Lapis, the YC-backed #1 AI ad generator rated 4.9/5 on G2, uses AI agents to create, launch, and optimize ads on ChatGPT, Google, Meta, and LinkedIn.
The AI ad generation mistakes that waste the most budget in 2026 are strategic rather than technical: variations instead of angles, one platform's creative pasted everywhere, invented proof, and ad platforms grading their own results. Lapis is built to prevent all four, because it writes from a brief, tests distinct angles as experiments, builds a page for every ad, and measures confirmed sales.
Generating ads with AI is now the default, since IAB found on January 15, 2026 that 83% of ad executives say their company has deployed AI in creative. Meta said on October 21, 2026 that nearly 10 million advertisers used its generative AI creative tools in the third quarter. Consumers have noticed: the Bentley University-Gallup survey published August 19, 2026 found 79% of Americans had recently seen ads that looked AI-made, and 49% view the practice negatively against 19% positively.
I built Lapis after watching good brands lose money on ads that looked impressive and taught them nothing. The twenty mistakes below are grouped by where the money leaks, from strategy and creative to compliance and measurement, and each one comes with its fix. A 20-point QA checklist, a worked cost example, and the workflow that avoids all twenty follow, and the companion comparison of ChatGPT vs Gemini vs Lapis for making ads covers which tool fits each job.
What Are the Most Common AI Ad Generation Mistakes?
The most common AI ad generation mistakes fall into four groups: strategy mistakes that make ads forgettable, creative mistakes that make them look generic, compliance mistakes that get them rejected or fined, and measurement mistakes that hide which ads sold.
| Group | Mistakes | What it wastes | Where it is fixed |
|---|---|---|---|
| Strategy | 1 to 5: no brief, variations instead of angles, one creative everywhere, a vague offer, clicks over revenue | Every dollar behind a forgettable message | The brief |
| Creative | 6 to 10: the generic AI look, garbled text, off-brand assets, ignored safe zones, broken character limits | Click-through, trust, and review time | Generation and QA |
| Compliance | 11 to 15: fake testimonials, unlabeled synthetic people, unsupported claims, likeness misuse, disclosure rules | Rejections, fines, and brand damage | Claims and label review |
| Measurement | 16 to 20: near-identical variations, early winners, no holdout, homepage traffic, self-graded results | Test budget and scaled losses | Test design and attribution |
The trust context raises the stakes on every group, because Cint reported on October 20, 2026 that 74% of US consumers believe brands have a duty to disclose AI use in advertising. Cint also found 44% would trust a brand less on discovering undisclosed AI, and Auros reported on October 16, 2026 that 53% of Americans do not trust products advertised with AI to be described accurately. An AI ad that looks synthetic, overpromises, or hides its origin buys distrust, which is why understanding how AI ad generators work matters as much as choosing one.
Which Strategy Mistakes Make AI Ads Fail?
Strategy mistakes make AI ads fail before a single image renders, because a model given a thin brief writes the category average, and the category average is what every competitor's AI is writing as well.
1. Generating Without a Brief or Positioning
Prompting from a blank box is the root mistake, because the model fills every missing input with an average that any brand in the category can run unchanged. IAB's January 2026 research advised marketers to "use AI to enhance creative quality, rather than to produce advertising assets more cheaply."
Write a one-page brief before generating anything: the buyer's situation in their words, the objection you answer, your proof, the offer terms, the landing page, and the confirmed metric. The AI ad creative brief template lists the inputs, and the 15 ways to write better AI ad prompts turns them into prompts.
2. Generating Variations Instead of Angles
Asking for 20 ads returns 20 versions of one idea, and Meta's ads system now treats that pattern as a single idea. Its December 16, 2025 guidance, "Demystifying Creative Diversification," says ads sharing key visual and thematic attributes are grouped as variations of one creative.
Meta reported in the same post that campaigns using its image generation saw 11% higher CTR and 7.6% higher conversion rates as advertisers leaned into diversification. Generate three to six angles first, each a different reason to buy, cut any two that share a promise, and use AI ad angles vs variations for the prompt that forces the difference.
