Uberall GEO experiment: YouTube content doubled AI mentions and citations

A 90-day test found that structured, answer-focused video content improved visibility across ChatGPT, Gemini and Perplexity, while exposing a bigger problem for multi-location brands.

SAN DIEGO, Calif. – YouTube may be a more important part of generative engine optimization than many search marketers realize.

That was one of the findings from a 90-day GEO experiment presented by Christian Hustle, Global AI Search Evangelist at Uberall, during BrightonSEO San Diego.

Hustle’s session, “Your GEO strategy works great… in one city,” examined how structured content influenced brand visibility across ChatGPT, Gemini and Perplexity, and why national AI visibility metrics may tell only part of the story for multi-location businesses.

The experiment produced a striking result: among the prompts Uberall was tracking, the brand’s AI mention rate increased from 20% to 40%, while its citation rate increased from 4% to 8%.

But perhaps the more interesting finding was where many of those citations came from.

YouTube.

Phase one: Publish videos and change nothing else

Uberall began the experiment with what amounted to a clean test. The company created videos addressing topics around GEO and AI search, but made no corresponding changes to its website.

Each video followed a deliberate format.

• Titles were written around questions or answers a potential customer might be looking for.
• The introduction quickly stated what the video would cover, generally within the first 30 seconds.
• The body then spent five to 10 minutes explaining the topic, examples and supporting information.
• Each video finished by summarizing the main points.

Hustle’s theory was that this structure would make the transcript easier for AI systems to understand and extract information from. The approach emphasized clear definitions and repeated explanations of important concepts from different angles.

48 hours to Success!

Within roughly 48 hours, Hustle said Uberall began seeing movement. Gemini was already surfacing the company’s YouTube channel and podcast for some of the GEO-related prompts being monitored.

As you know, traditional SEO experiments can take weeks or months before a measurable pattern emerges. In this case, Hustle said changes began appearing within days.

YouTube became a major citation source

At the beginning of the experiment, YouTube was roughly the 30th-ranked citation source within Uberall’s monitored data, according to Hustle. During the experiment, it climbed as high as No. 2 for the targeted prompt set.

For some queries, YouTube was cited more frequently than Uberall’s own website. That does not mean publishing videos automatically improves AI visibility. The experiment was limited to one brand, its prompt set and the AI platforms being monitored.

Consistency

Hustle attributed the gains partly to consistency. The videos repeatedly defined concepts such as GEO using similar language, while approaching those concepts from different questions and scenarios. His view was that this gave AI systems multiple pieces of corroborating material rather than one isolated page explaining a subject.

Phase two: Bring the transcripts onto the website

Uberall then expanded the experiment.

• The company embedded the videos on its website and published the transcripts alongside them.
• Those pages included conventional SEO elements such as headings, key takeaways, timestamps and other structured on-page information.
• That created two versions of the material:
  • the content living on YouTube and
  • a structured representation of the same information on Uberall’s domain.

The effect appeared to extend beyond the individual prompts directly targeted by the videos. Hustle said Uberall began seeing improvement across a broader portion of its tracked prompt library, suggesting that the impact was no longer limited to exact questions addressed in individual videos.

Across the larger data set, Uberall.com remained the company’s dominant citation source, but YouTube continued appearing among the leading sources across hundreds of monitored prompts.

Hustle reduced the experiment to three concepts that marketers can apply to GEO:

        Consistency. Repetition. Corroboration.

That does not mean publishing the same page or video repeatedly. Hustle said, that instead, brands can explain important concepts consistently across multiple questions, formats and trusted platforms.

For example, a company might answer a customer question in a video, publish the transcript on its website, reinforce the same expertise through social content and have third-party sources independently support the brand’s claims. The goal is to create a body of evidence rather than a single optimized page.

This represents an important difference between traditional SEO thinking and emerging GEO strategies. A search engine can rank an individual document.  An AI answer, however, may assemble its response after examining several sources, comparing claims and deciding which brands or entities it has enough evidence to recommend.

As Hustle put it, AI systems appear to favor answers they can more easily defend.

The local GEO problem

That leads to the second half of Uberall’s findings.

• A brand can look strong when AI visibility is measured nationally while performing poorly for the location-level questions customers actually ask.
  • A query such as “best seafood restaurant in San Diego” creates a very different competitive environment than a broad query about the best national restaurant chain.
  • The same is true for hearing clinics, pest control companies, auto repair businesses, coffee shops and virtually any multi-location company.

• Different locations have different competitors, reviews, citations and local signals.

Uberall’s argument is that AI effectively reconstructs the competitive set for each location. That means an aggregate GEO dashboard can show rising share of voice while individual stores or offices remain absent from valuable local AI recommendations.

The official BrightonSEO description of Hustle’s session summarizes the problem neatly: strong aggregate share of voice can still hide hundreds of local losses.

What SEOs should test

The lesson isn’t simply “make more YouTube videos.” The more useful takeaway is to rethink how authoritative information about a brand is distributed:

• Start with the situations and questions customers actually bring to AI systems, rather than converting a traditional keyword list directly into prompts.
• Then create answer-focused content around those scenarios.
• Structure the information clearly. Keep important descriptions and definitions consistent.
• Publish it in multiple formats where appropriate.
• Connect video content back to detailed information on the website.
• Finally measure AI visibility by topic, platform and, when relevant, location.

Hustle also cautioned that GEO is changing quickly. Citation patterns that appear important today can shift rapidly as AI platforms alter retrieval systems, source preferences and answer generation. That makes continued experimentation more useful than treating any single GEO tactic as a permanent ranking factor.

For Uberall’s test, however, YouTube remained a significant citation source after the initial experiment ended.

The broader lesson may be that GEO is becoming less about optimizing one page for one query and more about giving AI systems a consistent collection of evidence about what a company knows, does and can credibly be recommended for.

For multi-location brands, there is one additional catch. You still have to prove it in every city.

Uberall DL: