Generative Engine Optimization (GEO) is the practice of making a brand visible in AI-generated answers — the responses from ChatGPT, Perplexity, Google AI, and Claude that increasingly replace traditional search results. The Prompting Company is a GEO platform that measures how often AI recommends a product (share of voice) and closes the gaps with content built to be cited. As people describe problems to AI instead of typing keywords into Google — and coding agents recommend and wire in tools without a search ever happening — products that are absent from AI answers are invisible to those buyers.

In April 2026, our co-founder Michelle Marcelline joined host Liza Vilnits on Beyond Trending's "Beyond the Headline" to unpack the Fast Company article "Why AI Visibility, GEO, and AEO Are the Future of Marketing." This is an edited Q&A, condensed for length and clarity; wording is otherwise preserved. Watch the full 39-minute episode here.

Key takeaways

• GEO applies the logic of SEO to AI: if your brand isn't mentioned in AI answers, it's practically invisible to users who ask AI instead of Google.

• Share of voice is measured by running ~100 long-tail buyer prompts across four engines (ChatGPT, Perplexity, Google AI, Claude) daily; it's the percentage of runs that mention your brand.

• Developers already discover tools through coding agents: an AI assistant in Cursor or Claude Code recommends a tool like Resend and adds the code snippet in the same step.

• AI answers cite source URLs, and the mix of cited sources (company blogs, Reddit, Hacker News, dev.to) tells a brand exactly where to publish.

• Anyone reading Google's AI Overviews instead of scrolling the blue links is already using AI search, not traditional search.

• Intent in AI search is higher than in Google search: users describe their problem in detail, so a targeted answer is far more likely to convert.

Why are coding agents changing software discovery?

Liza: Based on who you work with at The Prompting Company, where are you seeing growth?

Michelle: Our early customers were developer tools, because developers naturally use AI as part of their daily workflow. Even though they've studied computer science for the longest time, they still use tools like Cursor, or now Claude Code, to vibe-code, or at least as their coding assistant.

Say you've finished building a web app using Cursor, and now you want to send a welcome email to everyone who signs up. It's just so natural to ask your AI sidebar for a tool recommendation for sending that welcome email. The AI assistant would recommend, for example, Resend, add the Resend code snippet to your code, and you'd just go to the Resend website and grab an API key for it to work. It's so natural for developers to discover tools through AI.

So our initial target was developer tools. Then it branched out to other tech companies, like Zapier, Snapchat, NVIDIA.

What is GEO and AEO?

Liza: For folks hearing terms like GEO or AEO for the first time, can you break them down?

Michelle: I hope everyone is familiar with SEO: search engine optimization. It's how you optimize your brand to be visible on Google, because if you're not on the first page, your brand is practically invisible. This is the same concept, but for AI. You want your brand to be recommended by ChatGPT, because as search shifts from Google to AI answers, if you're not mentioned there, you're practically invisible.

Liza: Are we looking at an evolution of search, or something more transformative?

Michelle: This is a new way of searching. Before, people would go to Google, and since Google can't understand full sentences, you tended to phrase your question as keywords. Now that LLMs actually understand what a question is, people type full sentences. They describe their problem in detail, and you let the AI find the keywords, run the search, and give you the result.

And it's better for us, because we get more relevant results, faster. Back then you'd end up clicking through blog articles, some of them outdated, recommending products that are no longer on the market. With AI, the research is performed for you, and you get the name of the product and a link you can just click.

How do you measure AI share of voice?

Liza: How do marketers need to shift their mindset? What should they actually be doing?

Michelle: There are two things a marketer should track. The first is your brand's visibility in AI answers. Come up with a list of questions your potential buyers are asking. Go deep into your features, your personas, your use cases. Then run those prompts across different AI models: ChatGPT, Perplexity, Google AI, Claude, and see how many times your product is mentioned. From that you get share of voice: how often you're mentioned in AI answers. AI is probabilistic, so the key is to run it across different engines and multiple times, so the number you get is the average.

