Visitors from ChatGPT are worth 4.4 times as much as those from Google — AEO is the way to attract them

Three Acronyms for a Single Battle

The marketing industry has never missed an opportunity to coin an acronym. Right now, it’s circulating three for the same concept: AEO (Answer Engine Optimization), GEO (Generative Engine Optimization), and LLMO (Large Language Model Optimization). Entire agencies have repositioned themselves around one or the other, and Profound—one of the leading AI visibility tracking tools—eventually published a post titled “AEO vs. GEO: Why They’re the Same Thing.” That’s exactly how I see it. Regardless of the label, the goal is the same: when an answer engine—ChatGPT, Perplexity, Google’s AI Overviews, Claude—answers a question from your potential customer, you want to be the source they read, cite, and recommend.

I’ll use the term AEO in this article, because “answer engine” accurately describes the new interface: the user asks a question, and the machine responds directly. Whereas traditional SEO optimizes for a list of links and hopes for a click, AEO optimizes to be read and cited by the machine that generates the response. The distinction isn’t just theoretical—it changes what you measure, and some of the things you do. But not as much as the sellers of “premium GEO” packages would have you believe, and that’s where the data gets interesting.

The dual trend: fewer clicks, but clicks that are worth more

Two studies published a few months apart together tell the whole story. The first comes from the Pew Research Center, in July 2025: among 68,879 actual Google searches by 900 U.S. participants, the presence of an AI summary caused the click-through rate on traditional results to drop from 15% to 8%. Worse still: the sources cited in the summary only received a click in 1% of visits, and 26% of sessions ended right on the results page, compared to 16% without a summary. In other words, when AI provides an answer, most people read it and leave.

The second finding comes from Semrush, which tracked traffic referred by LLMs to thousands of websites: the average visitor coming from ChatGPT or Perplexity converts 4.4 times better than a typical organic visitor, and this traffic skyrocketed by 527% between early 2024 and mid-2025. The logic is crystal clear when you think about it: the person who clicks anyway after reading an AI response has already been qualified, compared, and is halfway convinced. They arrive on your site at the end of their decision-making process, not at the beginning.

Put the two side by side, and you get the true AEO thesis: click volume goes down, while unit value goes up. The goal is no longer to attract a thousand curious visitors, but to be mentioned at the moment the machine makes its recommendation—and to convert the handful of highly qualified visitors who follow. For an SME in Sherbrooke that sells services, I’d rather have ten visitors with a 14% conversion rate than three hundred with a 2% conversion rate.

What Princeton Actually Measured

The problem with AEO in 2026 is that 90% of the advice circulating has never been tested. The only serious scientific basis remains the GEO study by Aggarwal and his colleagues (Princeton, Georgia Tech, IIT Delhi), presented at the KDD 2024 conference. They took 10,000 queries, tested nine content-modification strategies, and measured the effect on visibility in the generated responses. Overall result: up to 40% more visibility in generative AI responses for the best tactics.

And the winners aren’t the ones the industry touts. The three most effective strategies identified were: adding precise statistics, citing sources, and incorporating expert quotes. The most counterintuitive finding concerns small websites: for pages that rank poorly in traditional SEO, adding citations from sources caused visibility to jump by up to 115%. The answer engine does not replicate Google’s ranking hierarchy—it looks for content that appears reliable, verifiable, and well-supported. This is an unexpected opportunity for sites that don’t have twenty years of domain authority.

Conversely, keyword stuffing—the most instinctive SEO tactic—actually reduced visibility in their tests. And the researchers note that effectiveness varies by domain: what works for a historical query doesn’t necessarily work for a shopping query. There’s no one-size-fits-all formula, but there’s a clear direction: the system cites what looks like a source, not what looks like an ad.

If you follow this blog, you’ll notice that this is exactly the approach I’m taking here: every figure is dated and attributed, and every bold claim has a named source. This isn’t just a stylistic quirk—it’s literally the best-measured AEO tactic.

The Practical Approach on a Single Page

Let’s zoom in to the page level. The structure that works for a response engine is based on three principles, and none of them requires you to completely rewrite your site.

