How to Rank in ChatGPT

James Lawrence: Welcome back to the Smarter Marketer Podcast. I'm joined today by Joe Alder, Rocket's Head of SEO. Joe, how are you, mate?

Joe Alder: I'm good, thank you. Hi, James. Good to be here.

James Lawrence: What have you been up to today, Joe?

Joe Alder: Catching up on work. I spent a lot of time on the webinar this week.

James Lawrence: Great to hear. Obviously, it’s a topic of massive interest and importance to Australian marketers: the rise of ChatGPT. More specifically—how do you actually rank as prominently as possible in ChatGPT? How do you take advantage of what has become such an important channel in the marketplace? Joe’s role at Rocket is Head of SEO. Since ChatGPT launched in November 2022, everything Rocket has done around ChatGPT optimization has sat within Joe’s team. Over the last 18 months, we’ve been running optimization campaigns, doing first-party research, and studying the top 1,000 Australian websites monthly. That’s meant analyzing 68 million keyword queries and over 100 independent studies. There’s so much noise in the market about “how to optimize for ChatGPT,” but a lot of it isn’t based on running campaigns. The benefit Joe brings is he’s managing a team that’s at the coalface—testing what works, seeing what doesn’t, and tracking the impact of large language models. So, Joe—where are we at in the marketplace in terms of ChatGPT usage as of mid-2025?

Joe Alder: The first step, as everyone knows, is that it was the fastest-growing consumer app—100 million users in two months. For comparison, TikTok took seven or eight months. Since then it’s only grown. Right now, it’s around 700–800 million weekly active users, close to a billion. Almost 80% of knowledge workers are using ChatGPT. That’s important because they’re a huge part of the consumer base—people with cash to spend on products and services. From our research, ChatGPT currently holds about 85% of the LLM market share. So if people are coming to your site from LLMs, 85% of the time it’s from ChatGPT versus Gemini, Claude, Perplexity, and others. It’s basically the “Google” equivalent of LLMs. And finally, it’s the third most searched site in Australia, behind YouTube and another one we won’t mention.

James Lawrence: We do allow swearing on this podcast! Very British of you to politely avoid mentioning the large adult site. But yes, 34% of adults are using ChatGPT monthly, and 78% of knowledge workers. In a B2B context, HubSpot research found that 48% of B2B buyers and 39% of enterprise buyers are using AI tools to research software, plus 30% of consumers across the board. And that was late last year—the numbers will be even higher now.

Joe Alder: That’s a really important point. Most of those 48% are only using AI in the research phase, not necessarily clicking through as they would in traditional search engines.

James Lawrence: Exactly. Joe’s team has researched this: a lot of the “traffic down, revenue up” trends we see come from that change. Traditionally, people would Google, click to a blog, search again, click again, and so on. Now, a lot of that early and mid-funnel activity is happening inside ChatGPT. As a result, when traffic does come out of ChatGPT, it’s higher quality. An EMR study earlier this year found traffic from ChatGPT was 4.4x better at converting than general search traffic. For Rocket clients, Joe’s team found it was 5.1x better. Because users stay inside ChatGPT during research, by the time they click out, they’re further down the funnel and ready to convert.

Joe Alder: Yes, and if you think about how you personally use ChatGPT—it’s more specific than Google. Instead of just typing “computer mouse,” you’ll ask, “What’s the best computer mouse if you have wrist pain?” You do the research, compare products, read reviews inside ChatGPT, and only then go to Google or a site to buy. That’s why blog traffic has dipped, but it’s often more about Google’s zero-click searches and AI overviews than ChatGPT stealing traffic. Research shows most people are using search engines and ChatGPT together, not replacing one with the other.

James Lawrence: That’s a good point. Toward the end of today’s conversation, I want to talk about measurement—because so much of ChatGPT’s activity is locked down. But before that, let’s cover the practical side: how do you rank well in ChatGPT? To do that, you need at least a foundational understanding of how ChatGPT returns information.
Joe Alder: There are basically two main modes. First is database mode—ChatGPT uses its own training data. But it’s outdated—GPT-5’s cutoff is November 2024. That’s almost a year out of date now. Second is web search mode—it uses APIs, mainly Bing (Microsoft invested heavily in OpenAI). If Bing isn’t available, it falls back to Google. When it goes into web search, it does something called a query fan-out: breaking your prompt into multiple related queries. For example, if you ask about the best large SUV, it’ll generate queries like “What is a large SUV?” “What is the best large SUV?” “Best large SUV reviews,” and so on. It then pulls authoritative sites, awards, reviews, compares results, and personalizes the output based on your preferences, location, and previous prompts.

James Lawrence: Just to clarify for listeners: users don’t choose which mode it’s in. ChatGPT decides. If it already knows the answer—like “What’s the capital of Japan?”—it uses the database. If it doesn’t, it fans out queries and does a web search.

Joe Alder: Exactly. From our tests, it’s about 50/50 overall. But for products and services, it almost always uses web search. That’s where we have the most control, since it’s live information.

James Lawrence: Perfect. Let’s dive into how to optimize for this. We’ve broken it into four parts: technical do’s and don’ts—can ChatGPT access your content? Copywriting—how to write in a way ChatGPT likes. Types of content that perform well—based on Joe’s research. And authority—how ChatGPT decides which sources to trust. So Joe, starting with technical do’s and don’ts: what do listeners need to check?

