AI Video Ad Creative Goes Mainstream: Nearly Two-Thirds of Buyers Now Use GenAI

The pilot phase for AI ad creative just ended

A year ago, a media buyer who generated one AI cutdown for a Tuesday test had something to mention in the standup. On July 14, 2026, the Interactive Advertising Bureau published the full version of its 2026 Digital Video Ad Spend & Strategy Report and put a number on how fast that stopped being remarkable: nearly two-in-three video ad buyers now use GenAI for digital video creative, up from half in 2025.

The asset count moved with it. IAB reports that one-third of those buyers' ad assets will leverage GenAI this year, up from one-fourth in 2025, with that share projected to reach 43% by 2027.

Read the two together and the shape of the change is clear. AI creative is not an innovation line item any more. It is a third of the work, and an industry body just said so in writing.

This is an explainer for advertisers and brands. What the report actually says, which number belongs to which release (that part matters more than it sounds), and what the findings mean for how you plan creative between now and the end of the year.

What IAB actually published on July 14

The report comes in two halves, and they say different things

The 2026 Digital Video Ad Spend & Strategy Report was released in two parts, and confusing them is the easiest mistake to make with this data.

Part One landed on May 5, 2026, around the NewFronts, and carried the spend forecast. Part Two, the full report, landed on July 14 at the IAB Video Leadership Summit under the headline "Business Outcomes Are Just the Beginning." IAB says this second part, developed in partnership with Advertiser Perceptions and Guideline, examines the changes in how video ad spend decisions happen.

So the July release is a strategy document, not a spend document. Its subject is buyer behavior: what buyers trust, what they will pay a premium for, and how they are actually producing creative.

Finding	Release	Published
$80B US spend projection	Part One	May 5, 2026
11% year-over-year growth	Part One	May 5, 2026
Two-in-three buyers use GenAI	Full report	July 14, 2026
43% of assets by 2027	Full report	July 14, 2026
Humans in the loop (40%)	Full report	July 14, 2026

The adoption curve is the headline finding

The creative-production numbers are the part of the July release that is genuinely new, and they are the ones worth writing on a whiteboard:

• Nearly two-in-three buyers now use GenAI for digital video creative, up from half in 2025.
• One-third of their ad assets will leverage GenAI this year, up from one-fourth in 2025.
• That asset share is projected to reach 43% by 2027.

Note what each number measures, because they are easy to blur. The first is a headcount of buyers who touch the tools at all. The second is the share of finished assets those buyers actually ship. A doubling of the first without the second would be tourism. Both moved.

The spend headline belongs to May, not July

Here is the correction worth making out loud, because a lot of coverage will get it backwards.

The $80B figure is not July news. IAB published it on May 5, 2026, in Part One: "U.S. digital video ad spending is projected to surpass $80B in 2026," growing 11% year over year, which IAB describes as nearly 20% faster than the total ad market. The July 14 press release does not restate it. IAB's own July page refers back to Part One when it needs the volume context.

Three qualifiers travel with that number and get dropped constantly:

• It is US-only. IAB's own landing-page shorthand drops the "US," which is exactly how the scope error propagates.
• It is a projection, not a result. 2026 is not closed, so it is "projected to surpass," never "passed."
• "Nearly 20% faster" is a comparison of growth rates, not a gap in percentage points. Digital video growing 11% against a slower-growing total ad market is the claim.

For directional context on the mix, the July release states that social video (13%), CTV (11%), and online video (10%) are all projected to see increases. Those are growth rates, not dollar splits. If you need the segment dollars, open the report itself, because the labels get swapped in the coverage chain.

Why a third of ad assets flipped to AI in a single year

The bottleneck was never the idea

Ask any performance team what is limiting their creative testing and almost nobody says "we ran out of angles." They say the angles are sitting in a doc because producing them costs a shoot.

That is the constraint GenAI removed first. Not creativity, throughput. A hook you can only test if you book a creator is a hook you probably will not test, and the ideas that die in the doc are disproportionately the weird ones that would have won.

Cost per usable ad, not cost per video

The metric that actually changed is not the price of a video. It is the price of finding a winner.

