience
AI-assisted human footage	A real person’s recording enhanced with editing, dubbing, translation, captions, or localization	The underlying human performance, provided the edits do not materially misrepresent it

The third and fourth categories deserve particular separation. Translating an authorized creator video with AI is not the same as inventing a presenter. Nor is removing pauses or adding captions equivalent to generating a testimonial from scratch.

An avatar can read a supportable claim such as “This app includes a weekly planning view.” It cannot truthfully say, “I have used this every morning for six months and it changed my life,” because the avatar has no lived experience. If a real person supplied the underlying experience, the brand must still consider whether the synthetic presentation accurately communicates who had that experience and whether the relevant permissions cover the intended use.

The central distinction is therefore creator-style presentation versus genuine user testimony. A video can look informal, personal, and social-native without containing any customer experience at all. Teams that preserve that distinction will write more accurate scripts, choose safer use cases, and avoid treating visual realism as evidence.

How an AI UGC Video Moves From Brief to Export

Most AI UGC generators follow a recognizable workflow:

Submit a product URL, product image, existing video, reference asset, or campaign brief.
Select a stock avatar, generate a synthetic character, or use an authorized digital likeness.
Enter a script or ask the system to draft one.
Choose the voice, setting, tone, language, and output format.
Generate an initial render.
Review and rerender weak sections.
Add captions, B-roll, music, transitions, or manual edits.
Localize and conduct a language and cultural review.
Obtain internal approval.
Export the usable file.

A workable brief should identify the product, audience, campaign objective, target benefit, evidence available for that benefit, desired setting, brand tone, voice, language, format, and call to action. A clean product image and accurate source information may reduce visual errors, but no input method eliminates the need for review.

A simple short-form script structure is:

Hook: Identify the problem, desired outcome, or point of curiosity.
One supportable benefit: Explain one useful feature or advantage without overloading the clip.
Call to action: Tell the viewer what to do next.

One vendor-authored workflow recommends a 15–20-second script and a 9:16 vertical export. Those are practical starting points for short-form social placements, not universal standards. The appropriate duration and aspect ratio still depend on the placement, product, message, and test design. The same workflow recommends changing one element at a time when creating variants, such as the hook or call to action. Magnific’s published process covers product inputs, scene generation, scripting, localization, review, and export.

Common controls include appearance, clothing, background, voice, tone, pacing, captions, music, B-roll, language, and aspect ratio. More advanced workflows may add emotion direction, product handling, reference-motion recreation, reusable characters, clothing visualization, or app demonstrations.

Avatar choice introduces three different production and rights models:

Stock avatar: A ready-made presenter supplied by the platform. It is easy to deploy, but the buyer should review the provider’s terms to understand permitted commercial uses and any plan or media restrictions.
Custom synthetic character: A newly generated persona that is not intended to reproduce a specific real person. Teams should test whether its appearance remains consistent across scenes and campaigns.
Licensed digital twin: A synthetic representation of a real person. Do not assume an ordinary appearance release covers AI replication. As a risk-management measure, the agreement should expressly address the intended AI use, likeness and voice processing, modification, media, territory, duration, compensation, termination, retention, and deletion. Vendor legal guidance likewise recommends obtaining permission that specifically covers AI use when a real person’s image, voice, or likeness is involved.

Generation is not the end of production. Every output needs a mandatory quality-control pass covering:

Product shape, color, packaging, dimensions, and visible labels
Prices, feature descriptions, and objective claims
App interfaces, button locations, and depicted workflows
Hands, fingers, object contact, shadows, reflections, and continuity
Lip sync, facial movement, eye line, and body motion
Brand, product, and place-name pronunciation
Caption accuracy and on-screen text
Music, voice, font, and stock-asset permissions
Cultural fit, idiom, humor, gesture, and local expectations
Any implied use, satisfaction, endorsement, or result

A polished render can still be commercially unusable because a label is malformed, a phone interface is fictional, a hand passes through the product, or the voice stresses a brand name incorrectly. Rejected renders, rerenders, editing time, localization review, approvals, and export limits are part of production—not exceptions to it.

