An AI ad film is a commercial where generative video tools produce some or all of the imagery instead of a camera. In 2026 that stopped being a novelty and became a budget line: roughly 40% of video ads this year are expected to involve generative AI, and close to 90% of advertisers say they are using it or plan to. The question worth asking is no longer whether to use it. It is which half of the job you hand over.
We build both kinds of film — conventional shoots and generative ones — which puts us in an awkward but useful position. We can see exactly where the savings are real and exactly where they get quietly repaid later.
How much does AI actually save on video production?
A great deal, on paper. Industry benchmarks put generative video at roughly $0.50 to $30 per minute against $1,000 to $5,000 for freelance production and $15,000 or more for complex agency work. One widely cited figure has per-minute cost falling about 91%, and a 60-second cut that took thirteen days now taking under half an hour. Cost is now the single most-cited benefit, named by 64% of respondents — up from fifth place two years ago.
That number is true and slightly misleading. It measures generation, not production. It does not count the direction, the selection, the retries, the grade, the sound design or the brand-compliance pass — and those are where the hours go on anything a client actually signs off.
Does AI-generated advertising hurt brand trust?
It can, and the data is uncomfortably consistent. 78% of consumers say they trust video featuring real people more than AI-generated content, and among viewers who spotted AI in an ad, 36% said it lowered their trust in the brand behind it. Around 65% of US adults report being uncomfortable with generative AI being used to make ads at all.
Read that carefully, because it is not an argument against the tools. The penalty lands on detected AI. The failure is not that the model was used; it is that the seams showed. Audiences are not objecting to a technique — they are objecting to being able to tell.
The risk was never that AI looks bad. It is that AI looks generic, and generic is the one thing a brand paying for distinction cannot buy back.
Why hybrid pipelines outperform fully automated ones
The most useful figure in this year's research is not about cost at all: hybrid workflows — AI for production speed, humans for creative strategy — outperform fully automated approaches by more than 40% on brand equity measures. That gap is the whole argument. Automate the iteration; keep the judgment.
This is the split we work to, and it maps cleanly onto where each approach is strongest.
| Task | Best handled by | Why |
|---|---|---|
| Concepting | AI | A moving, sounding mood board before a day is booked |
| Pre-visualisation | AI | Camera moves and blocking tested in hours, no crew on standby |
| Paid-social variants | AI | Ten hooks generated and tested so media spend backs a winner |
| Impossible shots | AI | Imagined worlds and sequences that would break a live budget |
| Hero product frames | Camera | Exact pack, exact lockup, exact colour — models drift here |
| Human performance | Camera | Faces and hands are still where detection happens |
| Final grade and sound | Human | The pass that decides whether the seams show |
How to commission an AI ad film without the trust penalty
If you are briefing this work in 2026, the following sequence keeps the savings and avoids the damage.
- Decide what must be photographic — the product, the pack, the faces. Protect those before anything else is discussed.
- Use AI upstream, not downstream — concepting, previs and variants, where speed compounds and nothing ships unfinished.
- Do not marry a tool — the right model this quarter may be deprecated next. OpenAI wound down its consumer Sora apps barely a year after resetting everyone's expectations.
- Budget the finish, not the generation — grade, sound and compositing are what convert a generated clip into a brand asset.
- Test for detection — show the cut to people outside the project and ask what feels off. If they can name it, so can your audience.
- Keep a human name on the direction — the 40% brand-equity gap is a judgment gap, not a rendering gap.
What this means for smaller brands
Adoption is running fastest below the top tier: smaller brands expect about 45% of their digital video to be built with generative AI this year, against 36% at the largest advertisers. That is rational. A challenger brand with twelve products and no film budget now has a route to motion it simply did not have in 2023.
It also means the floor has risen for everyone, and a competent generated film is no longer a differentiator. When the tools are available to all, the advantage moves back to the thing they cannot supply: knowing which idea is worth rendering. We built twenty-four AI films for brands including Dettol, Finish, Al Ain and Bisconni on exactly that basis — the model produced the frames, the direction decided which frames were worth producing.
The tools got remarkable this year. The standard did not move.