The cost-per-winner equation: cost per winner = cost per variant multiplied by variants tested to find one. Most teams optimize the first term and never touch the second, which is backwards, because the second is the one with the exponent attached to your results.

Put illustrative brackets on it. If one in eight variants beats your control, a variant that costs $200 to produce makes a winner cost $1,600, while a variant that costs $3 makes the same winner cost $24. The hit rate did not change. The budget required to reach it collapsed. That is the whole mechanism behind the adoption curve IAB measured, stated without the vendor gloss.

A worked example

Take a skincare brand with one proven angle: "dermatologist-recommended." Under the old model, testing five new angles means five briefs, five creators, and one to two weeks of back and forth before a single impression is served.

Under the new one, the same five angles become five scripts and five renders in an afternoon. In Novoads, a five-second clip on the cheapest video model costs 1.5 credits, and the $49 Inicial plan carries 50 credits a month, so the entry plan covers roughly 33 clips at that setting. Heavier models cost more credits per clip. The point is not the exact count. It is that the unit you budget for stopped being the video and became the variant, which is the same conclusion IAB's asset-share number reaches from the demand side.

Buyers want the speed, and they want a hand on the wheel

Humans in the loop is the majority preference, not a hedge

The July report pairs the adoption curve with a caveat that any honest reading has to carry with it. Nearly everyone (96%) agrees there is a role for agentic AI, IAB found, and there is no consensus at all on what that role should be.

What buyers do agree on is where the brakes go. 40% want humans in the loop, and fully half (50%) of small and medium spenders feel strongly about it. That is not a rounding error in a survey. It is the largest single preference in the cluster, expressed by the people writing the checks.

Read next to a two-thirds adoption rate, it says something specific: the industry adopted AI creative and simultaneously refused to hand it the keys. Those are not contradictory positions. They are the same position.

The audit trail is the sleeper finding

The number nobody will quote is the one that will matter most in eighteen months. IAB found 36% of buyers want an AI agent audit trail for explainability, and 31% want guardrails that limit what agents can do.

An audit trail sounds like compliance vocabulary until somebody asks you a real question:

• Which version of the claim ran in which market?
• Who approved the before-and-after shot?
• Which model produced the asset a platform just flagged?

Teams that already keep an ad label and disclosure discipline can answer all three today. Teams that generate into a downloads folder cannot answer any of them.

You do not need a vendor to ship this. A naming convention, a shared folder, and a line in the brief recording who approved what will satisfy every question you are realistically going to be asked this year.

The other half of the report: buyers doubt what they are buying

Confidence in inventory quality is low across the board

Most of the July release sits somewhere that looks unrelated to creative and is not. Buyers were asked how confident they are in the quality of the inventory they are buying, and the answers get worse the further you move from a direct relationship:

Buying method	Somewhat to no confidence
Direct I/O, programmatic guaranteed, self-serve	43%
Private marketplaces (PMPs)	55%
Open exchange / RTB	67%

IAB puts the erosion down to two forces: bad actors introducing invalid inventory into the marketplace, and uncertainty around the origin and placement of otherwise legitimate inventory. One is fraud. The other is just fog, and fog is the harder of the two to price.

Why that pushes weight onto creative

Nearly all (93%) buyers agree that live content is worth more than all other video content, IAB found, pointing to higher attention (45%) and better business outcomes (38%). The report also notes that signal loss from IP degradation and the rise of AI-driven traffic mean targeting and audience reach now rank as equally important as business outcomes.

Line those findings up and a pattern appears. Buyers are paying premiums to buy certainty about where their money lands, because the supply side stopped handing that certainty over as standard.

Creative is the one input in that chain nobody else can degrade. You cannot audit every impression, but you can control every frame of the asset. When two-thirds of a market distrusts open-exchange supply, the lever that still responds predictably is the ad itself. That is a quieter and more durable explanation for the GenAI adoption curve than "the models got good."

The smaller-buyer squeeze

96% dissatisfied is an appetite number, not a rejection

The single sharpest statistic in the July release is easy to misread. Among smaller buyers, IAB reports, 96% are not satisfied with their current level of GenAI use for creative ad production.