This quality-control framework is a practical recommendation for evaluating output. It is not an established industry standard, and teams should adapt it to their product, market, and risk level.

AI UGC Versus Human Creator UGC

The useful comparison is not “cheap machines versus expensive people.” Each production model supplies a different form of value.

Decision factor	AI UGC	Human creator UGC
Production speed	Often faster after the workflow is established	Requires sourcing, scheduling, filming, and review
Variant volume	Strong for repeated hooks, voices, scripts, and backgrounds	More expensive and logistically demanding at high volume
Localization	Can reuse a structure across languages	May require local creators, dubbing, or separate shoots
Creative control	High control over script and presentation	Briefed, but delivery includes human interpretation
Lived experience	None unless accurately based on an authorized person’s real experience	Can include genuine use and personal context
Emotional credibility	Simulated performance	Can draw on spontaneous reactions and personal history
Product handling	Can visualize simple interactions, with error risk	Better suited to tactile, precise, or complex demonstrations
Audience reach	No built-in community unless distribution is arranged separately	Some creators bring an audience and social context
Community value	Limited	Can support relationships, comments, advocacy, and organic content
Rights administration	Platform terms and asset licenses require review	Creator agreements may cover usage, whitelisting, exclusivity, and revisions

AI’s clearest advantages are operational. It can reduce dependence on casting, product shipping, location access, and talent schedules. A team can keep the structure constant while testing a new hook, voice, language, or background.

Human creators offer a different bundle: talent, physical product interaction, spontaneous delivery, personal credibility, and sometimes distribution to an existing audience. Their pricing may also include value that does not exist in an AI render, such as paid-media rights, whitelisting, exclusivity, negotiated revisions, raw footage, product handling, or publication through the creator’s account. Human production can involve sourcing, negotiation, shipping, revisions, usage rights, and whitelisting rather than a single video fee. Sepia’s comparison describes these creator-production components while also positioning AI as a method for high-volume testing.

These commercial details are part of the broader working economics of UGC. Larping Agency’s editorial focus on creator rates, contracts, usage rights, whitelisting, exclusivity, and briefs reflects why a creator quote cannot be compared fairly with a bare generation fee.

Neither format needs to replace the other. Allocation should depend on the campaign objective, product risk, channel, and relationship the brand wants with its audience.

A practical rule of thumb is:

Use AI when the main problem is testing capacity. Use people when real experience or trust is part of the claim.

A hybrid system can use AI to explore hooks and positioning quickly, then brief human creators on promising concepts. That transfer still needs testing: a script may perform differently when the presenter, setting, timing, physical demonstration, and perceived credibility change.

Good Fits, Weak Fits, and Red-Flag Uses

AI UGC works best when the job is controlled communication rather than personal evidence.

Good fits include:

Paid-social hook and call-to-action testing
Frequent creative refreshes
Product-launch introductions
App and software explainers
Marketplace product videos
Simple feature summaries
Multilingual campaign variants
Early storyboard or concept visualization

These uses benefit from controlled scripts, repeatable presenters, quick revisions, and matched variations. An app company, for example, can create several versions of a short feature explanation while holding the interface footage, offer, and call to action constant.

Weaker fits include:

Genuine customer testimonials
Founder stories
Community-led organic content
Complex physical demonstrations
Products whose fit, texture, assembly, or performance must be shown accurately
High-consideration purchases
Messages relying on personal history, identity, or long-term use
Brand-building campaigns centered on creator relationships

A generated presenter can perform a scripted visual sequence. It cannot supply genuine satisfaction, product use, clinical improvement, financial success, or another result it never experienced. Adding an AI label does not make an unsupported testimonial true.

Healthcare, wellness, finance, and other regulated or trust-sensitive categories warrant stronger substantiation and appropriate legal review. The central issue is not merely whether the presenter looks realistic. It is whether the script, imagery, and context imply an endorsement, result, or body of evidence that does not exist. Vendor guidance itself warns that simulated testimonials should not make unsupported claims or imply untrue experiences, and that disclosure or platform rules may vary.

