Correction, 2026-10-02: we checked this guide again against first-party sources and 8frame's own prices, and several statements went further than the evidence. The 2024 agency figures (61% using, 30% exploring) came from a news report; Forrester's own release says "more than 60%" of US agency decision-makers were using generative AI and a further 31% were exploring it. We said ad platforms' automation needs "far less manual input than in 2024" and "optimizes best" when fed many creatives; neither is on the platforms' pages, so we now quote what Meta, Google and TikTok say about their own products. We stated as fact that 2026 video models clear "the bar for paid placement in most categories", that generative models produce "ad-grade creative", and that "every team getting real value" works one way; we had not measured any of that, so those lines now give IAB's survey figures, the output resolutions in the 8frame app, or advice. Our FAQ said Meta, TikTok and Google all require advertisers to disclose realistic AI content. Per their own pages, TikTok's ad policy asks for a label on completely AI-generated content, while Meta and Google Ads require advertiser self-disclosure only for some political, election or social-issue ads; the FAQ also left out the New York, California and EU disclosure rules. We removed an unsourced claim that hidden AI "costs more trust than a visible label", a claim that 8frame has "every leading model", a mention of "Topaz-class upscaling" that could read as video upscaling (8frame has no video upscaler), a promise that a first campaign flow takes "about ten minutes", and wording that made 448 credits look like the most a clip can cost. Legal points are now marked as general information, not legal advice, and match our disclosure and likeness guides. We also added that the Montage node joins the picture only: the merged file has no sound, so voiceover, music and the clips' audio go on the Video Editor timeline or in your own editor.
Correction, 2026-09-30: we re-checked every statistic on this page against its source. Removed: a "40% negative" figure the IAB report does not state (it reports a 12-point rise), a Salesforce "87%" figure we could only find second-hand, an agency hourly-rate range and a claim that AI creative performs on par or better, neither of which we could source. Reworded the IAB video-buyer figure to what the report says, and replaced the generation-cost ranges with 8frame prices.
AI in advertising stopped being a trend in 2026 and became the substrate. According to IAB and Sonata Insights' study "The AI Ad Gap Widens" (January 15, 2026), 83% of ad executives say their company has deployed AI in the creative process, up from 60% in the 2024 study. The question for a brand or agency team is no longer whether to use it. It is where in the pipeline it earns its keep, where it does not, and how to build a stack that turns adoption into an advantage rather than a liability. This guide maps the whole landscape: the four zones where AI operates, what changed between 2024 and 2026, how to build your own stack, and the limits you should design around. It is the hub for a cluster of deeper guides, linked throughout.
The landscape: four zones
AI touches advertising in four distinct places. It is easy to over-invest in the loud zone (creative generation) and under-invest in the quiet ones (measurement, personalization). All four matter.
1. Creative generation
This is the zone with the most visible change and the clearest economics. Generative image and video models make stills and clips at a cost per variant you can read off a price list (numbers below), and they differ by job: cinematic hero video, fast social variants, product stills, UGC-style clips. If you are choosing between them, start with the best AI video generator 2026 roundup and the Veo 3.1 vs Kling v3 comparison (Sora is no longer available on 8frame).
The workflows in this cluster span the full production chain. Pre-production comes first: the AI storyboard workflow for commercials, AI pre-visualization for agencies, and turning a board into a testable animatic. Production covers everything from a 30-second commercial made with AI to high-volume UGC ads and the AI ad variant testing workflow that makes creative volume affordable. Audio is in scope too: AI music for ads and AI voiceover for commercials.
The core economic shift is simple. On 8frame, an image costs 2 to 42 credits and most 5-second video clips 4 to 189, with a credit worth about $0.01 at pack rate. That does not make production costless, but it removes production cost as the reason you can only test two ideas. See what is creative volume for the case that volume matters more than any single model choice.
2. Media
The second zone is buying and placement. In part one of IAB's 2026 Digital Video Ad Spend & Strategy report (May 5, 2026), "Two in three buyers are live (21%), testing (20%), or planning to use (25%) agentic AI for digital video campaigns in 2026." The platforms' own automation covers much of the rest. Meta says its Advantage+ end-to-end campaigns work by "automatically applying AI across your campaign's audience, placement and budget". Google describes Performance Max as "a goal-based campaign type that allows you to access all Google Ads inventory" from a single campaign, with Smart Bidding setting the bids. TikTok says Smart+ automates "campaign and audience targeting, optimization, and creative".
