Are you getting the GEO value from all that video?

The GEO opportunity in video may begin before the camera turns on: decide what the organisation still needs to learn, then use the interview to acquire the evidence.

A lot of B2B brands are putting serious money into video — customer films, executive interviews, event footage, social clips, agency shoots with crews and producers and locations and several rounds of edits.

The obvious question afterwards is whether the video performed. I think there are two better questions.

What unique evidence did the shoot produce?

And, increasingly:

Did we know what evidence we wanted before we asked the questions?

Let me be careful about what I’m not claiming. Publishing a video does not magically improve how a brand appears in AI search, and I’d be suspicious of anybody selling that. The relationship is considerably less direct than the current enthusiasm suggests.

The interesting thing happens earlier, during the interview design and the shoot itself.

A good customer or practitioner interview can produce something that is becoming genuinely scarce: first-hand information that exists nowhere else on the web. A customer explaining precisely how a regulation changed their workflow. An accountant describing a problem their clients never see. A founder giving numbers, language or an operational detail that no generic article could reproduce.

Then a three-minute film gets published, and the remaining fifty minutes go onto a drive.

That’s a strange use of expensive access at a moment when every organisation is worrying about differentiation.

But even that framing is incomplete. The bigger opportunity is not simply to rescue unused footage afterwards. It is to use GEO and audience-question intelligence before the shoot to decide what the organisation needs to find out.

Some of the budget is buying internal visibility

It’s worth being honest about why this persists, because it isn’t stupidity.

A shoot produces something that looks unambiguously valuable from inside the organisation. There is a crew, a location, proper lighting. Senior people appear in it. The edit comes back looking superb, and it is obvious to everybody who sees it that serious work has been done. It goes up on LinkedIn, and a good share of the organic engagement is people who work for the company saying how good it looks. Then paid media goes behind it to reach the audience it was actually made for.

There is nothing wrong with that in itself — paid distribution is distribution, and buying reach is a legitimate way to get something in front of people. The problem is that three quite different things are being counted as one.

Internal visibility is the effect on colleagues and leadership: proof that the team is producing significant work. Paid reach is bought, and would have been available for almost any asset. External organic value is whether anybody outside the organisation, who was not paid for, sought this out or found it useful enough to pass on.

Only the third is evidence that the thing was worth making, and it is consistently the least reported of the three.

I should put myself inside this rather than describe it from outside. I’ve worked on content whose clearest measurable outcome was that people internally thought it was good, and I was pleased about it at the time. That is a genuinely nice thing to happen and it isn’t nothing. It is also not the same as somebody outside the organisation deciding the work was worth their attention, and the two feel far more similar than they should.

I’d rather not turn this into a cheap point, either, because internal visibility is not illegitimate. A film that finally makes a sales team understand a customer’s situation, or that gets a leadership group to back a programme, has done real work — and internal persuasion is often the harder problem. The failure is narrower than that: it’s when internal enthusiasm gets read as evidence that the audience wanted it, which is the one thing it cannot tell you.

What it sounds like at the other end

The reason this matters beyond the accounting is that it has an output, and the intended audience can recognise it immediately.

Technology companies talking to finance leaders tend to produce material that sounds like every other technology company talking to finance leaders. It names the pressures a CFO is under without telling them anything they didn’t already know, describes a category rather than a decision, and could be re-badged for a competitor with a find-and-replace. I’ve spent a long stretch working on research for that audience, and the gap between what finance leaders actually argue about and what gets said to them in marketing is not small.

Video tends to make this worse rather than better, and the mechanism is structural rather than creative. An expensive, highly visible asset attracts more approvals than a blog post, and every specific claim is something an approver can reasonably object to. The safe version survives each pass. What comes out is fluent, beautifully shot, and says nothing anybody could disagree with — which is generally the same as saying nothing anybody needed.

Approvability and usefulness are different properties. A process weighted towards internal comfort will select hard for the first, and the finished film is where that selection is most visible, because it is the format with the most people standing behind it.

Which is the actual case for doing the evidence-gathering, rather than a nice thing to do alongside it. A shoot that captures something specific — a number from the customer’s own operation, a case where the standard advice failed, the language they use rather than the language the category uses — produces material that an approval process cannot easily flatten, because there is no generic version to retreat to. You can water down a claim. It is much harder to water down somebody’s account of what actually happened to them. Specificity isn’t a stylistic preference in this argument. It’s the only durable defence against a process that would otherwise arrive at the safe version by default.