Product visualization must also be separated from product evidence. A generated garment moving naturally does not establish its real fit. A synthetic app screen does not prove that the live interface works that way. A fictional unboxing does not verify what arrives in the package. Treat generated visuals as illustrations until they have been checked against the actual product.

The Real Cost Is Per Approved Asset, Not Per Render

Subscription price and advertised generation cost are easy to compare because they are visible. They are also incomplete.

A more useful calculation is:

Cost per approved asset = (subscription and credits + scripting + prompting + failed generations + rerenders + editing + localization review + approvals + export costs + human labor + allocated testing costs) ÷ approved usable assets

This is a proposed operating formula, not a standardized accounting rule. Its purpose is to prevent teams from treating every generated file as a finished commercial asset.

Suppose a plan produces 30 renders. If only 12 meet the brief, and each approved clip requires editing and review, the operational denominator is 12—not 30. A cheap render becomes expensive when approval rates are low or corrections consume staff time.

Credit systems add another complication. Model choice, video duration, resolution, voice generation, and retries may consume different amounts. A monthly allowance therefore does not guarantee a fixed number of approved videos.

Pricing snapshot: The figures below reflect the vendor pages supplied for this article’s August 2026 review. They are promotional, plan-dependent snapshots rather than durable price guarantees. Buyers should verify current prices, taxes, billing periods, credit rules, and restrictions before purchasing.

MakeUGC displayed promotional monthly prices of $59 for 500 credits, $79 for 1,000 credits, and $149 for 2,000 credits. The vendor says credits refresh each billing cycle, do not roll over, and vary in cost by model, duration, and resolution. MakeUGC’s product and pricing page describes the displayed plans and credit conditions.

CreateUGC displayed annually billed equivalents of $19.90, $29.90, and $49.90 per month after a three-day trial. Some plans publish estimated monthly output, but the vendor states that actual output depends on the AI model selected. Higher plans also list different avatar, duration, language, and product-in-hand allowances. CreateUGC’s plan table explains its displayed prices, credits, and model-dependent output estimates.

Neither structure can be compared fairly with human creator production by looking at the smallest visible number. Creator costs may include sourcing, briefing, shipping, talent, filming, revisions, raw files, paid usage, whitelisting, exclusivity, and audience access. AI costs may include subscriptions, prompt development, rejected generations, manual editing, quality assurance, rights review, and staff time.

The units must match. Do not compare a 10-second unedited AI render with a 30-second edited creator deliverable. Do not include project-management labor on the human side while treating AI prompting and review as free. Do not use production cost alone to infer advertising effectiveness.

Track at least:

Cost per approved asset
Render-to-approval rate
Average retries per approved asset
Editing minutes per asset
Localization-review minutes
Time from brief to approval
Cost per tested concept
Cost per winning concept
Media spend consumed before a decision
Rights and licensing costs by channel and duration

These measures show whether the tool is reducing the cost of learning or merely generating more files.

How to Compare AI UGC Generators Without Falling for the Sales Page

There is no defensible universal “best AI UGC generator” in the available evidence. A buyer creating app walkthroughs has different needs from a clothing brand, a multilingual agency, or a team developing an authorized digital twin.

Start with the use case, then compare three layers.

1. Avatar and visual-production options

Ask whether the product provides:

A stock-avatar library
Custom synthetic characters
Licensed digital-twin creation
Product-in-hand support
App or screen presentation
Clothing visualization
Emotion and gesture controls
Consistent reusable characters
Reference-image and reference-motion inputs

2. Creative and editing controls

Compare:

Manual script entry and AI script generation
Product-link ingestion
Product and scene reference assets
Voice, tone, pacing, and emotion
Language support and translation
Lip synchronization
Maximum duration
Aspect ratios
Captions and text styling
B-roll, music, transitions, and trimming
Export resolution and file type

3. Operational and contractual controls

Look beyond the demo output and ask about:

Team collaboration and approval workflows
Automation and API access
Download formats and resolution
Trial limits and watermarks
Credit expiry and rollover
Concurrent generation
Model-dependent usage
Treatment of failed renders
Cancellation and post-cancellation access
Project and asset export
Uploaded-file retention and deletion
Support response for rights or safety issues

The following comparison is limited to capabilities advertised on vendor pages reviewed for this article. It is not an independent quality ranking.