3. Pasting One Platform's Creative Everywhere
A Facebook Feed ad pasted into Reels, LinkedIn, and ChatGPT fails four different ways, because each placement has its own frame, keep-out zones, limits, and reading context. Reels covers the bottom 35% of a 9:16 frame, LinkedIn truncates headlines past 70 characters, Google stops search headlines at 30, and OpenAI recommends ChatGPT ad titles of 16 to 24 characters.
Build a placement set per angle rather than one master asset, and adapt the message to each channel as well as the dimensions. The guide to AI for Facebook and Instagram ads and the ChatGPT ad image specs cover the two placements teams most often get wrong.
4. Ignoring the Offer
A model given a product but no offer writes generic benefits, and a model given an offer without its terms drops the dates, eligibility, and exclusions that make it true. Either way the ad gives no reason to act today or promises more than checkout delivers, and an offer is a claim held to the same substantiation standard as any other.
Put the full offer in the brief and require every ad that mentions it to name the end date and match the landing page word for word. The breakdown of how much it costs to make an ad shows why a cheap ad with a broken offer is the most expensive one you will buy.
5. Optimizing for Clicks Instead of Confirmed Revenue
AI makes click-through rate easy to raise, which is exactly why it is the wrong scoreboard for AI creative. A Google Display Network field study by NYU Stern and Emory researchers, revised December 11, 2025, found fully AI-created ads lifted click-through by up to 19% over human-made ads.
Neither that lift nor any other CTR tells you which ad produced a profitable customer. Name the confirmed conversion in the brief, tell the model to ignore CTR, and read every test against your store or CRM through Lapis Wayfinder, the same discipline the ROI measurement framework applies to ChatGPT Ads.
Which Creative Mistakes Make AI Ads Look Generic?
Creative mistakes make AI ads look generic or wrong, and audiences punish both. Taboola's January 28, 2026 study found ads perceived as AI-generated underperformed regardless of who made them, and Cint found on October 20, 2026 that 46% of US consumers dislike AI-generated people in ads.
6. The Generic AI Look and Uncanny People
The generic AI look, with flawless stock faces, impossible lighting, and products floating nowhere in particular, signals low effort, and synthetic people carry the heaviest penalty. The Bentley-Gallup survey of August 19, 2026 found 62% of Americans call AI-created people or voices in ads unacceptable even when disclosed, rising to 73% among adults aged 18 to 29.
Taboola found a large, clear human face was among the strongest signals that made an ad feel trustworthy, and AI ads that did not look AI-made earned the highest engagement of any group. Use real customers and licensed creators for faces and real product photos for scenes, and see the AI vs human ads performance data and 10 ways to scale UGC campaigns with AI.
7. Garbled Text and Wrong Product Details
A misspelled headline or a product with the wrong label or color turns a promising ad into evidence of carelessness, and consumers already expect the worst. Auros reported on October 16, 2026 that 43% of Americans have felt duped by products advertised with AI, and only 20% correctly spotted real products heavily enhanced with AI.
Generate the product from reference photos rather than a description, proofread every in-image word at full zoom, and follow Google's documentation, which recommends generating the text before the image. Regenerate any frame with a wrong character instead of patching it, because a swapped letter is the error people screenshot.
8. Off-Brand Colors, Fonts, and Logos
A model asked to "add our logo" draws an approximation, and an approximation of your own trademark in a paid ad is worse than no logo at all. Color and type drift the same way, so a brand blue three shades off and a substituted font look counterfeit next to your real site.
Supply the logo file with an instruction to place it as supplied, give exact hex codes and a licensed font, and check every asset against the brand sheet. Lapis learns your logo, colors, voice, and audiences from your website, closing the brand-memory gap described in the ChatGPT, Gemini, and Lapis comparison.
9. Ignoring Safe Zones and Specs
AI models output a frame rather than a placement, so a neatly centered headline ends up under Meta's caption bar or cropped off a Gmail logo circle. Meta tells advertisers to keep the top 14%, bottom 35%, and 6% of each side of a Reels frame clear, and the bottom 40% when the ad carries required legal text.
Google crops 21.46% of a square Demand Gen logo when Gmail renders it as a circle, so keep marks away from the corners. Turn on Meta's safe zone guardrail, preview every placement on a phone, and use the 12 Meta Ads AI features and 12 Google Ads AI features to decide which automated crops to allow.