Joe Alder: The first thing is making sure ChatGPT can actually read your content. A simple test: pick four key pages (homepage, blog, service page, product page) and ask ChatGPT, “Can you read this page and tell me the last line of the content?” If it can’t, it usually says so—because the page is behind a paywall, uses JavaScript to render, or is blocked by the host. Paywalls are self-explanatory. If your site relies on JavaScript to render text, ChatGPT can’t see it. And some hosting platforms block AI bots—Cloudflare began auto-blocking AI crawlers mid-2025. You can also ask ChatGPT, “Is this page dependent on JavaScript?” and it will tell you. Another issue is site speed. ChatGPT “bounces” if a page takes longer than ~3 seconds to load, just like a human user. Use Google PageSpeed Insights and check “Core Web Vitals” (look for LCP under 3 seconds). Also, set up Bing Webmaster Tools. Most SEOs only focus on Google, but since ChatGPT relies on Bing, you need visibility there. Finally, structured data. ChatGPT doesn’t yet use it, but others do, and Google still does. It future-proofs your site. For products, you can even submit URLs to OpenAI’s product discovery form.

James Lawrence: That’s a great practical checklist. We’ll include links to slides for listeners. Let’s shift to copywriting. This isn’t about using ChatGPT to write content, but about how to structure content so ChatGPT selects it. Joe, what have you found?

Joe Alder: Think about it like a user. First, ChatGPT sees your page title and meta description through the search API. If your page just says “Best Cars,” it’s less likely to click. If it says “Best Large SUVs 2025,” much better. We tested including the answer in the meta description—that increased pickup. For example: “Based on our tests, the best large SUV is the Nissan XYZ.” Second, ChatGPT only pulls the relevant snippet, not the whole article. So structure content with clear H1s and H2s, short snippable paragraphs (65–120 words), and a TL;DR summary at the top. Third, include tables, bullet points, and FAQs. These make it easier for ChatGPT to extract content—and they improve user experience too.

James Lawrence: So, when writing for ChatGPT and SEO, how do you balance the two?

Joe Alder: The good news: none of this harms SEO. In fact, much of it aligns. What’s different is we now emphasize snippable sections, bullet points, tables, FAQs, and freshness. ChatGPT prioritizes fresh content. Updating existing guides with “last updated” dates helps both ChatGPT and Google. And yes—ChatGPT loves original data, insights, and first-party research. Anyone can churn out generic “Ultimate Guides.” What stands out is unique content: studies, analysis, surveys, reports. That’s what gets cited.

James Lawrence: That’s a key point. The barrier to “average” content is gone. What performs best is unique research, unique perspective, and fresh insights. So Joe, what types of content perform best?

Joe Alder: ChatGPT favors content that helps it minimize sources. That means: listicles (e.g. “Top 10 Large SUVs”), comparisons and reviews, FAQs, and location-specific content. The more specific and structured, the better.

James Lawrence: Let’s talk about authority.

Joe Alder: Authority in ChatGPT isn’t about backlinks the same way as Google. It’s about brand mentions. When it does a query fan-out, it checks top sources. If you’re mentioned in industry publications, reviews, directories, awards, Reddit, or Wikipedia—you’ll rank better. Reviews are also critical: Google Reviews, Trustpilot, ProductReview. Interestingly, ChatGPT often uses affiliate or directory sites as sources. So being listed there can matter too.

James Lawrence: And the beauty is we can just ask ChatGPT: “Can you explain how you gathered this information? What sources did you use?” Then do the same for competitors and identify gaps.

Joe Alder: Exactly. It gives you the blueprint: which sites it pulled, which queries it ran. That’s gold for strategists.

James Lawrence: One more point: Wikipedia and Reddit are disproportionately influential. Our studies showed ~9% of responses cite Wikipedia. Not every business can have a Wikipedia page, but some can, and it’s worth pursuing. Also, awards matter enormously. For agencies, ChatGPT leans heavily on industry awards when recommending providers.

James Lawrence: Let’s zoom out. How do SEO and ChatGPT optimization overlap?

Joe Alder: There’s definitely overlap. About 60% of sites that rank in Google also appear in ChatGPT. But ChatGPT queries are longer and more specific (average prompt is 21 words), so comprehensive, structured, deeply covered content performs better. Like Wikipedia—it covers subtopics, cites sources, and is comprehensive. That’s the model. So yes, authority building still matters, but we target the sources ChatGPT actually uses.

James Lawrence: Now, let’s talk about measurement.

Joe Alder: This is the hardest part. Click-through rates from ChatGPT are very low—around 1% of queries. Most research stays inside the app. You can filter traffic in Google Analytics, but it’ll be small. There are AI tracking tools, but none are perfect. That’s why we’re building our own at Rocket—tracking specific queries, correlating results across LLMs and Google, and surfacing visibility. In the meantime: add GA filters for AI traffic, track your top 20 products/services with AI-focused queries, and ask customers directly how they found you—some will say “AI tools.” It’s imperfect, but it’s the best available right now.

James Lawrence: That’s excellent. It’s a different paradigm—more activity inside the platform, less top-of-funnel traffic to websites. Marketers need practical ways to measure visibility, even without a “Search Console for AI.” Joe, thanks for joining today. Any final thoughts?

Joe Alder: Just one: you might see advice online to create an “LLM.txt” file to direct ChatGPT. At the moment, none of the LLMs use it—so don’t waste time on it. Otherwise, that’s all from me.

James Lawrence: Perfect. Joe, thanks for coming back onto the podcast.

Joe Alder: Thanks, James.