The question before the camera turns on

If a business already knows which questions it wants to be authoritative on, those questions can go into pre-production rather than being discovered afterwards.

The sequence isn’t complicated:

audience / AI-search question → current answer → evidence check → missing knowledge → interview prompt → first-party evidence → stronger written answer

That middle step matters.

A weak GEO answer is not always a content problem. Sometimes the page is badly structured, incomplete or difficult for machines to retrieve. That can be fixed editorially or technically.

But sometimes the organisation simply does not know enough yet.

Suppose research shows that a group of professionals frequently does work outside the original scope of an engagement but often does not charge for it. That establishes the behaviour. It does not necessarily explain the mechanism.

A writer could infer plausible reasons: fear of damaging the relationship, fragmented requests that feel too small to bill, unclear boundaries, fixed-fee models, a culture of helping clients.

Those may all be reasonable hypotheses. They are still hypotheses.

If audience or GEO analysis shows that people are asking why this happens, the stronger move may be to use the next practitioner interview to find out.

Instead of asking:

Do you struggle with scope creep?

ask:

Tell me about the last time a client asked you to do something you knew was outside scope, but you didn’t charge them. What stopped you?

Now the interview is doing more than supplying a quote for a film. It is acquiring evidence the organisation did not previously have.

The film can still be made exactly as planned. What changes is that somebody has cross-checked the interview brief against the questions the written content will eventually need to answer.

That is the part of the workflow I think is easy to miss.

GEO should influence some interview questions, not all of them

There is an equally obvious way to ruin this idea: turn every video interview into a keyword interrogation. If every question comes from a prompt tool, the result will be rigid, predictable and probably worse television. The useful model is selective.

Start with the existing editorial or production concept. Understand what the interviewer already intends to ask. Then compare that question set with the important audience questions and known evidence gaps.

Most of the time, leave the interview alone. Where an important question is already covered, do nothing. Where it is almost covered, sharpen a follow-up.

Where there is a genuinely important knowledge gap that this particular person is unusually well placed to answer, add a question. That is very different from handing a producer a list of twenty GEO prompts. The editorial role is to decide which gaps deserve the scarce access.

The written article changes too

This matters particularly when video and written content are being produced around the same subject.

The weak workflow is:

film interview → edit film → writer watches film → article summarises film

The stronger workflow is:

written audience question → existing evidence → missing explanation → interview designed to capture it → practitioner answer → article built around stronger first-party evidence

That means the accompanying article does not have to be a derivative asset.

It can become the most complete, crawlable expression of what was learned.

The video supplies voice, personality and reach. The article can supply the concise answer, underlying data, attributable practitioner explanation, context, supporting evidence and links to related material.

The two formats do different jobs while drawing from the same reporting effort.

This is a more useful definition of “reuse” than simply turning one asset into six smaller assets. The thing being reused is knowledge, not just footage.

Captured evidence still needs a route

Since writing this, I have been testing the post-shoot side against live content plans. In one case, I took a short expert interview recorded at an industry event and compared the transcript with several briefs that were already in production. Three answers looked reusable, but the useful question was not “how many clips can we make?” It was “which live question does each answer actually help?”

One answer described the setup, education, chasing, review and preparation that happen behind a regulatory submission. That clearly strengthened guidance about quarterly reporting. Another described slow responses, missing information and the responsibilities on each side of the workflow, so it had a plausible home in both reporting and record-keeping guidance. A third answer was strong, but it was really about workload and capacity inside professional practices. Forcing it into user-facing support would have weakened both the clip and the page. The speaker also mentioned year-end work, but did not actually explain the year-end process, so I rejected that route too.

That exclusion is the part I want to preserve. Reuse should not mean maximum distribution. It should mean appropriate routing. A strong answer can be routed into a live brief, retained for a better future use, or rejected for a particular destination because it does not answer the question well enough.