The creative-side implication: these systems choose among the assets you give them. Google asks Performance Max advertisers to "Provide text, images, and videos", and TikTok says Smart+ "will also attempt to shift budget towards the highest-performing creative and pause poor performers". Give them genuinely different options to choose between, which loops back to zone one; how many is worth testing on your own account. This is also where dynamic creative optimization lives, assembling ads at delivery time from a library of generated elements.
3. Measurement
The third zone is quieter and less mature. AI is used for creative analytics (which hook, which frame, which variant is driving results), incrementality modeling, and pretesting before spend. The newest angle is synthetic audiences: LLM personas that simulate a target segment's reaction to creative. Treat their output as a directional signal, not a forecast, and read pretesting ads with synthetic audiences for what it can and cannot predict before you rely on it.
4. Personalization
The fourth zone is delivering the right message to the right segment at scale. AI makes it affordable to produce a version of a creative per audience, per market, per placement, rather than one master and a few crops. This is where creative generation and media automation meet: a codified brand system, expressed as reusable workflow templates, is how you keep those variants on brand across formats. See brand consistency across formats and performance branding with AI for the argument that AI collapses the old trade-off between brand craft and performance volume.
What changed from 2024 to 2026
Three shifts define the two-year gap, and they compound.
Adoption became the norm. In June 2024, Forrester's survey with the 4A's found that more than 60% of US agency decision-makers said their agency was already using generative AI, and a further 31% said it was exploring use cases. By January 2026, IAB and Sonata Insights put creative-process deployment at 83% of ad executives' companies, up from 60% in their 2024 study. Using AI no longer sets a team apart; how well it uses it does.
Generated video moved into paid production. In IAB's full 2026 video report (July 14, 2026), nearly two in three video buyers said they use generative AI for digital video creative, up from half in 2025, and IAB says one third of their ad assets will use it this year, up from a quarter in 2025, with 43% projected for 2027. Output specs rose as well: on 8frame, Veo 3.1 Standard and Fast and Kling 2.6 Pro return 1080p, and Hailuo H3 returns 2K. Whether a given clip is good enough for paid placement is still a call you make shot by shot, at full size, on the screen it will run on. For connected TV specifically, see AI for CTV ads.
Consumers noticed, and got skeptical. This is the shift that is easiest to underweight. IAB found 71% of Gen Z and Millennial consumers believe they have seen an ad made with AI, up from 54% in 2024, and the share who feel somewhat or very negative about such ads is 12 points higher than in 2024. The perception gap is the sharper number: 82% of executives think these consumers feel positive about AI ads, while 45% of consumers actually do. A Gartner survey of 1,539 US consumers, conducted in October 2025 and published on March 16, 2026, found 50% would prefer to give their business to brands that do not use generative AI in consumer-facing messages, advertising and content. So the tools got cheaper just as audiences grew more critical of what they make. See the full numbers in AI advertising statistics 2026.
The synthesis: the cost of an ad variant fell, output specs rose, and consumers became more skeptical. The practical lesson is to use AI to do better work at higher volume, not cheaper work at the same quality.
How to build your stack
A working AI advertising stack has layers. You do not need best-in-class at every layer on day one, but you need something at each. The full agency-scale build is covered in the ad agency AI stack 2026; here is the shape of it.
Generation canvas. The core layer: where image and video creative gets made. The 8frame approach is to put image, video and audio models from several providers on one canvas (among them Veo 3.1, Kling, Seedance, Wan, Hailuo, Nano Banana, Midjourney and ElevenLabs), so you route each job to the model that suits it without a separate subscription per provider, and save repeatable jobs as workflow templates. This is where the creative strategist works.
Copy and concept. LLMs for scriptwriting, headline variants, and concept development. This layer feeds the generation canvas with prompts and the media layer with copy variants.
Audio. Voiceover and music, covered in AI voiceover for commercials and AI music for ads. It is often the last layer teams build, although it is a small line in the budget: on 8frame, an ElevenLabs voiceover costs 13 credits per started 1,000 characters and an ElevenLabs Music track 81 credits per started minute.
Editing and finishing. Upscaling, cleanup, and assembly. Plan a finishing pass before anything ships: upscale stills, clean up artifacts, add brand marks as overlays rather than asking a model to draw them, and assemble the cut. On 8frame, upscaling is for images only (Topaz, Clarity and BRIA; there is no video upscaler), the Montage node joins 2 to 10 clips with transitions for 10 credits (picture only: the merged file has no sound), and the separate Video Editor mode has a timeline with video, text, caption and sound layers, which is where the ad's sound goes. This is where craft gets protected.
