For about fifteen years, demand gen has run on a quiet assumption. Someone has a problem, they search, they click, they land on something you own, and from there you have a shot. Your whole top-of-funnel machine, the SEO, the content calendar, the gated assets, the form, the lead score, the routing, all of it sits on top of that one move. The click.
That assumption is coming apart, and the data this quarter is hard to argue with.
SparkToro’s 2026 study found that 68% of U.S. Google searches now end without a click. Two years ago that number was 60%. Search Engine Land’s breakdown of the same data calls it the steepest two-year jump since anyone started measuring it, driven mostly by AI Overviews, which cut click-through rates by nearly 60% when they show up.
At the same time, new G2 research found that 51% of B2B software buyers now start their research inside an AI chatbot. In April of last year that was 29%. G2’s own writeup puts it plainly: 71% of buyers now lean on AI chatbots for software research, and a majority say the research feels more productive than a regular search.
Most of the coverage I’ve seen on this lands in SEO circles. Practitioners arguing about schema markup and whether AEO is a real acronym. That conversation matters, sure, but it skips the people who are about to feel this first and explain it worst: demand gen operators who built a pipeline number on top of organic inbound.
👋 Hi, it’s Kaylee Edmondson and welcome to Looped In, my newsletter exploring demand gen and growth frameworks in B2B SaaS. Subscribe to join 2k+ readers who get Looped In delivered to their inbox every Sunday. If this was forwarded to you by a friend, firstly, thank them, and then subscribe below. See you around!
Why this hits demand gen harder than it hits SEO
An SEO person loses rankings and traffic. That’s painful, but it’s legible. You can see it in the dashboard, you can point at it, you can build a recovery plan.
A demand gen operator loses something quieter. You lose the first touch, and you lose it invisibly.
Walk the path. A buyer at one of your target accounts has a problem. A year ago they Googled it, clicked your blog post, maybe grabbed a guide, and entered your world as a trackable contact. Now they open ChatGPT, Perplexity, or Claude, describe the problem in a sentence, and get an answer that may or may not include your name. They read three vendor comparisons that were synthesized for them. They form a shortlist. None of that touches a property you own. None of it fires a pixel.
By the time they do hit your site, they’re typing your brand name into the search bar or going direct. So your attribution model, which is almost always weighted toward last touch or first trackable touch, credits branded search or direct. The research that actually moved them, the work that put you on the list, happened in a room you weren’t allowed into (at least not yet anyway).
Onto more gruesome realizations…When that buyer converts, your dashboard will say direct or branded. It will not say “the comparison content we seeded eight months ago that an AI model later quoted.” So the channel doing the heavy lifting at the top will look like it’s contributing nothing, and the channel catching the demand at the bottom will look like a hero.
I watched a version of this play out with a client this spring. Non-branded organic clicks down quarter over quarter, branded search up, direct up, and a sales team reporting that more discovery calls were opening with “yeah, I’d already looked into you.” The funnel looked like it was shrinking at the top and getting stronger at the bottom. But we know it wasn’t. The top had moved somewhere the analytics couldn’t follow.
If you manage to a pipeline number, this is the trap. The motion that builds demand is going darker, faster, in your reporting at the exact moment you need to defend its budget.
The dark funnel was already a problem. Now it’s basically the default.
We’ve talked about dark social and the dark funnel for years. The podcast someone heard, the Slack community recommendation, the LinkedIn post that did the convincing, all the influence that happens off-platform and shows up as “direct” later. Most teams treated it as a known gap. A rounding error you acknowledged and moved past.
AI research takes that gap and moves it to the center of the buying process. The single most important stage, the part where a buyer decides who’s even in the running, is now the least trackable stage. The dark funnel stopped being the edges. It’s the middle now.
I’m pretty sure this changes our jobs, y’all. For a long time the demand gen mandate was capture: get in front of in-market buyers and pull them into a trackable path. Capture still matters. But capture assumes the buyer ends up somewhere you can measure. When the research happens inside a model, you can’t capture what you were never present for. You have to be in the answer itself.
The four moves to make before Q3
None of this is a reason to panic, and none of it means SEO is dead. LinkedIn posts are always trying to claim something else in our toolkit is dead…this isn’t that. It means the inbound model needs a second engine bolted onto it, and the demand gen team should own that engine rather than tossing it over the fence to SEO. Or shoot, maybe you are SEO, but either way.
Here’s the play I’m running with clients right now.
1. Audit whether you exist in the answers. Before you change anything, find out where you stand. Take your fifteen highest-intent buying questions, the ones a prospect asks right before a shortlist forms, and run them through ChatGPT, Perplexity, Gemini, and Google’s AI Overviews. Write down who gets named, who gets cited, and where you don’t appear at all. This is your new rank tracking. Do it as a recurring monthly check because the answers will shift.
2. Write content a model can quote, not just a human can skim. Answer engines pull from content that states a claim cleanly, backs it with a number, and attributes it to a source. That means clear definitions near the top of the page, real data with citations, direct answers to direct questions, and structure a machine can parse. The fluffy, keyword-stuffed SEO post that ranked in 2021 is the worst possible input for an AI answer. Your best original research and your sharpest point-of-view content are the best. This is a reason to spend on primary data and strong opinions, not generic volume.
3. Re-instrument attribution around self-reported and branded signals. Stop pretending your multi-touch model sees the whole journey. It doesn’t, and it’s about to see less. Add a “how did you first hear about us” field to your demo and contact forms, and report on it. Track branded search volume and direct traffic as demand indicators because a rise in both is often the visible shadow of invisible upstream work. When you present pipeline to leadership, name the dark funnel explicitly so a branded-search spike gets read as demand you created.
4. Build presence where the research actually happens. If buyers are forming opinions in AI chats, communities, podcasts, and peer conversations, that’s where your demand creation budget belongs. Review sites and third-party comparison content feed the models, so your G2 and Capterra presence is now an AEO input, not just a sales-stage proof point. Founder and executive presence on LinkedIn shapes the conversations that models later summarize. Original research gets cited and quoted in ways a product page never will. This is signal-driven demand gen in the truest sense: you’re planting the inputs that the machines, and the humans, pull from later.
The takeaway I’d write
The buyer’s research moved into rooms you can’t measure and can’t gate. What you can do is be present in those rooms, feed the systems that summarize you, and fix your reporting so the work doesn’t look like it’s failing when it’s actually doing its job.
If you do one thing this week, do the audit in move one. It’ll take you an afternoon, and it will tell you more about your real top-of-funnel position than your rankings report has.
Here’s to growth,
Kaylee ✌
the “yeah, i’d already looked into you” part is audible on our end too. i listen to our sales call recordings to build icp and messaging, and the first five minutes have quietly become an attribution report. people volunteer where they heard about us before anyone asks. the form field is a good move, and the recordings are the same data already sitting there, in their own words