INSIGHTS / B2B GEO RESEARCH

Should B2B Teams Make YouTube Videos for AI Search?

YouTube deserves a place in a B2B content plan when a video helps buyers assess something real. A process demo, an installation constraint or a product test may be easier to judge on screen. The AI search data makes video worth a test. It does not, on its own, show how much you should spend.

7 min read

TL;DR

  • Make a video when buyers need to see a process, product or trade-off that text alone explains poorly.
  • A strong link between YouTube brand mentions and AI visibility does not prove that starting a channel will increase citations.
  • A study's 94% long-form share describes observed YouTube citations, not the chance that a new video will be cited.
  • Give a useful video a clear written companion, then track buyer use, AI citations and inquiries as separate results.

What a buyer can learn from the footage

Start with what a buyer can learn from the footage. A video that only reads out a sales page adds little evidence. A clear demo may answer a question that the page leaves open.

What the research says about YouTube

Ahrefs studied 75,000 brands and reported an approximate 0.737 correlation between YouTube mentions and AI brand visibility. Mentions included video titles, transcripts and descriptions. The sample filtered for domains with Domain Rating above 40 and a highest-volume keyword of at least 800 monthly searches.

That is a link seen across brands. It is not a test of what happens when a firm posts a video. A well-known brand may get both more video coverage and more AI mentions. Nor does a mention in someone else's video mean the brand owns an active channel. Use the finding to look at video use in your market. It cannot tell you what your next upload will earn.

OtterlyAI's March 2026 study examined more than 100 million citation instances over 30 days. Among its YouTube citations, 94% concerned long-form videos, 5.7% Shorts and 0.3% other formats. Its video sample covered work that AI had already cited. Read the result as a description of that group.

What the research says about YouTube
YouTube formatShare of YouTube citation instances
Long-form videos94%
Shorts5.7%
Playlists, channels and livestreams combined0.3%

Source: OtterlyAI, March 2026. Shares describe observed YouTube citations, not all AI citations or the success rate of published videos.

The missing comparison matters. We do not know the mix of all videos the AI could have used. These shares therefore cannot show which format has a better chance of being picked. A small number of heavily cited videos could also account for many citation instances. Ask whether buyers need a source they can return to.

Figure 1. The format mix within observed YouTube citations% of YouTube citation instances in the study
Long-form videos94%
Shorts5.7%
Playlists, channels and livestreams combined0.3%
050100

Already-cited YouTube material in a 30-day study. These are not shares of all AI citations or selection rates for new videos. OtterlyAI, March 2026.

Brand presence and source selection are different results

Brand presence and source selection are different results
StudySelected result and methodUseful lessonDecision it cannot settle
Ahrefs, December 2025Approx. 0.737 brand-level correlation; 75,000 brands with entry filtersYouTube discussion is worth examining alongside wider brand presenceWhether an owned channel will cause more AI mentions
OtterlyAI, March 202694% long-form share among observed YouTube citationsReference video is a plausible format to testWhether long-form beats Shorts for an otherwise equivalent new video

These findings fit together without measuring the same thing. One concerns which brands appear in AI answers; the other concerns the format of cited YouTube material. Both firms sell search tools. Neither study tests what a small B2B firm gains by starting a channel.

That leaves a useful, narrower conclusion: find where video could show buyers something they still need to see. Avoid turning a broad citation pattern into a rule to publish more footage.

Start with a decision that benefits from seeing

Use this editorial decision table before a shoot. Use it to choose the job each piece should do. It is not an AI ranking rule.

Start with a decision that benefits from seeing
Buyer questionWhat a video can showWritten companionKeep the claim bounded
Will this fit our workflow?A complete task on a stated setupRequirements and supported versionsShow which steps were manual or excluded
Can this machine handle our material?A test with a named input and conditionsTest method and result tableDo not generalize one test to every material
How difficult is installation?Access, tools and assembly sequenceDimensions, prerequisites and exceptionsState when specialist installation is needed
How do two service options differ?The work performed under each scopeA fair scope comparisonKeep footage consistent with the actual terms

For example, take a fictional firm that sells packing machines. Buyers keep asking how the machine switches from one job to the next. Its sales page says the machine is flexible but does not show the work. An engineer could film the switch, name the starting setup and show each change made.

The companion page could record the setup, steps and limits, with links to approved specs. Buyers could see the proof and share the details with their team. This is an example of a plan, not a client result. It does not show that the pair will earn citations.

Pair the formats when each adds value

A written companion should do more than paste in an unchecked transcript. It can state the question, explain conditions, provide a readable result table and link to the right part of the demo. Check that the video and page describe the same version and scope.

An OtterlyAI follow-up tracked eight interview stories in video and article form. Written pages earned 79.5% of format citations on branded prompts in its early observation window. That small study did not assign formats at random. Still, it gives a reason to test text alongside video. It does not establish that every B2B story needs both formats.

For a team with limited resources, begin with one repeated buyer question and one piece of evidence you can show well. Set a goal before filming. You might aim for fewer calls to clear up basic points, better briefs from buyers, or use by sales staff. Track citations as a separate result.

YouTube and B2B AI Search: Common Questions

Do we need a large channel before testing?

You do not need a large channel to make a useful buyer demo. The cited studies do not give a minimum subscriber count for success. Start from a buyer need you can serve and a result you can measure.

Should we stop making Shorts?

No. Citation share is only one measure of a format's use. A Short may serve discovery or a quick update, while a fuller demo helps buyers assess the work. Choose the format for its task.

Does a brand mention count as a recommendation?

No. An answer can name a brand without endorsing it. It can cite a video without sending a visitor to your site. Define those outcomes separately when reviewing results.

What should we check after publication?

Check whether buyers can understand the setup and limits, whether the written page agrees with the video, and what answers actually cite. Record the question, platform, date and source URL. Do not treat one appearance as a stable trend.

Discuss a focused video test

If your team has useful footage but cannot link it to buyer questions, bring a sample video and its related page to SemShift. We can discuss which evidence is missing and how a focused video test could fit your GEO work.