← Back to blog
AiAbout 6 min read

Static Product Photos Become Vertical Video Now. The Liability Moves on November 20.

Published Nov 8, 2026
Static Product Photos Become Vertical Video Now. The Liability Moves on November 20.

On November 2, Meta added Image Animation to Advantage+ Creative. It takes a single product photograph and applies diffusion-based motion: a slow pan, a slight zoom, subject movement, particles around the product. The output is a 9:16 video for Reels and Stories.

For a brand with a good photo library and no video team, that is a real unlock. A catalog of product shots becomes a catalog of vertical video without a shoot.

The same update is bundled with a change to Meta's Commercial Terms, and that part has a date attached.

What ships on by default

Image Animation joins two other generative video capabilities already in Advantage+ Creative. Video Expansion resizes existing assets to fit placements it was not shot for. Catalog-to-Video generates product videos from feed listings, which is the version that scales to thousands of SKUs.

All three are enabled by default in new Advantage+ Shopping and Sales campaigns. For existing ads, Image Animation is opt-in. The distinction matters because "default on" is how most advertisers will meet these features, without ever opening the enhancement panel.

There is a fourth piece worth knowing: Brand Memory, which trains the system on a brand's visual guidelines. It is the mechanism that keeps generated output inside the brand's look, and it is also the reason a brand should care about what it has been fed.

The clause that changed

Meta's revised Commercial and Self-Serve Ad Terms take effect November 20, 2026. They were due October 30. On October 23 a Meta spokesperson told Reuters the date was moving so the company could explain what is changing and what is not, after publisher backlash over a separate clause about crawling business websites. The UK's Competition and Markets Authority has described the rollout as on hold.

The AI Features section is the part to read:

> Meta may use or make available products and features powered by artificial intelligence... You acknowledge and agree that AI features may make mistakes and may not always produce accurate, complete, or suitable results. You are solely responsible for and assume any and all risks associated with such AI features and any resulting Content, including evaluating the suitability of any Content for business or commercial purposes. You must not alter or remove any watermark or content-authenticating metadata generated by AI in connection with Content.

The old terms tied liability to a business choosing to use Meta's AI tools. The new wording does not contain that condition. As reported by Adweek and summarised by several agencies, the "you chose it" framing is gone.

Asked directly whether an advertiser is liable for a feature they never switched on, Meta's spokesperson declined to say.

Why the date matters more than the feature

November 20 lands two weeks before Black Friday. That timing is the whole story. The group with the most live creative in flight, and the least appetite for an audit, gets the new terms exactly when changing anything is most expensive.

The mechanics compound it. Advantage+ enhancements are toggled per ad, inside a collapsed creative menu, not as one switch at account level. An account running forty ads needs forty checks. Most teams have never opened that panel because the features arrived switched on.

There is a second, quieter problem. Nobody has published a straight answer on whether turning a feature off today limits exposure for edits Meta already made while it was on. The clause is written about content in your campaigns, not about your settings at a point in time.

What an audit actually looks like

The practical answer is unglamorous. In Ads Manager, open each active ad, scroll to Advantage+ Creative enhancements, and record which generative features are on. Then look at the creative that is actually running, not the creative that was uploaded.

The reason is documented. Advertisers including Bookshop.org have described AI-added illustrations and edits appearing on ads they did not ask for. Under the new terms, that content is theirs either way.

Three categories deserve a harder look than the rest. Regulated verticals, where an AI-rewritten claim can become a compliance event. Product photography where accuracy is the commercial promise, which is where motion synthesis is most likely to misrepresent. And any account running automated translations, because a translated offer is a new offer and nobody approved it.

A fourth is easy to miss: the watermark clause. Removing AI provenance metadata is now explicitly prohibited, which makes any internal process that strips metadata a terms violation rather than a housekeeping preference.

The agency version of the same problem

Agencies have an extra layer, because the Commercial Terms require them to promise that any third party they act for will follow the terms, and to have authority to bind that third party. A separate line extends sole responsibility to the acts and omissions of any third-party software or service authorised on an account.

Read together, that covers the MCP servers and creative agents agencies have been wiring into client accounts all year. If an agent writes a headline or swaps an image and it runs, the exposure is described the same way as Meta's own AI making the change.

Meta's liability for all of this stays capped at the greater of $100 or what the advertiser paid Meta in the previous twelve months. That number is worth holding next to the size of a Black Friday budget.

The performance case, and its shape

None of this argues the features are bad. One agency reported 20% higher conversion across seven UK brand accounts using the AI enhancements, which is a data point published without a control group. When the automation picks the right crop or the right translated line, the work is genuinely done for you.

The honest framing is that the automation is a bet on the average case and the terms are a transfer of the tail. Most generated variants will be fine. The ones that are not will be visible to customers under your brand name, and after November 20 the paperwork says that was your decision even when it was not.

There is a cheaper reading of the timing as well. Moving from October 30 to November 20 pushed the effective date past the point where most brands had already locked their holiday creative, which means the first wave of ads built under the new terms will be the ones made in a hurry. A deadline set by legal review and a deadline set by campaign calendars are rarely the same date, and this one lands where the two collide.

What to do before the twentieth

Two things, in order.

Decide which ads can carry generated creative at all, and turn the rest off. The output is usually fine. The review cost of approving every variant is not, and the deadline is the wrong day to discover that.

Then write down who approves a generated variant before it scales. Under the new terms the approval is the control. There is no technical mechanism that makes an unreviewed AI edit safe, and Meta's own answer to "how do I turn it off" is the only control it has offered.

The feature is a convenience. The clause is a transfer of risk with a stated date, and the date is the twentieth.

Related articles