That gives post-production a clearer loop:

brief requirement → source evidence → selected evidence object → route / retain / reject → owned answer, social use or future brief

An evidence library without routing logic is still mostly an archive. The editorial work is connecting what was actually said to the question that needs answering, while resisting the temptation to use everything simply because it exists. That is also why the brief matters so much: as I argue in The brief is where real-world insight should enter the system, a live brief gives the evidence somewhere meaningful to land.

Video is probably the capture layer, not the whole answer

This distinction is worth being precise about, because it’s where the claims tend to get loose.

Google has a well-established video-search layer, and its guidance is explicit that video can appear across the main search results, video mode, Google Images and Discover. It recommends making videos discoverable and indexable, using dedicated watch pages where appropriate, and supplying useful metadata. VideoObject structured data helps Google understand the video, and key moments can be marked through structured data or YouTube timestamps. Google’s video SEO guidance and its VideoObject documentation are both clear that the page around the video matters as much as the file.

That establishes a real video-search opportunity. It does not establish that a finished video creates generative-search visibility, and the two keep getting conflated.

For AI search specifically I’d be more cautious. OpenAI’s publisher guidance says public web content needs to be accessible to OAI-SearchBot for it to be available for inclusion in ChatGPT search summaries and snippets — which makes the crawlable layer around valuable source material important before we even reach the harder question of how different systems interpret video itself.

So my working hypothesis isn’t that video improves GEO. It’s that video can create proprietary evidence, and that GEO can help decide which evidence is worth trying to create in the first place.

Then editorial and technical work can make more of that evidence useful and discoverable. That’s a narrower claim than “video is good for GEO”, but a much more interesting one because the organisation controls the whole process.

The expensive part has already happened

There’s also a fairly mundane economic argument underneath all this.

If an organisation has already paid for the customer access, the agency, the crew, the location and the edit, then asking several better questions during the interview costs approximately nothing. The same is true afterwards: producing a transcript and turning two strong answers into usable web material is trivial next to the cost of arranging that access again.

Access is the scarce input. Everything else is comparatively cheap, and organisations routinely spend heavily on the cheap part while discarding most of the scarce one.

Which suggests a specific and limited role for editorial teams around video. Not another sign-off layer, not attempting to become producers, and definitely not turning every interview into twenty mediocre pieces of repurposed content.

Before a strategically important shoot, ask:

Which audience questions matter here?
Which of those can we already answer well?
Where do we have a content gap, and where do we actually have an evidence gap?
What can this particular person tell us that nobody else can?
Are the existing interview questions likely to acquire that evidence?

And afterwards:

What unique knowledge did we capture?
Where should it live in a crawlable form?
What unanswered questions did the interview create?

That’s the Evidence Engine applied to video production, and it connects directly to the role I described in Becoming the GEO journalist: using demand intelligence not only to decide what to publish, but occasionally to decide what to go and learn.

Events can use the same question bank

The same model extends naturally to events, without requiring a separate “event GEO strategy”. If an editorial team already has a small set of important questions where the organisation’s answer is weak, an event becomes another opportunity to find the right source.

You do not need to ask every attendee the same list. Instead ask: which unresolved question is this person unusually qualified to answer?

That turns event coverage from “what can we make while we’re here?” into “what can we learn while we have access?”

The resulting evidence can still become video, social clips, event coverage and customer content. But those formats are downstream of a clearer reporting purpose.

I’d test this rather than build it

I don’t think the next move is a video GEO framework rolled out across every production, and I’d be wary of anybody who does.

I’d try it on one existing video series or customer shoot.

Choose a handful of important audience questions. Check what the organisation already knows. Identify one or two places where the current answer is under-evidenced. Cross-check those gaps against the questions the producer already plans to ask. Add or sharpen only the prompts that genuinely matter.

Make the normal film.

Then publish the strongest underlying evidence in written and video form, and watch what happens over the following months — whether the written answer becomes more specific and useful, whether the evidence gets reused, whether the page surfaces for relevant searches, whether it appears in AI-generated answers, and whether the organisation has gained knowledge it did not have before.

The result might be nothing much. That would be worth knowing too, and it would be considerably cheaper to establish than the alternative of assuming it works.

But if B2B organisations are going to keep spending at this level on original video, the question is no longer only whether they are getting enough value from the footage.

It is whether they are using the access to answer the questions they still do not know how to answer.