Vendor	Advertised capabilities in the supplied page	Important qualification
MakeUGC	More than 1,000 actors, product-in-hand output, custom actors, motion recreation, captions, music, B-roll, and trimming	Counts and features are vendor claims; availability can vary by plan or model
HeyGen	More than 1,100 ready-made avatars and more than 175 languages and dialects; appearance, outfit, background, voice, tone, translation, and lip-sync controls	Breadth does not establish natural delivery or cultural accuracy. HeyGen lists these advertised avatar and language options
CreateUGC	Product-link or script-based generation, multiple aspect ratios, multilingual options, and product-in-hand output on some plans	Output estimates and features depend on the selected plan and model
Arcads	More than 1,000 actors, custom actors, product, app and clothing presentation, emotion direction, and localization in more than 30 languages	One-click editing for B-roll, music, captions, and transitions was marked “Soon!” rather than available. Arcads distinguishes current features from the forthcoming editing tool
Creatify	More than 140 languages, captions, music, editing, and exports for TikTok, Instagram, Facebook, and YouTube	Export destinations and language options do not prove platform compliance or local naturalness. Creatify lists these advertised production capabilities

Before buying, ask:

What commercial-use rights attach to generated footage?
How does the provider describe authorization for stock avatars and voices?
Can a custom avatar and its source files be deleted permanently?
What happens if the person behind a digital twin withdraws permission for future work?
How long are product images, voice samples, and reference videos retained?
Are uploaded assets or outputs used for model training?
Which voice, music, font, and stock-media licenses apply?
Can reference-ad recreation create copyright, trademark, or passing-off concerns?
What indemnity, if any, does the provider offer?
What remains accessible after cancellation?
Do rights differ by plan, model, geography, or export method?

These are due-diligence questions, not statements that every provider grants or withholds a particular right. Governing terms—not a sales-page ownership slogan—determine what the provider promises, while third-party likeness, voice, music, trademark, copyright, and publicity issues may still require separate review.

Avatar counts, language counts, generation speed, and sales-page adjectives do not establish product fidelity, natural acting, cultural accuracy, legal clearance, or commercial suitability. The available evidence supports a comparison of advertised capabilities, not confirmation of every vendor’s contractual protections or a definitive quality ranking.

What the Performance Evidence Does—and Does Not—Show

No broad independent benchmark in the available evidence proves that AI UGC universally outperforms human creator video—or the reverse.

The most direct comparison in the supplied material is an unnamed advertiser test reported secondhand by Playcut, a company that sells AI UGC tools. The report says the Meta campaign covered $100,000 in spend, 220 creatives, and three months. Human UGC recorded a 2.4% click-through rate, compared with 1.9% for AI creative, while AI recorded 2.8× ROAS, compared with 2.3× for human UGC. The original methodological appendix, campaign setup, attribution rules, and advertiser calculation were not available in the supplied evidence. Playcut says the advertiser attributed part of AI’s ROAS advantage to lower production costs. Playcut’s review reports the figures and acknowledges the first-party and methodological limitations.

This should be treated as a limited, commercially interested report—not an industry benchmark. It does not produce a simple winner. Instead, it illustrates that different metrics answer different questions:

Click-through rate asks how frequently impressions produced clicks.
Conversion rate asks how frequently visits or clicks produced the target action.
CPA asks how much media spend was required per acquisition.
Production cost asks what it took to make the creative.
ROAS compares attributed revenue with the spend included in the advertiser’s calculation.
Production-adjusted economics incorporate the cost of making and replacing assets.

Lower production expense can improve campaign economics even when viewers click less often. That does not prove stronger audience preference for synthetic presenters. Conversely, a higher click-through rate does not guarantee better profitability if traffic quality, conversion rate, order value, media cost, or production expense differs.

