Product Page Video Is a Different Job Than the Ad You Just Shipped

Buyers Switched Back to Acquisition. Here Is Where It Settles.

Product page video is a separate deliverable from the ad you just shipped, and in 2026 it is the surface where new-customer budgets actually settle. The IAB's September update put customer acquisition back at the top of the buyer's list, named as the leading media investment goal by 63% of decision-makers, while focus on using generative AI inside campaigns fell to 69%. The cut that earns the click and the cut that closes the sale are not the same piece of work.

The IAB's 2026 Outlook Study: September Update, released on 10 September 2026 and drawn from more than 200 brand and agency ad investment decision-makers, raised its full-year US ad spend growth forecast to 12.3%, up 2.8 percentage points from the 9.5% projected in January. Customer acquisition is the top media investment goal for 63% of buyers, nine points higher than in January, while brand equity rose six points to 43% and repeat purchase held essentially flat at 24%. The top media investment challenge, named by 44% of buyers, is adapting to evolving consumer behavior including AI-driven search, with AI slop following at 38%. Optimizing content for AI-generated answers is the number one area of increased focus at 76%, while attention to the models themselves sits at 72%.

The buyer's attention moved off making things with AI and onto being found by it. Social media is expected to grow 16.5% this year, CTV 15.6% and commerce media 13.6%, while linear television declines 1.5%. Eighty-six percent of buyers are changing, or expect to change within twelve months, how they measure media because of conversational AI, and 45% name comparing an AI-driven journey with a traditional one as their hardest measurement problem.

For a video team the practical consequence is a question about surfaces. Acquisition money does not convert in the feed; it converts somewhere the brand controls, and for most e-commerce businesses that surface is the product page. The feed ad earns the click. The product page video has to close it, on the same page where the shopper is comparing the product against three competitors in another tab.

The Ad Cut and the Product Page Video Are Different Jobs

A campaign cut is built for interruption. It has to earn the first second from someone who was not shopping, put the brand on screen early, and land one claim. A product page cut is built for a shopper who has already clicked and has arrived with a specific worry: will it fit, how big is it really, does it do the thing the photos imply. The finish and the length that win one of those jobs lose the other.

Retail media now ships {{link}}, and that is exactly why the on-page job gets neglected: one of the two deliverables arrives with a media plan attached and the other does not. Commerce media was among the fastest-growing lines in the IAB's channel table precisely because it sits next to the checkout, but the asset it generates is still an advertising asset.

The social channel solved its own version of this by treating {{link}} as a standing asset class rather than a campaign output. Product pages have not had that treatment. Most e-commerce catalogues run one template across thousands of SKUs, and the video slot in that template is usually filled by whichever clip the last campaign happened to produce, cropped to fit.

Retail media now ships the retail-media pipeline that turns a product page into a finished ad, and that is exactly why the on-page job gets neglected: one of the two deliverables arrives with a media plan attached and the other does not.

The social channel solved its own version of this by treating the social-commerce content engine as a standing asset class rather than a campaign output.

Half of Major Retailers Still Ship a Mediocre Product Page

Baymard Institute's Product Page UX benchmark covers 343 leading US and European e-commerce sites against more than 110 product page guidelines, and its latest update is not flattering. Only 49% of benchmarked sites reach a decent or good product page UX performance, which leaves 51% at mediocre or worse, and no benchmarked site reaches perfect or state of the art. That is the context in which a brand decides whether its product video underperformed or its page did.

The methodology matters because the headline is easy to dismiss. The findings come from 25 rounds of moderated Think Aloud usability testing covering more than 4,400 participant and site sessions, plus in-lab eye-tracking, and the wider catalogue has run 54 benchmarking rounds across 650-plus guidelines. Participants hit more than 1,300 product page usability issues on sites run by multi-million-dollar businesses, and repeatedly abandoned otherwise suitable products over resolvable page problems.

The operational lesson is that placement runs before content. A correct demonstration sitting in a gallery nobody opens, loading below the fold, or hiding behind a UI pattern the shopper never triggers has the same commercial effect as no video at all, and it still costs the render and the review round. Before commissioning another cut, confirm which slot the clip occupies and what actually triggers playback.

On a Product Page, the Frame Is a Claim

The European Commission's own summary of the AI Act is blunt about generative output. Providers of generative AI have to ensure that AI-generated content is identifiable, and certain AI-generated content should be clearly and visibly labelled, namely deep fakes and text published with the purpose to inform the public on matters of public interest. The transparency rules come into effect in August 2026. For commerce video that is a labelling regime delivered through the asset rather than through the campaign brief.

The machine-readable half of that regime is Content Credentials, the open technical standard maintained by the Coalition for Content Provenance and Authenticity for establishing the origin and edits of digital content. C2PA describes the mechanism as functioning like a nutrition label for digital content, giving anyone access to a record of what produced and changed the file. The limit is built into the design: provenance records history, it does not adjudicate accuracy. A manifest can prove a clip was generated and edited. It cannot prove the generated product looks like the product.

That gap belongs to the production team, and it is narrower on a product page than anywhere else in the media plan. On a feed ad a frame is an impression. On a product page the same frame is a claim about colour, scale, texture and performance, and the shopper is holding the physical item a few days later. No metadata field discharges that claim.

Crop It Later Is Not a Delivery Spec

A product page video is delivered into a gallery slot with its own constraints. It competes with stills for the first glance, it usually autoplays muted, and its first frame doubles as a thumbnail. Platforms have started doing the crops themselves, which is {{link}} to any video that enters a paid placement, and the same automation applied to a product gallery produces the wrong first frame, landing on a hand, a background or a logo instead of the product.

Deliver per-placement masters instead of one master plus crops. Keep the product on screen early, keep the demonstration legible with the sound off, and keep generated lettering out of the frame entirely so nothing needs re-cutting when a legal line changes. If colour, scale or texture is what sells the product, that attribute has to be legible in the first frame, because for a large share of shoppers the thumbnail is the whole video.

Platforms have started doing the crops themselves, which is what platform-side reframing already does automatically to any video that enters a paid placement, and the same automation applied to a product gallery produces the wrong first frame, landing on a hand, a background or a logo instead of the product.

Measure the Page, Not the Viewer

Most product video reporting measures the shopper rather than the video. Viewer-level engagement asks whether the people who pressed play bought more, and the people who press play were already further down the funnel; that self-selection gap is wide enough to make almost any clip look effective. The number that returns a verdict is the page's own conversion over the same weeks with the video and without it. The honest version is {{link}} applied to one page template rather than to a whole campaign.

That test is cheap on a product page because the catalogue supplies the control group. Split a product group, put video on half the SKUs, leave the other half unchanged, and compare page conversion, add-to-cart and return rate over the same period. Return rate is the part campaign reporting never sees and the part a generated frame influences most: a clip that flatters colour or scale lifts add-to-cart and then comes back through the returns desk, months after the media report has closed.

A short pre-render checklist follows. Confirm the gallery slot and the playback trigger. Write down the single claim the cut is making. Check the frame against the physical product rather than against the previous campaign. Confirm the provenance record and any label the platform will attach on its own. Then hold out half the catalogue and let the page, not the player, decide whether the clip earned its place.

The honest version is the holdout discipline behind incrementality testing applied to one page template rather than to a whole campaign.