10. Copy That Breaks Character Limits
AI copy that runs past a field limit is rejected at upload or truncated mid-promise, and the limits are tight: 30-character Google search headlines, a 27-character Facebook Feed headline, 70 characters on LinkedIn, and 50 for a ChatGPT ad title. Google's editorial policy also rejects repeated punctuation and gimmicky capitalization, the two habits a prompt for "punchy" lines invites.
Require a character-count column, verify it with a counter rather than trusting the model, and ban repeated punctuation and all-caps emphasis outright. The ChatGPT ad copywriting guide shows how to write inside the tightest limits without losing the angle.
Which Compliance and Trust Mistakes Get AI Ads Rejected or Fined?
Compliance mistakes turn wasted budget into legal exposure, and the 2026 rules are specific: fake testimonials carry FTC civil penalties of up to $53,088 per violation, undisclosed synthetic performers are fined in New York, and platforms reject undisclosed AI in political ads.
11. Fake Testimonials and Synthetic Reviews
An AI-written testimonial from a customer who does not exist is illegal in the United States. The FTC's final rule on fake reviews and testimonials, announced August 14, 2024 and in force since October 21, 2024, bans them, AI-generated ones included, and the FTC's December 22, 2025 warning letters cite penalties of up to $53,088 per violation.
On April 13, 2026 the FTC charged supplement seller TruHeight over employee-written reviews and about 176 fake Facebook and Instagram profiles whose comments, a contractor wrote, were "generated through ChatGPT." The July 15, 2026 final order carried a $4 million judgment, partially suspended on payment of $750,000, and India's ASCI also bars fabricated testimonials in guidelines published September 29, 2026. Use only real customers' words with permission, and label AI avatars as AI.
12. Unlabeled Realistic Synthetic People
A realistic AI person or scene now needs a label in a growing list of places, and skipping it costs money. New York's synthetic performer law, in force since June 9, 2026, fines advertisers $1,000 for a first violation and $5,000 for each later one. TikTok rejects or restricts ads with fully AI-generated or significantly AI-modified media that lack a label, caption, watermark, or sticker.
Meta has auto-labeled detected AI media since June 1, 2026, Google shows visible labels on ads aimed at the EU, India, New York, California, South Korea, and Vietnam, and the EU AI Act has required deepfake labels since August 2, 2026. The IAB's August 2026 framework asks for labels where AI use is material, citing an NYU Stern finding that disclosure cut click-through 31.5%, so label synthetic people and realistic scenes and leave AI-drafted copy unlabeled.
13. Unsubstantiated Claims
A language model will write "clinically proven" or "#1 rated" whenever the prompt leaves room, and every such phrase needs evidence before the ad runs. The FTC requires a reasonable basis for advertising claims, OpenAI's ad tools terms make advertisers responsible for substantiating AI-generated claims, and the TruHeight case began with allegedly unsupported claims that its vitamins made children grow taller.
Give the model a proof library, forbid any claim outside it, and have it write [NEEDS PROOF] wherever a line needs a fact you did not supply. Cut every such line before launch, because the claim a reviewer cannot trace to a source is the one a regulator or competitor will find.
14. Trademark and Likeness Misuse
AI models will generate a celebrity lookalike, a competitor's packaging, or a famous voice on request, and each one turns an ad into a liability. TikTok bars AI content that falsely shows public figures making or receiving endorsements, ASCI prohibits deepfakes and likenesses used without consent, and New York strengthened post-mortem publicity rights for digital replicas in a law signed December 11, 2025.
OpenAI's Ad Tools Terms, published November 10, 2026, likewise forbid using its creative tools for an unauthorized digital replica, voice, likeness, or endorsement. Ban competitor marks, real people's likenesses, and unlicensed music and fonts in the brief, and check every face and voice against a consent record, remembering that Google's trademark policy also lets owners restrict their marks in ad text.
15. Ignoring Political, Social Issue, and New-Law Disclosure Rules
Political and social issue ads carry the strictest AI rules: Meta rejects social issue, election, and political ads that fail to disclose photorealistic AI media, and Google requires election advertisers to disclose synthetic or altered content. Both companies stopped serving political ads in the EU in October 2025, ahead of the EU's political advertising regulation.