Creator-style advertising statistics also require careful classification. A comparison between casual creator-style video and polished brand creative is not an AI-versus-human test unless the study specifically changes the production source while controlling other material variables.

Vendor testimonials, logos, dashboard screenshots, and case-study summaries should be treated as promotional evidence unless the publisher supplies methods, baselines, sample definitions, attribution rules, and independent verification. They can suggest a hypothesis; they cannot establish a typical result.

The practical conclusion is to test AI and human work against the brand’s own offer, audience, funnel, and production process. Imported ROAS promises are not substitutes for a controlled campaign.

A Fair Test for AI and Human UGC

A useful test begins with a hypothesis, not a folder of unrelated videos. Examples include:

A problem-led hook will produce more qualified clicks than a benefit-led hook.
A human creator will improve response for an experience-based claim.
A synthetic presenter will enable lower-cost localization without reducing conversion.
A demonstration will outperform direct-to-camera explanation.
A particular language or voice will improve response in a target market.

Hold the following constant where possible:

Core script or message
Offer and price
Call to action
Duration
Aspect ratio
Placement
Audience
Spend or delivery rules
Landing page
Attribution window
Production-quality threshold

Change one major variable at a time. If one version changes the hook, actor, duration, music, captions, offer, and landing page, the result will not explain which change mattered.

Separate response metrics from production metrics.

Delivery and response metrics:

Impressions and reach
Video starts and completion
Click-through rate
Landing-page conversion rate
CPA
Revenue and ROAS
Conversion quality, refunds, or downstream retention where relevant

Production metrics:

Subscription and generation cost
Human labor time
Number of retries
Approval rate
Editing time
Time to launch
Cost per approved asset
Cost per tested concept

Do not declare a winner from a tiny spend or an early fluctuation. Set a predetermined budget, duration, sample threshold, or business decision rule appropriate to the campaign. The purpose is not to wait for perfect certainty; it is to prevent the team from promoting whichever screenshot looks best on day one.

Add a transfer test to a hybrid workflow. If an AI-presented concept wins, commission a human creator version using the same core message, then test it. The presenter, setting, pacing, and credibility will change, so performance may not transfer automatically.

Keep losing concepts and quality-control notes. A weak result may reflect a poor idea, unnatural voice, malformed product, slow opening, inaccurate claim, or mismatch between presenter and category. Without those records, the team may wrongly conclude that an entire production format failed.

This testing framework is a practical recommendation developed for campaign evaluation. It should be adapted to the team’s measurement capability, media budget, and decision risk.

Disclosure, Testimonials, Likeness Rights, and Copyright

Synthetic testimonial-style content creates a transparency risk because viewers may interpret a generated performance as the statement of a real customer. Disclosure can help reduce that risk, but it does not cure an inaccurate claim, invented experience, or unauthorized use of a person’s identity.

Before publication:

Confirm that every objective product claim is supportable.
Remove language implying unverified use, satisfaction, outcomes, or endorsement.
Do not call a synthetic actor a customer, user, reviewer, patient, investor, or client unless that description is true.
Review the full visual context, not only the spoken script.
Escalate ambiguous claims rather than relying on realism or disclaimers.

Vendor-authored legal guidance for AI UGC recommends truthful product claims, disclosure where required, current platform review, and appropriate permission before using a real person’s image, voice, or likeness. It also warns against simulated testimonials that imply experiences that are untrue. These recommendations are general rather than jurisdiction-specific legal authority.

Before cloning or materially reproducing a real person’s image, voice, likeness, or performance, obtain permission appropriate to the planned commercial use. Do not assume a conventional release necessarily covers synthetic replication. A prudent agreement should expressly address AI processing, modification, approved media, duration, territory, compensation, review or approval rights where negotiated, source-file retention, deletion, and what happens after termination.

Provider terms also need direct review. Check the provisions governing stock avatars, custom voices, uploaded product assets, generated footage, music, fonts, reference-ad recreation, retention, deletion, model training, cancellation, and indemnity. Treat statements that a customer “owns” an output as a prompt to read the governing contract—not as proof that all third-party rights have been resolved.

