Beyond the SERP: Winning the Visibility Layer and the Trust Stack

Good morning, Sydney. My name is James Norquay. Just kidding. James redirected me to this time slot, you could say he applied a penalty. He outranked me. I'll stop the SEO jokes now. We tried to make this happen for a long time, this is the third year James asked, and every year I said "next year." This time I made it happen.

I want to start with a quick experiment. Raise your hand if you would buy a $10 product on ChatGPT. Keep it up for $100. For $500, a few hands. For $1,000, really high, about two. Now, would you buy a $10 product on Google? Almost everybody. $100, $500, $1,000? Significantly more hands. Why is that?

The answer goes to the heart of what I'm talking about today. People, brands, and AI all change, but not at the same speed. Users are usually a bit ahead of brands, and the question is how you match that speed. I study AI closely, and I brought a lot of research to make sense of where we are now.

The rift

Twelve months ago something important happened, and I call it the rift: the separation of organic traffic from organic rankings. Up until March 2025 the two moved in parallel, a strong correlation where better ranks meant more traffic. Then it changed. What changed? AI Overviews. In March 2025 Google significantly ramped up the number of AI Overviews in the results, which separated ranks from traffic. So we need new ways to make sense of how much Google likes our website.

We underestimate how many SERPs actually show AI Overviews, because most studies look at a broad range of keywords. I looked at the three largest sites on the web by traffic, YouTube, Reddit, and Wikipedia, and checked how many of their non-branded keywords show AI Overviews. I get to about 50%. We are moving toward a world where the majority of keywords show AI Overviews. This problem is not going away. It is the new reality.

Traffic is no longer a leading indicator

There is an important Pew Research study of over 900 people showing that when Google shows AI Overviews, the click-through rate on organic results is cut in half. More interesting: only 1% of users click links inside AI Overviews. That is barely anything. This is not a click driver, it is a new experience.

It is not just AI Overviews. A leak of ChatGPT data showed that answers with over 600,000 impressions get a 0.69% click-through rate, which is abysmal and surprisingly close to that 1%. And AI mode has almost 100% zero-click share. These environments are not built to send traffic out, they answer the user's question right then and there.

On top of that, I looked at a range of B2B websites and found referral traffic from ChatGPT has been stagnating. The breaking point was August last year, when it dropped by over 40%. Until then there was a realistic chance AI chatbot referral traffic might become significant. I think that hope has died.

Here is one client example: organic search traffic declining 50% year over year, not something you want to brag about at an SEO conference. But conversions grew almost 20% year over year. Why? Purchase intent has not gone away. People still want to buy, and they still search. Traffic just is not a leading indicator anymore.

That brings me to three core challenges. AI dominates everything, mobile apps, web search, chatbots, you can even search the web in Excel. Click-through rates are going down and will not return to today's volumes. And traffic is detaching from conversions, though conversions are not going away, because people still need things.

What the user studies show

To make sense of this, I ran two user-behavior studies, one on AI Overviews and one on AI mode, with 35 and 70 participants. Picture a lab where people search at a computer while we record everything they see and say. We gave them tasks, had them think out loud, and used AI to codify the results finely.

First finding: when it comes to the final click or final answer, only 20% of people get it from AI Overviews. 80% get their final information from the classic search results. So even as traffic drops, classic results still matter, and there is still value in ranking highly.

Second: where do people go when they leave Google? About 18% of the time they click UGC platforms like Reddit and YouTube for their final answer. When Google shows AI Overviews, that jumps to 30%. From the transcripts, they are validating the AI answer, especially when the decision feels risky. They get the answer from AI, then double-check it with what they see as other users. So being present on those platforms matters.

The next finding changed my understanding of search completely. We looked at the evaluation process, what people ask themselves before they click or read. The number one filter is not relevance, it is trust. When people see a brand, they first ask "do I know and trust this brand?" and only then "is it relevant?" If you are not in your users' minds before they search, you have a big problem, you are working uphill.

In AI mode, engagement is high. Sessions averaged over a minute, about how long people brush their teeth. Most results are text, so people are reading, not scanning, and it matters to appear in those answers to influence the purchase journey. Of the few clicks that still happen, most go to shopping results when they appear, but 18%, one in five, open a new tab and navigate directly to the website instead of clicking in AI mode. Good luck attributing that.

The through-line: trust beats relevance. Trust was always the T in E-E-A-T, the most important factor, and this finally made sense to me. People genuinely engage with and read AI answers, but organic results still matter.

The playbook: the visibility layer

Based on these studies, here is my playbook for the new era, built on two frameworks, the visibility layer and the trust stack, with four areas each.