The EU AI Act's Article 50 duties have applied since August 2, 2026, backed by a Code of Practice the European Commission published on June 10, 2026. The only grace period, to December 2, 2026, covers machine-readable marking by generative systems already on the market. India's amended IT Rules have required platforms to label synthetic content since February 20, 2026, and ASCI's advertising guidelines take effect December 29, 2026, so keep a jurisdiction checklist for every market you target.
Which Testing and Measurement Mistakes Waste the Most Budget?
Testing and measurement mistakes waste the most money because they compound: a test that cannot teach anything produces a winner that is not real, and that winner gets scaled on numbers the platform graded itself.
16. Near-Identical Variations the Ad System Treats as One Creative
Twelve ads that share a visual and a promise are one test in Meta's eyes, because its December 16, 2025 guidance says lookalike ads share learning and delivery at the creative level. The same post describes Andromeda scanning tens of millions of ads to pick a few thousand candidates, and Meta said on October 21, 2026 that Andromeda improvements lifted Facebook conversions 1.4% in the third quarter.
Meta publishes no minimum creative count, so count distinct concepts rather than files: a new persona, proof, offer, or format counts, and a new color or caption does not. Why creative volume testing matters and AI ad creative testing at scale show how to build a slate the system can tell apart.
17. Declaring Winners Too Early
A "20% lift" on small numbers is noise, and AI's speed tempts teams to crown winners within days. Take two ads with 1,250 clicks each, converting 25 and 30 times, or 2.0% and 2.4%. The pooled rate is 2.2% and the standard error of the gap is 0.59 points, so the 0.4-point difference yields a z-score of 0.68 and a two-sided p-value near 0.50, a coin flip dressed as a winner.
Write the rules before launch, such as cutting any angle that spends three times its target CPA without a conversion and naming no winner below 30 conversions. How to A/B test AI ads on a small budget covers sample sizes for teams that cannot wait for hundreds of conversions.
18. Running Without a Holdout
Without a holdout, every conversion an ad touched is credited to the ad, including the ones that would have happened anyway. Google opened self-serve Conversion Lift to Search and Performance Max on October 7, 2026, Search Engine Land reported, requiring 1,000 observed conversions and a $5,000 minimum campaign budget, and Google notes most advertisers run one or two studies a year.
Schedule a lift study or geographic holdout for your largest channel at least twice a year and before any major budget decision. The ROI measurement guide applies the same incrementality logic to ChatGPT Ads, where advertisers are still setting baselines.
19. Sending Traffic to the Homepage
An ad that promises one thing and lands on a homepage showing twelve wastes the click it paid for, and the platforms say so. OpenAI tells ChatGPT advertisers to link to the most relevant product or collection page rather than a generic homepage, and Google counts landing page experience as one of three Quality Score components.
Give every angle a page that repeats the ad's headline, offer, and proof in the first screen. Lapis RapidDomain builds that page for every ad and optimizes it on real conversions, and dynamic AI landing pages and the custom landing page evidence explain the mechanics.
20. Letting the Platform Grade Itself
Each ad platform reports its own conversions, credits its own AI, and never deduplicates a sale another platform also claims. Madison and Wall estimated, as Business Insider reported on March 18, 2026, that $57 billion of US ad revenue runs through AI-powered platforms in 2026, 12% of the market. Meta's GEM model delivered 18.9% more conversions by Meta's own count (October 13, 2026), and no platform reconciles its count with the others, so the advertiser who owns that reconciliation decides what really scales.
Reconcile every channel against store or CRM orders on one set of attribution rules, and scale only the lift your store or CRM confirms. The Lapis vs Advantage+ and Lapis vs Performance Max comparisons show how an independent layer sits above each platform's AI.
The AI Ad QA Checklist: What Should You Check Before Every Launch?
The AI Ad QA Checklist is a 20-point pre-launch test, one check per mistake above, and an ad ships only when it passes all twenty, because a single failure in claims, labels, or measurement can erase the gains from the other nineteen.