That is not a satisfaction score for AI tools. It is a statement that smaller advertisers want to be using more of this than they currently are. Dissatisfaction pointed at your own usage level is demand, not churn.

What they say is missing

IAB is specific about the blockers. More than four in ten of those smaller buyers want more proof of performance and easier integrations with platforms and DSPs.

Both are honest asks and neither is about output quality:

• Proof of performance means the category has been sold on speed and cost, and under-served on evidence. If you cannot connect a generated asset to a result, you have bought a production tool and called it a growth tool. This is a creative analytics problem before it is a generation problem.
• Easier integrations means the last mile still costs a human hour: export, rename, resize, upload, tag. The generation step got a hundred times faster while the shipping step did not move.

The gap shows up per channel, not in the abstract. A single vertical platform like Snapchat can absorb a dozen finished videos a month once you run enough variants to separate them, which is precisely the volume that made the old creator-rate model unaffordable for smaller buyers.

If you are a smaller advertiser reading your own experience in those two lines, the productive response is not to wait for the category to mature. It is to fix the measurement side yourself, because that is the half you control.

What to change in your creative process this quarter

Budget for variants, not videos

Set the monthly plan in variants. Pick a number of distinct angles you will test, then work backwards to the tool and the spend, instead of buying a production package and rationing it. Teams running structured creative operations already think this way; most brands still budget per shoot.

A workable default for one month:

• Three angles minimum. One angle is a bet, two is a coin flip.
• Two hooks per angle, because the opening seconds carry most of the variance in a UGC-style ad.
• One untouched control, so every read has a baseline instead of a memory.

Keep the approval step, on purpose

IAB's 40% is a good default for you too. One named person reviews every asset before it goes live, and the review is logged. This costs minutes and it is the difference between "we use AI creative" and "we cannot say what our AI creative claimed."

Hold AI creative to the same measurement bar

The fastest way to waste the throughput is to run twenty variants and read them like twenty separate campaigns. Give each angle enough spend to say something, kill on evidence rather than vibes, and remember that a high CTR is not a result. Volume without a decision rule is just a bigger pile.

Do not let quality drift with volume

The one real risk in the curve IAB measured is that cheap variants become sloppy variants. The quality gap in AI ads is not usually the render, it is the brief: a generic script produced ten times is still a generic script. Volume multiplies whatever discipline you already had.

How Novoads fits the human-in-the-loop workflow IAB describes

Novoads generates UGC-style video ads from a script and an AI actor, or from an uploaded product image, and every render is a file you preview, approve, and download yourself. Nothing publishes automatically, which is the workflow shape IAB's 40% is asking for, arrived at by architecture rather than by policy.

What the loop actually looks like

Four steps, and a person sits at the end of all four:

• Write or auto-generate a script, then pick an AI actor to deliver it.
• Or upload a product image. Product-to-Ad takes an image file (JPEG, PNG, or WebP) rather than a URL, and produces an ad image at 0.3 credits.
• Render, and watch what came back before anything else happens.
• Decide. Keep it, rerun it, or throw it away.

Voices are available in 31 languages, so the same angle can be tested in more than one market without booking a second shoot.

Realism, honestly, is a tie. Every serious tool in this category rents the same frontier video engines, so nobody wins that row. The row that decides your quarter is what it costs to reach an ad you can actually run, and how many attempts that budget buys you.

What it costs

The $49 Inicial plan carries 50 credits a month. At 1.5 credits for a five-second clip on the cheapest video model, that is enough variants to run a real test cycle rather than one hero video. You can try it for $1 for 3 days first, and cancel whenever you want.

Creative volume is the media plan now

The interesting thing about IAB's numbers is not that AI won an argument. It is that the argument is over and almost nobody announced it. Half of buyers in 2025, two in three today, and a projected 43% of assets by 2027 is not a technology story any more. It is a planning assumption.

Which means the competitive question has quietly moved. When everyone can produce a hundred variants, producing them stops being an edge and choosing well among them becomes the whole job. The teams that win the next two years will not be the ones that generated the most. They will be the ones that kept a human on the approval step, kept a record of what ran, and could tell you why.

Creative is not the bottleneck any more. Judgment is.