WHY BRANDS NEED TO RETHINK THEIR VIDEO CONTENT STRATEGY FOR AI SEARCH

From nine dollars to one billion hours per day

The history of video marketing begins long before social media. On 1 July 1941, a ten-second advertisement for Bulova watches was broadcast before a baseball game in New York: a clock placed on a map of the United States, accompanied by a phrase destined to become famous. Bulova spent nine dollars - about 240 current dollars, or 180 euros - for what is considered the first official television commercial.[1]

For the next six decades, the rule was simple. Brands bought a moment in the middle of a programme watched by a mass audience, while a handful of broadcasters decided what was worth seeing.

The internet broke this model in stages. In April 2005, a nineteen-second video titled "Me at the zoo" became the first video uploaded to YouTube, a platform that then presented itself as a "video directory for the Internet".[2] Two years later, the iPhone put a video screen and a search engine in every pocket.[3]

Twenty years later, YouTube alone boasts billions of monthly users, who watch more than one billion hours of video every day.[4] By the end of 2025, video represented about 75% of all global mobile data traffic, and Ericsson predicts that total traffic will multiply by another 2.2 times by 2031.[5] Marketers have followed the audience: in 2026, 91% of companies surveyed by Wyzowl used video and 82% of video professionals believed it generated a good return on investment,[6] while 58% of B2B marketers surveyed by the Content Marketing Institute ranked video as the format producing the best results, ahead of case studies, white papers, and short articles.[7]

The debate on consumption is therefore settled. The open question in 2026 lies elsewhere: can a machine find, cite, and recommend what a brand publishes?

Search is no longer a list of links

Google, ChatGPT, Perplexity, and other generative interfaces interpret intent, explore multiple paths, compare sources, and construct an answer. The phenomenon has already entered mass usage: in June 2026, Google indicated that AI Overviews had more than 2.5 billion monthly active users, that AI Mode had crossed one billion, and that query volume in AI Mode had more than doubled each quarter since its launch.[8]

Search is also becoming visual. Search Live allows users in more than two hundred countries and territories to converse with Google while showing it what they see through their phone's camera.[9] A query can start from a product on a shelf, a mechanical part, a plant, an outfit, or an entire storefront, with the question being asked out loud.

The interface is changing behaviours. Instead of typing "men's running shoes", internet users ask: "Which shoes should I choose to resume running after a knee injury, mainly on the road, three times a week?" The question sounds like a conversation with a salesperson who knows their job - and it expects the same quality of answer.

The click is fading away along with the keyword. A study by the Pew Research Center of nearly 69,000 Google searches shows that users clicked on a traditional result in 8% of visits when an AI summary appeared, compared to 15% when there was none. They only opened the links cited in the summary in 1% of visits.[10] Bain estimates that about 80% of consumers already rely on zero-click results or AI summaries for at least 40% of their searches.[11]

Being third on a results page matters much less when almost no one clicks. The decisive indicator is now: is the brand's expertise cited in the answer itself?

From now on, it is the AI that watches

To answer a complex question, a generative system needs demonstrations, comparisons, actions, testimonials, and evidence. Video gathers precisely all of that.

YouTube makes this shift very concrete. With Ask YouTube, a user can formulate a complex request - for example, getting advice on teaching a child to ride a bike - and then ask follow-up questions. In practice, it is the AI that watches: it scans long videos and Shorts to find the precise passage that answers the question, then builds a structured response around that moment.[12]

Compare this to how most brand videos are still produced. A flagship film, adapted into a few social formats, broadcast for three weeks, and then replaced by the next campaign. The video may generate views and engagement, but it answers no lasting question. Its title remains vague. Its description is two lines long. It has no transcript, no chapters, and no dedicated page. Its entire value depends on someone pressing "play".

For an AI assembling an answer, this video is invisible. Nothing in it is readable by the machine; therefore, it never passes pre-selection. The videos that are retained are answer-content: resources built around a question that a person - or the algorithm accompanying them - is actually asking.

The brands that took the lead

The best examples come from brands that made usefulness a media strategy long before people spoke of "AI optimization".

Decathlon publishes guides and videos that answer practical questions: how to start running, how to choose your shoes, how to select your ski boots.[13] Leroy Merlin structures its tutorials around precise actions, like installing a bathtub, fitting an RJ45 socket, or levelling a floor, and associates each video demonstration with a written explanation.[14] HubSpot Academy transforms its products and methods into courses, demonstrations, and lessons hosted on structured pages.[15]

These resources were created for humans. They also possess, almost by nature, all the qualities that new search engines need: an identifiable question, a precise answer, a visible expert or action, text surrounding the video, and a usefulness that outlasts a campaign.

The numbers confirm the strength of this straightforward approach. Wistia's analysis of three-to-five-minute videos reveals an average engagement of 74% for educational content, compared to 43% for all videos of the same length.[16] For a "how-to" query, a clear demonstration is worth more than epic music and a drone shot at sunset.

The best answers are in the minds of the teams

In most companies, the most useful answers live in the minds of employees, not in campaign scripts. Think of the engineer capable of explaining why one material lasts longer than another, or the technician who diagnoses a breakdown in three seconds, just by the sound. Every salesperson hears the same three objections at every meeting and has answers that never leave the room.

For years, this knowledge remained scattered among conversations, sales presentations, tenders, and internal documents. Video can make it visible. AI search further increases the value of this approach, because everything that is published can now be understood, summarised, and recommended. Everything that remains inside the company remains invisible to the machines that build the answers.

An SME will never publish as much as a content factory. However, it can identify the fifteen questions to which it responds better than anyone else, then film those answers and back them up with evidence. A manufacturer can show the concrete consequences of a technical choice. A software publisher can dedicate each video to a precise use case or problem. A consultancy firm can explain the decisions that its clients too often make in the wrong order. A family business can put the people who have carried its expertise for twenty years in front of the camera.

Today, when anyone can generate a decent script in thirty seconds, generic competence is worth nothing. An identified expert, a real figure, a hand executing an action repeated ten thousand times: no prompt can make that exist.

Give every video a text layer

One obstacle remains. An AI does not watch a video in the exact same way as a human; it relies on the information surrounding the file, on the text that identifies the content, contextualises it, and makes its meaning extractable.

Google's recommendations are concrete. Important information must remain available in text form. Videos benefit from being hosted on indexable pages, with unique titles and descriptions. VideoObject structured data, chapters, timestamps, and Clip or SeekToAction markup help systems locate key moments.[17]

All of this boils down to one instruction: give every video a text layer that tells the machine what it shows and which question it answers.

Treated in this way, a single interview with an expert can produce a main video, an edited transcript, a page answering a specific question, several short clips, an article, sales collateral, and a training tool. Each element gives context to the others. The video carries the proof; the surrounding text explains to machines and hurried readers what that proof means.

At Trinity, we have seen strategic video content contribute directly to tenders representing more than 500 million euros in contracts won.[18] In those rooms, the film's role was simple: to allow the evaluation committee to see the expertise, rather than read assertions about it. A structured library of expert answers had more impact than any showcase film.

Explain better, or be summarised instead

All companies now have access to the same generation tools. Result: all channels are filling up with the same sentences and the same visuals, barely distinguishable from an image bank. Volume no longer costs much. Deserving to be cited, however, does.

The brands that will stand out will be those that place real experts in front of the camera to answer real questions, and whose videos will continue to be cited five years after their publication. For eighty years, brands have mainly used video to buy attention. The decade ahead will reward those that use it to transmit knowledge.

The next time a customer points their phone at a shelf and asks which product to choose, a brand's expert will be part of the answer. The work described in this article determines which one.