First, understand that successful content in 2026 is polymorphic, it lives in an open relationship with different platforms. In the pre-AI era content mostly had to satisfy Google. Now content has many masters: Google, ChatGPT, Perplexity, Reddit, YouTube, and it has to be viable for AI training, for RAG, for web retrieval, and now for agents.

Visibility comes down to four things. Additive content is the opposite of evergreen. There is little marginal utility in writing another "what is" or "how to" guide, because AI answers those. Additive content is new, original, thought leadership and research that AI cannot yet provide. One of the best sources of additive topics is transcripts. Classic keyword research sends everyone after the same volume-ranked terms. Instead, analyze transcripts, for example your sales calls, to find the topics and questions people actually care about. With one client we analyzed 40 to 50 call transcripts from prospects who became clients, mined them with AI (in that case a tool called Humata AI), and found topics we would otherwise have missed. That content outperformed our keyword-research content on both organic search and conversion rate, because the topics were top of mind for people in the market.

Freshness. LLMs value fresh content more than search engines do. An Ahrefs study found cited content tends to be younger, because LLMs have a training cutoff and rely on live retrieval for anything newer, where freshness is the number one signal. Ahrefs found a rough three-month threshold, though it varies by topic, some need content under a week old, some are fine over a year. Freshness matters across every intent. With one client, citation rate tripled after one simple change: adding a last-updated date to landing pages so LLMs can see how fresh the content is. This was across hundreds of thousands of pages, and yes, we made real content updates too, but signaling freshness alone had a profound impact.

Style. Google did not much care about style as long as the information was relevant. LLMs do. Definitive, very clear writing gets almost twice the citations, from an analysis of over 1 million AI answers. When you state clearly what you are talking about, without fluff or excess storytelling, you get cited more. Entities matter too: name the concepts and things specifically. "There are many good tools for this" gets cited less than "these are the top tools" followed by their names, because clear entities help LLMs understand what your content is about.

Reputation, or web reputation: how often and how you are talked about on the web. Studies from Ahrefs and Semrush show web mentions are one of, if not the, most important factors for how often you show up across AI chatbots and most use cases. I analyzed backlinks too, and they have a profound impact, but I think it is a bit of a red herring: LLMs seem to treat them like web mentions and care less whether it is a link or a mention. What matters most is quality, being mentioned in the right context on the right site beats sheer volume. To find where you need to be mentioned, use a traditional SEO process: take the keywords or prompts that matter, see which domains have the highest share, and prioritize them, breaking platforms like Reddit down by subreddit. Two clients do this well: Xero, whose XSBI research hub publishes original small-business economic data that gets consistently cited and linked, and Ramp, whose AI index tracks how much of card spend goes to AI, prime additive data for publishers and LLMs.

The playbook: the trust stack

Trust is the meta layer. Visibility is how you show up; trust is how you get people to go with your answer. First, evidence: back your claims, bring receipts, patents, certifications, or dense referencing of fresh, relevant data. With clients we look at the fact density of content.

Second, experts. People follow people. When credible, important people in a space speak well of your brand, it has a profound impact, the same validation principle as user-generated content. Influencers, creators, analysts, and journalists talking well about you leaves a lasting impression, so you are more likely to be believed, clicked, or bought from.

Third, validation. One of my clients is G2, and I looked at the relationship between software reviews and AI visibility. There is a small but robust, recurring correlation between reviews and brand visibility. Reviews, software or otherwise, Trustpilot, Google Maps, are a layer of consensus that lets LLMs check whether what a company claims matches what its customers say. Invest in reviews for AI visibility, they do not replace the basics, but they are a nice top-off.

Fourth, reputation, which is hard to quantify. Branded searches are not very indicative. Dan Petrovic's concept of selection rate is useful: ask an LLM the same prompt several times and look for the stable patterns. LLMs answer slightly differently each time, but if you run a prompt five times and some brands show up consistently, the model has high confidence they are critical to the answer. You can simply ask an LLM about your brand many times and see what it says consistently versus not, that is your reputation with LLMs.

From clicks to influence

To bring it home: the classic click funnel is breaking, and classic playbooks do not work as well anymore. We are moving from a world of clicks to a world of influence, where you influence users along the purchase journey. The two components are visibility (showing up) and trust. Visibility is additive, fresh content with a good reputation. Trust is evidence, experts, validation, and reputation.

We are moving from a single platform, Google, to many, Reddit, YouTube, Google, ChatGPT, Perplexity, Claude. From broad intent and short-head keywords to specific intent, since prompts are on average five times longer than search keywords. From evergreen to additive content. From clicks to influence, and from relevance to trust. Brands, people, and AI are all changing at different speeds, and the ones that win are the most trusted. Thank you very much.