With so many people using ChatGPT like a search engine, the natural next question for marketers, brand strategists, and PR agencies is: How can I make sure my content makes it into AI responses?

The result is a flood of AI influencers proclaiming that SEO is dead and GEO (Generative Engine Optimization) is the next big strategy to master. But a few weeks ago, Benjamin Houy shut down his GEO platform, Lorelight – not due to lack of demand, he had paying customers – but because he doesn’t think it’s a necessary product to pay for.

Here’s why he thinks you don’t need an expensive GEO strategy, and what he says you need instead.

What is GEO, and how is it different from SEO?

Ben says that GEO and SEO have a lot of overlap – where they diverge is predictability.

SEO is all about optimizing your website so that Google’s crawler can understand it, assign authority, and rank it for relevant queries. At its core, SEO is structured around knowable ranking factors: content relevance, backlinks, authority, site performance, and a somewhat predictable algorithm.

GEO is the opposite: it’s messy, hyper-personalized, and impossible to fully track. That’s because GEO is more about optimizing how AI models talk about you, not how you rank.

AI systems don’t just search the web, they also pull from:

• Your conversation history
• User memory
• Your browsing behavior
• Real-time context
• Internal training data
• Model-specific quirks

Basically, AI is interpreting all these factors, not indexing by them – so there’s a lot that is out of your control. And those factors shift constantly as models evolve.

For example, ChatGPT currently pulls a disproportionate amount of its insights – something like 40% – from Reddit. So for now, your GEO strategy becomes about getting all over Reddit. But what happens when the next model shift favors Wikipedia – which is close behind at 26% of responses? Or a new site entirely?

Unlike SEO, GEO has no stable playbook you can optimize for.

GEO is currently a short-term strategy

Ben saw these issues firsthand in his customer lifecycle. He witnessed a significant amount of churn at Lorelight because GEO currently has no long-term strategy, for a few reasons.

Problem #1: GEO signals change constantly

Ben describes GEO tactics as “hacks” more than strategies. You’re not building muscle the same way you are with SEO, you’re appeasing the current AI model.

“I haven’t seen any GEO hacks work long-term. Even when people claim they know them, they can’t actually explain what they are.”

Ben’s talked to dozens of customers and marketers who make claims like:

• Putting out more press releases will increase the likelihood of being flagged in AI responses because it emulates the experience of your brand being talked about.
  
  
• You should be adding an LLMs.txt file to your website to tell AI models what pages they should or shouldn’t use for training or reference – similar to how marketers use Robots.txt files for SEO purposes to tell search engine crawlers what pages they can or can’t index.

But there’s nothing to substantiate these strategies yet.

“I haven’t seen any reports of LLMs.txt actually being used by any major LLM,” Ben says

Problem #2: GEO tools rarely offer surprises or real action items

Lorelight saw a trend in what it surfaced to clients, and it wasn’t novel. It tended to flag:

• The same authoritative outlets brands already knew mattered
• The same topics marketers were already writing about

Nothing fundamentally new emerged from GEO analysis. Some teams saw insights on the brand level that intrigued them about how AI talked about their company – one saw a ChatGPT claim that their contracts were “too hard to sign” – but it surfaced very little that changed how they marketed.

In essence, it didn’t uncover any new levers to pull.

“Companies mostly expected a magic bullet. They thought they would log in and it would say, ‘if you do this one thing, suddenly you will be number one in ChatGPT’. Instead they got: "if you keep doing the hard work, then you will show up.”

Problem #3: It’s difficult to see if it’s actually working

The non-deterministic nature of AI makes it quite difficult to not only set a strategy but to validate it. You can’t know whether AI is showing users the same results you're seeing.

“Companies kept asking me, ‘how do I know that’s actually what people search?’ and my answer was basically, ‘you can’t really’.”

Because AI results are personalized and contextual, you can’t reproduce another user’s AI output, even with the same prompt. Not only does this mean you can’t see if your strategy is working, it means no strategy will be 100% effective for everyone.

Forget GEO hacks and build a strong brand

The truth is, brands that show up in AI models are the brands people already know and talk about.

Ben shut down Lorelight because he realized that the best GEO strategy is a strong marketing strategy. There is no shortcut.

You need to:

• Have a product people actually love
• Deliver something useful, compelling, and worth recommending
• Build authentic relationships with customers
• Create meaningful differentiation
• Have a compelling point of view

“Everything companies needed for GEO was the same as what they already needed for good marketing.”

GEO is a side effect of strong brand presence – the same things that mattered before still matter with AI. Case in point: what Ben learned from Lorelight about his other company, a language learning tool called French Together, is that his brand was forgettable. He needed to create something people actually wanted to talk about, so he’s rebranding it to Copycat Cafe.

“Instead of trying to game a system that changes all the time, you may as well focus on fundamentals.”

How to lead your brand into the AI era

Ben set out to help leaders win with AI. And they still can if they’re willing to accept that the lessons of Lorelight are less about how to hack your way into results and more about building the infrastructure to look important to the models.

Here are his most critical takeaways from launching and shuttering a GEO tool:

1. Double down on your current strategies

Focus on getting mentions, strengthening PR efforts, continuing to push a wide variety of content on the topics that matter to your business, and catching the eye of authoritative publications.

Newsworthiness = conversation. Conversation = citations. Citations = AI mentions. This is basically just SEO for the age of AI.

2. Take a hard look at your brand

What do you want people to say about you? Instead of pushing short-lived content optimized for  AI responses, build content around the story you actually want to tell.

“Ask yourself the question: why would people even talk about your company? Don’t be too attached to your opinions – or even to your business.”

AI mirrors the market’s perception, so if your positioning is strong, GEO will follow naturally.

3. Stay open, curious, and unattached.

Don’t become dogmatic about any new AI trend. Hold your strategies lightly, iterate fast – and if the evidence tells you something isn’t adding value, drop it.

Ben caught some flack with VC-backed GEO competitors for “devaluing” the space in his Linkedin post. But his rebuttal: You’re too focused on proving that you’re right to learn that there actually might be a better way.

To sum it up:

• GEO isn’t a hoax, but it isn’t a strategy either. It’s a reflection of what the world already believes about your brand. That’s what needs work if you’re not getting AI mentions.
  
  
• SEO isn’t dead and AI isn’t going to get you out of doing the hard work.