| # | Check | Pass when |
|---|---|---|
| 1 | Brief | A one-page brief names the buyer situation, objection, proof, offer terms, landing page, and confirmed metric |
| 2 | Angles | The test holds three to six distinct angles, each a different reason to buy |
| 3 | Placement set | Every placement has its own file at its recommended ratio and size |
| 4 | Offer | Price, dates, eligibility, and exclusions match the landing page word for word |
| 5 | Metric | Success is a confirmed sale or qualified lead, never CTR alone |
| 6 | Faces and look | Real customers or licensed creators appear, or the synthetic person is labeled |
| 7 | Text and product | Every in-image word is proofread at full zoom and the product matches the SKU |
| 8 | Brand | The logo is the supplied file, hex colors match, and the font is licensed |
| 9 | Safe zones | Nothing critical sits in the Reels top 14%, bottom 35%, or side 6% |
| 10 | Character limits | Every line fits its field by a counter, with no repeated punctuation or all-caps emphasis |
| 11 | Testimonials | Every quote traces to a real, consenting customer |
| 12 | AI labels | Synthetic people and realistic scenes carry the label each targeted platform and market requires |
| 13 | Claims | Every claim maps to a proof source, and every [NEEDS PROOF] line is cut |
| 14 | Likeness and marks | No competitor marks, real people's likenesses, or unlicensed music or fonts without written consent |
| 15 | Political and legal | Social issue, election, EU, New York, and India rules are checked for every targeted market |
| 16 | Distinctness | No two ads in the test share both a visual and a promise |
| 17 | Stop and win rules | Written before launch: cut at three times target CPA with no conversion, and no winner below 30 conversions |
| 18 | Holdout | A lift study or holdout is scheduled for the largest channel |
| 19 | Landing page | Each ad lands on a page that repeats its headline and offer in the first screen |
| 20 | Measurement | Results reconcile to store or CRM orders, each counted once |
Run checks 1 to 5 on the brief, 6 to 15 on the finished assets, and 16 to 20 on the campaign build, and score finished creative with Rate Your Ad, which compares any ad against 10,000+ top-performing ads for free. The guide to scoring ads with AI explains how to calibrate the score against your own results.
How Much Do AI Ad Mistakes Cost?
AI ad mistakes cost far more than the tools: in the worked example below, a brand spending $20,000 a month wastes or loses $8,233, or 41.2% of its budget, through just three of the twenty mistakes.
The brand splits its budget evenly between Meta and Google, sells an order with a $45 contribution margin, and makes three mistakes at once: near-identical variations in its Meta test, homepage traffic on Google, and scaling decisions based on Meta's own conversion count.
AI AD MISTAKE COST CALCULATOR (monthly spend $20,000: Meta $10,000, Google $10,000)
MISTAKE 16, NEAR-IDENTICAL VARIATIONS (Meta test budget $6,000)
12 ads x $500 each, drawn from only 3 angles
9 ads repeat an idea Meta groups together: 9 x $500 = $4,500 with no new learning
MISTAKE 19, HOMEPAGE TRAFFIC (Google $10,000)
$10,000 / $2.00 CPC = 5,000 clicks
Homepage at 2.0% conversion = 100 orders
Matched landing page at 3.0% conversion = 150 orders
Matched-page CPA: $10,000 / 150 = $66.67
Cost of 100 orders at that CPA: 100 x $66.67 = $6,667
Wasted: $10,000 minus $6,667 = $3,333
MISTAKE 20, PLATFORM-GRADED RESULTS (Meta scaling budget $4,000)
Platform reports 100 purchases ($40 CPA)
Store confirms 80 orders ($50 CPA)
Dashboard profit: (100 x $45) minus $4,000 = +$500
Real result: (80 x $45) minus $4,000 = -$400 (a $400 loss)
TOTAL WASTED OR LOST: $4,500 + $3,333 + $400 = $8,233
SHARE OF MONTHLY SPEND: $8,233 / $20,000 = 41.2%The third line is the dangerous one, because the dashboard reports a $500 profit on a campaign that loses $400, and the natural response to a profitable campaign is to scale it. Fixing all three costs nothing but discipline, which is why a media buyer, an agency, or an AI employee should be judged on the waste they prevent. The comparisons in hire a marketer or use AI and the 15 signs your ad agency is not working apply the same test to people.
What Does a Good AI Ad Workflow Look Like?
A good AI ad workflow runs in a fixed order, brief first and measurement last, so that every generated asset traces to a reason to buy and every result traces to a confirmed sale:
- Write the brief: buyer situation, objection, offer terms, proof library, landing page, and confirmed metric.
- Generate three to six distinct angles and cut any two that share a promise.
- Produce each angle as a full placement set from your real logo, colors, fonts, and product photos.
