Test, don't theorize. Cassie Clark on GEO and video.

Cassie Clark is a fractional content strategist and AI search optimization expert. She has spent months running live experiments on her own personal brand and her own website as a sandbox before taking any of it to clients. We talked about why SEO is still the foundation under GEO, how to actually measure if you are showing up in AI answers, and the bad-hair-day Q&A video that landed her in Google's AI Overviews for months.

https://www.youtube.com/watch?v=CC6hqR5KM_E

Key Takeaways

• SEO is still the foundation. GEO is the context layer that gets you cited in AI answers on top of it.

• Track recommendation rate and AI share of voice weekly — not once. LLMs are probabilistic and give consensus answers, not single answers.

• A low-budget Q&A video filmed on a phone with no editing can show up in AI Overviews and stay for months.

• Build a Q&A video plan from customer language. Use the top 10 questions or queries your customers actually ask.

• Agentic commerce is coming. If you are not in the answers when agents start buying on behalf of humans, you will not be in the answers when it matters.

How is GEO different from SEO?

SEO is the foundation. GEO is what gets you cited.

Cassie frames it cleanly. SEO indexes you so a search engine can present you as an answer. GEO is the context layer that determines whether AI search engines actually cite your brand when they generate that answer. You need both. SEO without GEO leaves you indexed but invisible to AI search. GEO without SEO is a house on sand.

The labels matter less than the goal. AEO, AIO, GEO — Cassie groups them together. "What we're talking about is showing up in AI search results. It's just that simple, no matter what we call it."

There is also a Bing layer that most B2B marketers underweight. "Bing is, like, really big now across a couple of different platforms," Cassie said. The search interface behind Perplexity and a few other AI tools is built on Bing. If your brand is invisible on Bing, you are invisible to the layer feeding several AI search engines.

The other shift is how broad the surface area is. GEO is not a keyword exercise. You need to be present on video, on LinkedIn, in Reddit comments where they actually matter for your category, and your positioning has to be crystal clear across every surface. A tight marketing strategy is the prerequisite, not the byproduct.

Why do LLMs give different answers every time?

Because they are probabilistic, not deterministic.

This is the source of every "is GEO even measurable?" debate. Two people ask the same AI engine the same question and they can get different answers. Run the same prompt five times and you might get five overlapping but different responses. So how can you measure anything?

Cassie's answer: stop looking for a single answer and start looking for consensus.

"These engines are probabilistic, so they don't give the same answers consistently, but they give a kind of consensus," Cassie said. "If you're looking for the best software for my HR team, it might give you five different suggestions, but those same five different suggestions might show up repeatedly with one or two filtering in and out based on how many times you run the prompt."

That means the measurable variables are:

• Recommendation rate. How often does your brand show up in the answer for a given prompt across multiple runs?

• AI share of voice. When the AI is naming brands in a category, how often is yours one of them, and are you helping shape the answer or just present in it?

You run the prompt multiple times. You average. You track on a weekly basis. You watch the trend, not the snapshot. "You can't just run it one time and be like, oh yeah, we're in," Cassie said. "That's not how it works."

What does a low-budget GEO video test look like?

A phone, a question, an answer, and a year of patience.

This is the most quotable part of the conversation. Cassie ran the test herself, on her own brand, with no production budget and no editing. The story is now her favorite proof point.

"I filmed it on my phone, no editing. It's, like, very embarrassing, please don't go look at these old videos."

She picked one specific question that her audience was searching: "What is AI share of voice?" Then she made a single Q&A video where the title was the question and the video defined the term, explained how to track it, and walked through what it means. That was it.

"It did, and I was really mad about it because it was a day I had a bad hair day. And, like, here it is in the AI Overview and it stayed there for months." — Cassie Clark

The bad-hair-day video showed up in Google AI Overviews. And it stayed there for months. Not a polished, scripted, edited, branded production. A phone video. The question is the title. The answer is the body. The format does the work.

For the 60-second version of this story in Cassie's own words, watch the short clip.

How should a B2B brand actually run this test?