Asset management and governance. Version history, brand references, and a QA gate. This is the layer that makes volume safe rather than dangerous. See AI brand safety for advertisers for the review workflow that belongs here.
Measurement. Creative-level analytics and, optionally, synthetic pretesting before spend.
The build-vs-buy logic is consistent across layers: buy the layers where the market is mature and commoditized (generation models, platform media tools), build the connective tissue and governance that is specific to your brand. The one layer worth centralizing early is the generation canvas, because that is where fragmentation costs the most in switching time and inconsistency.
The team that runs it
Tools do not make advertising; people using tools do. The org shift is as real as the technology shift. One role worth defining early is the AI creative strategist, who combines taste, prompt craft, and performance literacy in one seat. Building the surrounding team is its own project, covered step by step in how to build an AI creative team. For the broader question of how AI reshapes the agency model and its pricing, see how agencies repriced after AI and AI vs traditional creative team.
Our recommendation: do not use AI only to shrink the team and cut costs. Use it to raise output and move the work toward strategy and direction, with a human on the judgment calls and the quality gate. The AI for marketing agencies playbook works this out in unit-economics detail.
The honest limits
A complete guide has to name the boundaries, because knowing them saves credits and reputations. The legal points below are general information, not legal advice: our guides on disclosing AI-generated ads and synthetic talent and likeness quote each rule from its source, and a lawyer should check your specific campaign.
AI does not replace the idea. Models generate execution, not strategy. A weak concept generated 40 ways is still a weak concept. The creative direction, the insight, and the brand judgment remain human work.
It does not fix substantiation. A generated ad can invent a claim as easily as a scene. Every factual claim still has to trace to something the brand can back up. In the US, the FTC's Endorsement Guides say an endorser "could be or appear to be an individual, group, or institution," and the FTC's rule on consumer reviews and testimonials, in effect since October 21, 2024, is aimed at testimonials that materially misrepresent "That the reviewer or testimonialist exists" or that they "used or otherwise had experience with the product." Never present an AI-generated person as a real customer. This is not a model problem; it is a governance problem you solve with a QA gate.
It does not clear rights on its own. A model that generates a face tells you nothing about whether you have the right to run it in an ad. New York requires written consent for using a living person's likeness for advertising (Civil Rights Law section 50), and California requires prior consent (Civil Code section 3344). The federal NO FAKES Act is waiting for a Senate floor vote and is not law. Trademark and other IP questions do not disappear because a machine made the pixels either. See synthetic talent and celebrity likeness in ads and AI and brand IP 2026.
It does not win over a skeptical consumer by hiding. Disclosure is now partly a legal duty. New York (since June 9, 2026) and California (from January 1, 2027) require ads to disclose AI-generated "synthetic performers", the EU AI Act requires disclosing deep fakes from August 2, 2026, and TikTok's ad policy asks for a label on completely AI-generated content. On the consumer side, 73% of the Gen Z and Millennial consumers in IAB's study said that knowing an ad was made with AI would increase or make no difference to their likelihood to purchase, and Gartner analyst Emily Weiss advises marketers to "label AI-driven experiences, so people understand when and how AI is being used." Full detail in do you have to disclose AI-generated ads.
Some categories still need a shoot. Products where texture, scale, or real-world physics are the whole story (some food, some tactile goods) may still need real production, with AI handling variants and versioning around a real hero. AI is a production function, not a magic wand.
FAQ
Is AI-generated advertising legal to run?
None of the rules we checked bans making ads with AI. They regulate disclosure and deception. In the US, no federal law requires a "made with AI" label on every ad, but New York (since June 9, 2026) and California (from January 1, 2027) require ads to disclose AI-generated synthetic performers, and in the EU the AI Act requires disclosing deep fakes from August 2, 2026. Platforms add their own rules: TikTok's ad policy asks for a label on completely AI-generated content, Meta requires advertisers to disclose AI use in some ads about social issues, elections or politics, and Google Ads requires it for election ads with synthetic or digitally altered content. The FTC Act still bans deceptive ads, and the FTC's reviews rule targets fake testimonials whatever tool made them. This is general information, not legal advice: see do you have to disclose AI-generated ads for each rule quoted from its source, and ask a lawyer about your campaign.
Does using AI hurt ad performance?
We could not find a neutral, public dataset comparing AI-made and hand-made creative on click-through or return on ad spend, so we will not claim either way. What is measured is perception: in IAB's study, 45% of Gen Z and Millennial consumers feel positive about AI ads while 82% of executives assume they do, and Gartner found half of US consumers would prefer brands that do not use generative AI in consumer-facing content. The practical answer is craft, testing and disclosure, not avoidance. Numbers are in AI advertising statistics 2026.
How much does AI advertising actually cost?
At the generation layer, little: on 8frame an image costs 2 to 42 credits ($0.02 to $0.42 at pack rate) and most 5-second video clips 4 to 189 ($0.04 to $1.89). An 8-second Veo 3.1 Standard clip with audio is 448 credits ($4.48), and longer or higher-resolution clips cost more. Plans start at $19 a month for 1,000 credits. The bigger cost in 2026 is not compute; it is the team, the governance, and the strategy. Budget for the people and the QA gate, not the pixels. The full cost model for an agency is in the ad agency AI stack 2026.
Where should a brand start with AI advertising?
Start with one high-volume, low-risk workflow where the economics are obvious: variant testing for an active paid campaign. Lock a hero concept, generate 10 variants, test them, and measure. That single loop teaches your team the workflow, proves the unit economics, and produces a result you can point to before you invest in the full stack. The AI ad variant testing workflow walks it through end to end.
Sources
- IAB and Sonata Insights, "The AI Ad Gap Widens", January 15, 2026 (104 US ad executives, 505 US Gen Z and Millennial consumers; 83% vs 60%, 71% vs 54%, negative share up 12 points, 82% vs 45%, 73% on purchase likelihood): iab.com, checked 2026-10-02
- IAB, 2026 Digital Video Ad Spend & Strategy Report: part one release of May 5, 2026 (agentic AI: 21% live, 20% testing, 25% planning), iab.com; full report release of July 14, 2026 (nearly two in three buyers use generative AI for video creative, up from half in 2025; one third of ad assets this year, 43% projected for 2027), iab.com; report page (full text needs a free IAB account), iab.com, all checked 2026-10-02
- Forrester, "Forrester: US Agencies Are Currently Leading Generative AI Adoption", June 17, 2024 (The State Of Generative AI Inside US Agencies, 2024, fielded with the 4A's: more than 60% using, a further 31% exploring): forrester.com, checked 2026-10-02
- Gartner, "Gartner Marketing Survey Finds 50% of Consumers Prefer Brands That Avoid Using GenAI in Consumer-Facing Content", March 16, 2026 (1,539 US consumers, October 2025; Emily Weiss quote): gartner.com; the live page refuses automated requests, so we read the Internet Archive copy of September 29, 2026, checked 2026-10-02
- Meta, "Meta Advantage+": facebook.com/business, checked 2026-10-02
- Google Ads Help, "About Performance Max campaigns": support.google.com, checked 2026-10-02
- TikTok Business Help Center, "About Smart+ Campaigns" (last updated June 2025): ads.tiktok.com, checked 2026-10-02
- FTC Act, Section 5: ftc.gov; FTC Endorsement Guides, 16 CFR Part 255: eCFR; Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465: eCFR, effective October 21, 2024 per the Federal Register, checked 2026-10-02
- New York General Business Law section 396-b (chapter 617 of 2025): NY Assembly, Governor's announcement of June 9, 2026; New York Civil Rights Law section 50: nysenate.gov, checked 2026-10-02
- California SB 1050 (chapter 246 of 2026): bill text; California Constitution, Article IV, Section 8 (start date January 1, 2027): leginfo; California Civil Code section 3344: leginfo, checked 2026-10-02
- NO FAKES Act of 2026, S. 4591, actions: Congress.gov, checked 2026-10-02
- EU AI Act, Regulation (EU) 2024/1689, Articles 3, 50 and 113, consolidated text of 27 July 2026: EUR-Lex, checked 2026-10-02
- TikTok advertising policy, "Misleading and false content" (last updated April 2026): TikTok Business Help Center; Meta, "How Meta Is Preparing for the 2026 US Midterm Elections": about.fb.com; Google Ads policies, "Political content": support.google.com, checked 2026-10-02
- 8frame prices, output resolutions and features: the 8frame app, checked 2026-10-02
This guide is the map. The build is a canvas where your models, brand references and workflow templates live in one place; the workflow library has campaign flows you can start from.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.