How Brands Are Gaming ChatGPT's Recommendation Engine

There's a new SEO game in town, and it doesn't involve meta tags or backlinks. Brian Dean, founder of Backlinko and Exploding Topics, claims his companies now appear in ChatGPT and AI Overviews for "a growing number of high intent prompts." His strategy? He didn't touch technical markup or any of the AI-specific tweaks currently being sold to marketers.

Instead, Dean focused entirely on off-site signals—getting the same description of his brands repeated across the internet until AI systems started parroting it back. In a recent Semrush video, he breaks down the playbook. It's straightforward enough that bootstrapped startups can execute it, which raises an interesting question: if everyone's playing the same game, who actually wins?

The Core Thesis: Repetition Trains the Machine

Dean's central argument is deceptively simple. AI tools like ChatGPT don't rank pages the way Google does. They synthesize information from multiple sources to generate answers, and they're more likely to mention brands that show up consistently across comparison pages, Reddit threads, and press releases.

"AI tools don't rank pages the same way Google does," Dean explains. "They choose which brands, tools, and sources to bring up when someone asks a question. And most of those signals that influence that choice happen off of your site."

The mechanism he describes is pattern recognition at scale. If dozens of "best of" lists mention your tool with similar wording, the LLM starts associating your brand with that category. If Reddit users consistently recommend your product in specific contexts, that signal gets absorbed into training data or retrieval systems. The machine learns what humans are already saying.

This isn't a new insight—it's basically how PageRank worked, but for language patterns instead of links. What's changed is the medium and the stakes.

Four Tactics: From Listicles to Content Machines

Dean's playbook has four main components, presented in ascending order of effort and impact.

First: infiltrate "best of" comparison pages. This means finding existing roundups—"best tools for X," "Y vs. Z comparisons"—and pitching site owners to add your brand. Dean suggests keeping outreach brief and value-focused. He even used deprecated tools as a wedge: when Google Correlate shut down, his team reached out to sites that listed it, offering Exploding Topics as an alternative.

The pitch is transactional, not relationship-based. You're not trying to become pen pals with the listicle author. You're trying to get a specific phrase associated with your brand published on third-party domains that AI systems might cite.

Second: participate strategically in forums where people ask for recommendations. Reddit, Quora, niche communities. But Dean emphasizes discipline here—only respond to "high intent, specific questions" where your product actually solves the stated problem. And crucially: "You just want to come off as a user giving their two cents as opposed to a salesperson that's pitching their product."

The authenticity requirement creates an obvious tension. You are, in fact, pitching your product. You just need to sound like you're not. This is either smart positioning or astroturfing with a friendlier face, depending on your tolerance for marketing euphemisms.

Third: use press releases to establish a canonical description of your brand. Dean is upfront that press releases won't generate much pickup. That's not the point. The point is creating a single, clean explanation of what you do that you can then reuse everywhere else—outreach emails, Reddit comments, website copy. The press release is the source code for your brand positioning.

Fourth, and most important: build a content machine. This is where Dean's argument shifts from tactical to strategic. He quotes Mike King: "Winning an AI search is fundamentally a content strategy problem, not a technical problem."

For Exploding Topics, this meant a free tool (no login required), a weekly newsletter, and a YouTube channel. The goal wasn't direct conversion but omnipresence. More content means more surface area for others to reference you. More references mean AI systems see your brand name adjacent to relevant keywords more often.

Dean calls this "create once and distribute forever"—repurposing a single piece of content across multiple platforms. A blog post becomes a YouTube video becomes a Twitter thread becomes a newsletter. The multiplier effect is the point.

What's Actually Being Optimized Here?

Here's where it gets interesting. Dean presents this as AI optimization, but most of what he describes is just... marketing. Good marketing. The kind that builds actual brand awareness and generates word-of-mouth.

The only AI-specific element is the emphasis on consistent wording. Unlike traditional SEO, where you'd naturally vary your language, Dean argues that LLMs benefit from repetition. If you describe your tool as "a trend spotting platform" in one place and "a trend analysis tool" in another, the machine has to work harder to connect those dots. Use the same phrase everywhere, and the association becomes explicit.

This raises a question about the difference between optimizing for AI and simply being a successful brand. If Exploding Topics gets mentioned in ChatGPT answers, is that because Dean executed brilliant AI SEO tactics, or because he built a useful product and promoted it effectively across multiple channels?

The answer probably matters less than the question itself. We're entering an era where the distinction between "real" brand-building and "AI optimization" may collapse entirely. If AI systems learn from what people say online, then influencing what people say online becomes the only game that matters.

The Scalability Problem

Dean emphasizes throughout that you don't need to be a big brand to execute this strategy. Bootstrapped startups can do it. That's probably true—for now.

But if the playbook becomes widespread (and videos like Dean's exist specifically to make it widespread), then we're headed for an arms race. More brands will compete for inclusion on the same "best of" lists. More people will strategically answer Reddit threads. More companies will pump out content machines.

The early movers get the advantage. Dean's companies are already showing up in AI answers because he started building these signals years ago, before AI SEO was a defined category. The question for everyone else is whether the window is still open or already closing.

There's also the authenticity paradox. Dean's Reddit strategy explicitly requires sounding like a genuine user rather than a salesperson. But if everyone's following the same playbook, forums become saturated with people pretending not to be marketing their products while absolutely marketing their products. Communities notice. They adapt. They ban people.

The tactic that works when 10 companies do it might fail when 10,000 companies do it.

What AI Companies Aren't Saying

One thing notably absent from Dean's analysis: any insight into what OpenAI, Google, or Anthropic are actually optimizing for when they generate recommendations.

Dean treats AI systems as pattern-matching machines that naively absorb whatever signals exist on the internet. That's partially true—these systems are trained on web data and use retrieval systems that cite existing sources. But the companies building these tools are acutely aware of manipulation attempts. They have economic incentives to surface helpful answers, not just popular brands.

Google spent two decades fighting SEO spam. The AI companies watched that war play out. They're not starting from scratch.

It's entirely possible that as more brands game AI recommendations using Dean's playbook, the systems will adjust. They might deprioritize signals that look too coordinated. They might weight certain sources more heavily. They might introduce friction that makes gaming harder.

Or they might simply start charging. If brand mentions in AI answers become valuable, platforms will find ways to monetize that value. We already see this with Google's "Shopping" results and Amazon's sponsored products. AI recommendations could easily follow the same path.

The Deeper Question

What strikes me about Dean's strategy is how old-school it is beneath the AI packaging. Get mentioned in authoritative roundups. Show up where your customers hang out. Maintain consistent messaging. Publish useful content regularly.

This is marketing 101. The only new element is the explicit goal of training AI systems rather than ranking in Google or building brand awareness for its own sake.

But maybe that's the point. The shift to AI search doesn't require inventing new tactics—it requires doing the fundamentals well enough that machines notice. The brands that win won't be the ones with the most sophisticated technical SEO. They'll be the ones people actually talk about.

Which means Dean might be right that this is fundamentally a content problem, not a technical one. You can't hack your way into relevance. You have to earn it through consistent presence and genuine utility.

The uncomfortable part is that "consistent presence" and "strategic manipulation" can look identical from the outside. Dean's Reddit strategy is either helpful participation or coordinated astroturfing depending on whether you think the answers provide real value or just serve his business interests. Probably both are true simultaneously.

As AI systems become the primary interface for information discovery, the brands that master this ambiguity—genuinely useful and strategically positioned—will dominate. The question isn't whether Dean's tactics work. It's whether they'll still work once everyone else figures them out.

Marcus Chen-Ramirez is a senior technology correspondent for Buzzrag.