Answer first, elaborate later. Under each section heading—formulated as a genuine question or statement—place the complete answer in the first two or three sentences—self-contained, quotable, and extractable as-is. Elaboration, nuance, and examples come afterward. LLMs break pages down into blocks; a block containing a complete answer is a quotable block. A block that begins with “First, you need to understand the context…” is not.

Be a consistent entity. Answer engines reason in terms of entities: who is this person, this company, this product. The same name, city, and specialties everywhere—website, LinkedIn, Google Business, directories. On my website, the JSON-LD ProfessionalService clearly states who I am, where I work, and what services I offer, and this consistency is reflected word for word in my external profiles. When an AI puts together its response to “web designer in Sherbrooke,” all the pieces of the puzzle fit together.

Date and sign your content. An anonymous, undated article is unverifiable, and therefore not very citable. An identifiable author, visible publication and update dates, and specific years mentioned in the text. This is also what protects you from content recycling: search engines prefer data from 2026 over data from 2023.

Schema Markup: Useful, but Not Magic

Let’s talk about schema markup, because it’s the most widely promoted—and most misunderstood—AEO tip. Yes, JSON-LD markup (Article, Person, Organization, FAQPage) helps machines understand the structure of your content, and it costs almost nothing to add. No, it’s not a magic button for getting cited. A quick reminder that the industry has a short memory: Google removed FAQ rich results for nearly all websites as of August 2023—reserving them for government and healthcare sites. The visual badge has been gone for three years; the remaining value of markup is semantic, not cosmetic.

As for the domains that AI systems cite most often, analyses by Semrush and Profound in 2025 consistently place Reddit and Wikipedia at the top of the list. Practical takeaway for a local business: your presence in third-party conversations—forums, industry directories, guest posts, reviews—carries as much weight in the models’ memory as your own website. AEO doesn’t stop at your pages.

What I Wouldn’t Do

The fear-mongering market is in full swing, so here’s my blacklist. I wouldn’t pay for an “AEO audit” marketed as some esoteric discipline: about 80% of the recommendations will be basic technical SEO (speed, clean HTML, structured data, quality content) repackaged to sound trendy. I wouldn’t rewrite an entire site in an artificial Q&A format: the “answer-first” structure applies to sections, not to a disguised giant FAQ that human visitors hate. And I wouldn’t rely on “magic” files: I documented in my article on llms.txt that 97% of these files receive no requests—it’s a zero-cost best practice, not a strategy.

The sorting rule is simple: if an AEO tip isn’t in the Princeton study, isn’t an established SEO best practice, and isn’t a matter of technical extractability, ask what data supports it. The answer is usually an awkward silence.

Where to Start This Week

Three steps, in order of return on investment. One: take your five most important pages and infuse them with sourced, dated statistics and named references—a tactic proven to boost results by 40%, implementable in a single afternoon. Two: Restructure each section so that the answer appears within the first three sentences below the headline. Three: Check for extractability—is your critical content in the source HTML, is your entity consistent throughout, and does your JSON-LD tell the truth?

And measure. AI sources appear in your analytics (chatgpt.com, perplexity.ai), AI crawlers show up in your server logs, and tools like Ahrefs’ Brand Radar now track brand mentions in generated responses—that’s what I use. SEO taught you to monitor rankings; AEO teaches you to monitor citations. Traffic will be lower. Each visitor will matter more. And honestly, for those of us who have always written to inform rather than to rank, this shift feels less like a threat and more like long-overdue justice.

Sources

• GEO: Generative Engine Optimization — Aggarwal et al., KDD 2024 (arXiv
  .09735)
• Google users are less likely to click on links when an AI summary appears — Pew Research Center, July 22, 2025
• We Studied the Impact of AI Search on SEO Traffic — Semrush
• Average LLM visitor worth 4.4x organic search visitors — MarTech
• AEO vs. GEO: Why They’re the Same Thing — Profound
• Search volumes: Ahrefs Keywords Explorer, July 2026