Cassie laid out the test any team can run this month:

1. Pull customer language from your audio recordings and call transcripts. Use the words your customers actually use, not your internal marketing language.

2. Pick the top 10 questions or queries your customers are searching.

3. Make a video question-and-answer for each one, discreetly on its own. No multi-topic videos. One question, one answer.

4. Post them. No fluff, no production overkill, no waiting for the perfect script.

5. Check the queries once a month, once a quarter. Watch if any of your videos show up in AI Overviews, ChatGPT, or Gemini answers.

6. When one shows up, track how long it stays. Eventually it will be replaced by something fresher. The data you collect tells you what is working.

The timeline is realistic. Cassie says 6 months is roughly when you should expect to start seeing yourself show up consistently. Check monthly. Stay patient.

How should you verify any of this for yourself?

"I am full-on test and experiment and track the things."

That is Cassie's answer in one sentence. The space is noisy. Every study says something different. Every vendor has a product. The way to cut through is to run your own queries against your own category and see what shows up.

Start with a video strategy. Pick a cadence — one or two long-form videos a week, or a sprint of one a day if you have the bandwidth and want to compress the experiment. Then start running your category queries against ChatGPT, Perplexity, Gemini, and Google AI Overviews. Write down which queries cite you, when, and for how long. Numbers are what leadership wants to see, so the math is also the political case for the program.

This is the GEO version of what good marketers have always done with paid media. Test, measure, adjust. The only difference is the surface area.

Where is AI search headed next?

Agentic commerce.

Cassie says she has been "dusting off the crystal ball" on this one. The shift is from a human running a prompt and reading an answer to an AI agent running prompts and acting on the answer. Booking the demo. Making the purchase. Filling the form.

When there is no person in the loop, the cost of not being in the answer is no longer a missed click. It is a missed sale.

"If agentic commerce is eventually going to be making a purchase on behalf of someone or signing up for a demo, and there's not a person in there pushing the button, it's going to be so much harder to be the answer when that happens if you've not already done the work now."

The doing-the-work-now list, in Cassie's framing:

• Active videos wherever you are posting them. YouTube first.

• Active presence on Reddit so models learn what your brand is, who it serves, what the use cases are.

• Specific, repeated, consistent positioning everywhere.

You want to be in the answers before the agents start asking. By the time agentic commerce is widespread, the brands that win will already be embedded in the training data and citation graph. The brands scrambling will be too late.

Frequently Asked Questions

What is the difference between SEO and GEO?

SEO is the foundation that makes your content indexable and findable by search engines. GEO, or Generative Engine Optimization, is the context layer that determines whether AI search engines like ChatGPT, Perplexity, and Google AI Overviews actually cite your brand when they generate answers. You need both. SEO without GEO leaves you indexed but invisible to AI search.

How do you actually measure if a brand is showing up in AI answers?

Track two things weekly: recommendation rate (how often your brand shows up in the answer for a given prompt across multiple runs) and AI share of voice (how often your brand is named in the answer for a category, and whether it is shaping the answer). LLMs are probabilistic, so you run prompts multiple times and average. A single run is not data — a trend over weeks is.

Does a low-budget phone video really show up in AI search results?

Yes. Cassie Clark tested it on her own brand. She filmed a single Q&A video on her phone, no editing, on what she calls a bad hair day. The video answered one specific question her audience was searching ("What is AI share of voice?"). It showed up in Google AI Overviews and stayed there for months.

How long does it take to start showing up in AI answers?

Roughly 6 months is when you should expect to start seeing yourself show up consistently. Check monthly. Track which queries cite you and how long the citations last. The signal is the trend across months, not the snapshot of a single week.

What should B2B marketers do to prepare for agentic commerce?

Get into the answers now. Publish active videos wherever you post (YouTube, LinkedIn). Be active on Reddit so AI models learn what your brand is, who it serves, and what the use cases are. Keep positioning specific, repeated, and consistent everywhere. When AI agents start buying and booking demos on behalf of humans, the brands already embedded in the citation graph will get picked. Those scrambling will not.