AI Ad Authenticity Checklist & Quality Score

Claim evidence review

Can this marketing claim be traced to evidence?

Build a compact review packet for one public-facing claim. This evaluates documentation completeness, not whether the claim is legally permitted or scientifically valid.

Exact public claimSource URL or evidence recordSupporting passage, dataset, method, or ownerEvidence dateLanding page presents the same claim and limitsNecessary qualification

How should I document evidence for a marketing claim?

Record the exact public claim, a retrievable source, the supporting passage or method, evidence date, necessary qualifications, and confirmation that the landing page carries the same promise and limits.

01 · Eight-part review

Calculate an AI-generated ad quality score.

Choose zero, one, or two for every criterion. Score the final ad as viewers will see it, including copy, image or video, disclosure, destination, and CTA.

Original human inputIs there a real observation, brief, concept, or judgment behind the output?

Product or process evidenceDoes the ad show what is sold, how it works, or what the buyer receives?

Supported claimsAre outcomes sourced, qualified, and consistent with the landing page?

Brand-specific voiceCould a competitor publish the same words without changing anything?

DisclosureAre synthetic or sponsored elements labeled wherever rules require it?

Consent and identity safeguardsAre real people, voices, testimonials, and likenesses authorized and accurate?

Visual plausibilityDo hands, labels, interfaces, products, and physical relationships remain credible?

Proportional CTADoes the next step match the evidence and commitment level?

02 · What quality means

AI ad quality is broader than polished execution.

Google’s Ad Strength guidance emphasizes relevance, quality, and diversity. Meta recommends diverse formats, placements, and creative concepts. Those production signals matter, but a publishable ad also needs truthful claims, brand ownership, consent, and a destination that fulfills the promise.

First-party references: Google Ads: About Ad Strength and Meta: Expand your ad creative strategy.

03 · Stop conditions

Some failures cannot be averaged away.

Do not publish merely because the total score is high. A single serious integrity problem can make the entire ad unsafe, misleading, or unusable.

Claims

Fabricated proof

Invented statistics, testimonials, customer logos, certifications, scarcity, or performance outcomes require removal—not softer wording.

Identity

Unauthorized likeness or voice

Do not simulate a real person, customer, employee, expert, or creator without appropriate rights and clear context.

Product

Misrepresented interface or result

A generated screen, package, ingredient, capability, or before-and-after result must not imply behavior the actual product cannot deliver.

Destination

Promise mismatch

The landing page must contain the offer, terms, availability, and evidence the ad leads viewers to expect.

04 · Review sequence

How to review an AI-assisted ad before publishing.

1
Freeze the source evidence
List the approved product facts, proof, offer, audience, and claims before evaluating the generated artifact.
2
Inspect copy and pixels separately
Check every word, label, logo, product detail, person, and implied physical relationship at full size.
3
Compare the destination
Confirm price, terms, CTA, availability, and promise are consistent from ad through landing page.
4
Record the approved version
Publish only the exact reviewed artifact. Regeneration creates a new version that needs another review.

FAQ

AI-generated ad quality questions.

Does a high score mean the ad will perform?

No. The score reviews readiness and authenticity, not predicted auction performance. Test the ad against a clear objective and business metric.

Does every AI-generated ad require disclosure?

Requirements vary by platform, jurisdiction, subject, and use of synthetic people or events. Check current placement rules and disclose material facts clearly where required.

Can AI-generated visuals be authentic?

Yes, if they represent the product and context honestly, respect rights and consent, and do not fabricate evidence or real-world events.