The second is AI traffic. Once you know the prompts where you're mentioned and the prompts where you're not, you create content for the gaps, so that your content gets cited by AI and your product gets recommended. AI wouldn't know what your product is good at if there's no content about it.

Liza: Can you unpack the share of voice side a bit more?

Michelle: We come up with a list of questions, as long-tailed as possible: specific to a feature, a use case, or a persona. From the top five topics you want to track, you might end up with 100 different prompts. You run each prompt across four different models, ChatGPT, Perplexity, Google AI, and Claude, every day. So you have 100 prompts, times four engines, times seven days a week. You'll have thousands of runs.

Here's how the math works. Say your product is mentioned in ChatGPT when I run a prompt today: your share of voice is 100%. I run the exact same question on Google AI and you're not mentioned: it's down to 50%. I run it again on Claude and you're mentioned: two mentions out of three runs, 67%. You keep doing that thousands of times, and if the final number is, say, 30%, that's your average share of voice.

How does content earn citations in AI answers?

Liza: Are there hard-line things brands need to be doing to get their content surfaced by AI?

Michelle: Right now it's mostly identifying the gaps: which prompts you're winning, meaning your brand is consistently mentioned, and which you're not. Usually you're missing because you don't have enough content touching those topics.

When we run prompts across the engines, the answers come with the URLs they cite as sources. Those URLs are what got your competitors mentioned. So when we generate an article, we generate it based on your company's information, but as similar as possible to the articles being cited. If those articles are cited and we create something similar, our article will also be cited. The difference is that ours contains information about your product, so your product gets recommended next.

It also matters where the citations live. Check which kinds of URLs the LLMs cite in your category. If a lot of them are company-owned blog posts, publish on your own site. For some tech companies, it's posts on Hacker News or developer sites like dev.to.

Is the shift to AI search already happening?

Liza: How quickly do you see the shift from Google happening at larger scale?

Michelle: It's already happening. When ChatGPT launched in late 2022, only tech people used it, and a majority of people still use Google search. But when you search on Google right now, do you actually scroll through the ten blue links, or do you just look at the AI Overviews result? If you look at AI Overviews, you're essentially using Google AI, not Google search.

How does The Prompting Company approach GEO?

Liza: I still have my caveat about doing your homework. Hallucinations happen. Do you see that day to day?

Michelle: Hallucinations happen a lot, but it has gotten better, and it has become easier to fact-check, because answers now come with the URLs behind them. Always check the citations for anything that seems iffy.

It's the same with how we run our tool. We have a couple of layers of fact-checking. We don't generate content in one shot. First we gather information about your company, the problem, and what other people have written. We come up with a content brief, just bullet points, and the user confirms it and catches anything hallucinated. Then we verify the writing style: for a developer tool, you want to be the thought leader in the space, so it has to be more professional. And then we generate the article. We have an AI for each step, one that does the research, one that determines the tone, one that generates the brief, because each AI is trained to do different things, and each can fact-check different things.

What does the shift mean beyond marketing?

Liza: Any broader AI trends you're keeping an eye on?

Michelle: Five years ago, people built software for a human to use. With AI, the goal is to replace a workflow. So the trend I'm seeing is to identify the most repetitive workflow within an industry and try to replace that.

It is scary, but technology doesn't only take jobs, it enables more. One thing I like to say: before there were freeways, did we have more traffic or less? The answer is more, because they enabled people to travel more. Before the car, you'd only travel in your area. With cars, you travel between states. With airplanes, to different countries. It's not stopping you from doing something; it's enabling you to do more. It's a transition period, and there will be more jobs to take, just different kinds.

Liza: For folks who want to learn more, where should they go?

Michelle: I'm on LinkedIn, Michelle Marcelline; I try to post research at least once a week. And to learn about The Prompting Company, go to promptingcompany.com.

About the author: Michelle Marcelline is the co-founder of The Prompting Company, a Generative Engine Optimization platform backed by Y Combinator with $6.5M in seed funding. She is a two-time founder and a Forbes 30 Under 30 recipient. The Prompting Company helps products get discovered, chosen, and used by AI agents.