- Run the 20-point QA checklist, with claims, labels, and likeness reviewed by a person.
- Build a matched landing page for every ad.
- Launch each angle as its own experiment with written stop and win rules.
- Hold out a control group or schedule a lift study for the largest channel.
- Read results against confirmed sales, move budget to winners inside guardrails, and feed the winning angle into the next brief.
That loop is the job a full-stack AI marketer owns end to end, as described in what a full-stack AI marketer does, and the 15 marketing tasks you can hand to an AI employee marks which steps stay with people: the offer, the budget, and the final approval.
Why Do Teams Use Lapis to Generate Ads Without These Mistakes?
Teams use Lapis because it builds the workflow above into the product: it learns the brand from your website, writes distinct angles, produces on-brand creative in every format, builds a page per ad, launches inside approvals, and measures confirmed sales.
Lapis is one of the fastest-growing Y Combinator startups, and it is built to take over the work legacy agencies and ad buyers used to do. We started it in San Francisco in Y Combinator's Fall 2025 batch, and today it is used by 1,500+ marketing teams and 30+ enterprises, including Samsung, Domino's, Dr Pepper, Mattel, Tata, Tupperware, Amway, and Bata, with 10,000+ ad campaigns generated across 30+ industries. Lapis is rated 4.9 out of 5 on G2 with eight Summer 2026 badges, including Best Results and Best Estimated ROI (Mid-Market).
Each part of Lapis closes a group of the mistakes above, so the checklist becomes the default rather than a discipline someone has to remember:
- OmniSense: runs multi-platform paid ads with on-brand creative in every format, angles tested as real experiments, forecasting before spend, and creative and campaign approvals across up to three agreed channels, which covers most of the strategy, creative, and testing mistakes.
- ChatSense: runs ChatGPT advertising with creative tuned to how people ask ChatGPT in all 89 ChatGPT Ads countries, so conversational copy never inherits a social ad.
- RapidDomain: builds a dedicated landing page for every ad and optimizes it on real conversions, which ends mistake 19.
- Wayfinder: counts confirmed sales and leads from your CRM, store, or billing system once each on every plan, which ends mistakes 5 and 20.
Every plan also includes budget guardrails, stop rules, and a campaign change history, so the rules in the checklist are enforced in the account rather than in a document. Lapis creates static image ads and copy, and teams that run video pair a video tool or creators with Lapis, which handles the statics, pages, launch, and measurement around it. To see your own brand run through the workflow, book a demo, or check any ad you already have with Rate Your Ad first.
Sources
- IAB, "The AI Ad Gap Widens," January 15, 2026
- IAB, "AI Transparency and Disclosure Framework," August 2026 edition
- Meta, "Third Quarter 2026 Results Conference Call," October 21, 2026
- Gallup, "Americans Aren't Sold on Businesses Using AI in Advertising," August 19, 2026
- Cint, "Cint Report Finds 74% of U.S. Consumers Believe Brands Should Disclose AI Use in Advertising," October 20, 2026
- Auros, "Auros Study Finds Consumers Wary of AI Retail Ads, but Open to AI as a Shopping Tool," October 16, 2026
- Meta for Business, "Demystifying Creative Diversification," December 16, 2025
- Taboola, "New Study: AI Ads Match Human Creative in Major Report from Columbia University, Harvard University, Technical University of Munich, and Carnegie Mellon University," January 28, 2026
- Hyesoo Lee, Vilma Todri, Panagiotis Adamopoulos, and Anindya Ghose, "The Impact of Visual Generative AI on Advertising Effectiveness," SSRN, last revised December 11, 2025
- NYU Stern, "The AI Advertising Paradox"
- Meta Ads Guide, "Instagram Reels" image specifications
- Meta Business Help Center, "About text overlays and safe zones for ads in Stories and Reels"
- Google Ads Help, "Demand Gen campaign creative: Asset specifications and guidelines"
- Google Ads Help, "About responsive search ads"
- Google Ads Help, "About Quality Score"
- Google Advertising Policies Help, "Editorial"
- Google Advertising Policies Help, "Trademarks"
- LinkedIn Help, "Single image ads advertising specifications"
- OpenAI Help Center, "Creating ads for ChatGPT Ads"