Disclosure requirements vary by jurisdiction, platform, format, and claim type, and they can change. Teams should consult current primary regulator guidance, applicable law, and first-party platform policies before launch rather than relying on a static article or universal disclosure phrase. This article does not establish whether a specific campaign requires a particular label.

U.S. copyright treatment also requires caution. Entirely AI-generated material may face human-authorship limitations, while potentially protectable human contributions—such as original writing, selection, editing, or arrangement—are assessed on their facts. It is unsafe to assume either that every AI output is protected or that routine human review automatically secures copyright. TechCXO’s overview discusses the U.S. human-authorship issue and recommends documenting meaningful human creative involvement.

Keep records of:

Prompts and script versions
Source images, footage, music, and licenses
Model and product versions
Generated drafts
Manual edits and arrangement decisions
Claim substantiation
Likeness and voice permissions
Internal approvals
Final exports and publication dates

Escalate for qualified legal review when a project involves regulated claims, cloned real people, customer-result statements, before-and-after claims, imitated ads, sensitive personal data, or unclear commercial rights.

This section provides general risk-management information, not jurisdiction-specific legal advice. Rates, licensing terms, and platform policies can also change; the site’s terms and editorial disclaimer recommend reviewing contracts and consulting a professional before agreeing to rights or exclusivity terms.

Conclusion

Treat AI UGC as a production method, not a shortcut to trust. Its strongest case is the ability to test, revise, and localize creator-style concepts with fewer filming constraints. Its weakest case is any execution that depends on a real person’s experience, relationship, or credibility.

The practical model is selective and hybrid. Calculate cost per approved asset, not cost per render. Test synthetic and human work under comparable conditions. Move promising ideas into the format best suited to the claim. Retain real people where physical demonstration, lived experience, community context, or personal trust is central. Resolve claims, disclosure, likeness, licensing, data, and platform questions before publication.

Frequently Asked Questions

Is AI UGC really user-generated content?

Not in the literal sense when the presenter, voice, and experience are synthetic. It is more accurately described as AI-generated creator-style content or synthetic UGC-style video.

The label “AI UGC” describes the format’s visual and narrative conventions rather than its source. Genuine UGC comes from an actual user; paid creator UGC comes from a human performer; fully synthetic AI UGC comes from a production system. Brands should not blur those categories when describing a testimonial or customer experience.

Can AI UGC replace human UGC creators?

It can replace some production tasks, especially repetitive script delivery, localization, variant creation, and early paid-social testing. It cannot automatically replace real product experience, physical dexterity, emotional spontaneity, creator reach, community relationships, or audience trust.

The strongest model is often hybrid: use AI to investigate hooks and messages, then hire people for executions where experience, demonstration, or credibility matters.

How much does an AI UGC video cost?

There is no reliable universal per-video price. Costs depend on the subscription, credit model, generator, duration, resolution, voice, editing requirements, retries, staff time, and approval rate.

Calculate the cost per approved usable asset rather than the advertised cost per render. Include subscriptions, credits, scripting, prompting, failed generations, rerenders, editing, localization review, approvals, labor, export restrictions, and media testing.

Do AI UGC videos need to be disclosed?

The answer depends on the jurisdiction, platform, format, content, and claim. Rules and platform policies can change, so current primary guidance and appropriate legal advice should be checked before publication.

Independently of whether a particular label is mandatory, marketers should consider whether a reasonable viewer could mistake the synthetic performance for a real customer statement. Disclosure does not make false claims, invented use, or unauthorized likeness cloning acceptable.

Can an AI avatar give a product testimonial?

An AI avatar can deliver a scripted product description or clearly framed dramatization. It cannot truthfully claim personal use, satisfaction, or results because it has no lived experience.

If the script says “I tried this,” “this worked for me,” or otherwise implies a real endorsement, the underlying experience must be genuine, properly authorized, supportable, and presented without misleading viewers about who supplied it. For most synthetic-avatar campaigns, factual product explanation is safer and more accurate than first-person testimony.