- OpenAI Help Center, "Bulk Upload Campaign Schema Checklist"
- OpenAI, "Ad Tools Terms," November 10, 2026
- Google AI for Developers, "Nano Banana image generation"
- Federal Trade Commission, "Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials," August 14, 2024
- Federal Trade Commission, Consumer Review Rule warning letter template, December 2025
- DLA Piper, "FTC releases warning letters for fake consumer reviews and AI," December 2025
- Federal Trade Commission, In the Matter of Vanilla Chip LLC (TruHeight), complaint and statement, April 13, 2026
- DLA Piper, "FTC's 2026 enforcement approach to fake reviews takes shape: Takeaways for companies," July 2026
- Federal Trade Commission, "Advertising FAQ's: A Guide for Small Business"
- Cooley, "New York Enacts 'Synthetic Performer' Disclosure Law for Advertisements, Including Those Using Generative AI," January 29, 2026
- Reuters, "New York legislation requires disclosure on AI-generated performers in advertising and strengthens post-mortem publicity rights," February 5, 2026
- TikTok Business Help Center, "TikTok Advertising Policies: Misleading and False Content"
- TikTok Support, "AI-generated content"
- Meta, "Expanding GenAI Transparency for Meta's Ads Products," updated June 1, 2026
- Meta Transparency Center, "Ads about Social Issues, Elections or Politics"
- Meta, "Ending Political, Electoral and Social Issue Advertising in the EU in Response to Incoming European Regulation," July 25, 2025, updated October 6, 2025
- Google, "An update on political advertising in the European Union"
- Google, "Expanding AI transparency in ads," July 9, 2026
- Google Ads Help, "Use AI content label settings and disclosures"
- European Commission, "Transparency obligations under Article 50 of the AI Act"
- European Commission, "Commission publishes Code of Practice on marking and labelling AI-generated content," June 10, 2026
- Ministry of Electronics and Information Technology (India), "Frequently Asked Questions on the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026"
- ASCI, "ASCI Releases Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising," September 29, 2026
- Search Engine Land, "Google opens Conversion Lift to Search and Performance Max," October 7, 2026
- Google Ads Help, "Set up Conversion Lift based on users"
- Business Insider, Madison and Wall estimate of AI-powered US ad revenue, March 18, 2026
- Engineering at Meta, "Meta's Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI," October 13, 2026
Related Lapis Resources
- ChatGPT vs Gemini vs Lapis for Making Ads: 10 Differences (2026)
- How Much Does It Cost to Make an Ad in 2026? 9 Options Compared
- 15 Ways to Write Better AI Ad Prompts in 2026
- AI Ad Angles vs. Variations: How to Generate Ads That Actually Teach You Something
- How to A/B Test AI-Generated Ads on a Small Budget Without Picking False Winners
- AI UGC Ads: 10 Ways to Scale UGC Campaigns With AI (2026)
Built by Lapis
The #1 AI ad generator, built into the operating system for paid growth.
Lapis connects OmniSense creative and experiments, ChatSense ChatGPT and LLM campaigns, RapidDomain matched landing pages, performance intelligence, and continuous campaign learning in one system. Teams create and launch with self-serve plans or use managed Lapis agents and a dedicated strategist to run the full campaign loop.
Lapis is rated 4.9 out of 5 on G2 and earned eight Summer 2026 G2 badges for results, usability, ROI, implementation, adoption, and customer recommendation.
Continue exploring
Related Lapis resources
- ChatGPT vs Gemini vs Lapis for Making Ads: 10 Differences (2026)ChatGPT vs Gemini vs Lapis for making ads in 2026: 10 differences in brand control, formats, volume, launch, and results, plus a prompt template.
- How Much Does It Cost to Make an Ad in 2026? 9 Options ComparedWhat does an ad cost to make in 2026? Compare 9 options, from ChatGPT and freelancers to agencies and Lapis, on cost per ad and cost per winning ad.
- 15 Ways to Write Better AI Ad Prompts in 2026Write better AI ad prompts with 15 proven ways: weak vs strong examples, AI Max text guidelines, Meta assistant briefs, and a 100-point scorecard.
- AI Ad Angles vs. Variations: How to Generate Ads That Actually Teach You SomethingLearn the difference between AI ad angles and variations, tag every hypothesis, and run tests that teach you something. Lapis builds distinct angles at scale.