Last month, we left you with a three-part prediction: June would give back some of May’s ChatGPT spike (the volatility tax), the engine mix would keep diversifying, and AI-referred conversions would hold. All three came true.
June was quieter than May, but quiet months are where the structural shifts show. We’re also halfway through 2026 now, which makes this a good moment to zoom out (more on that below).
The first force: the volatility tax came due. AI-referred sessions declined for just over half of our accounts, with moves ranging from -66% to +51%. May’s surge wasn’t earned traffic; it was ChatGPT flipping brand links back on, and a spike created by platform behavior settles once that behavior stabilizes. June is the settling.
The part worth internalizing: June settled well above April’s lows almost everywhere. This reads like a new baseline, not a collapse. One client just logged its fifth straight month of AI-referred conversion growth, and another booked 40% more meetings from AI traffic in a month where those sessions fell 11%.
The second force: AI search isn’t a ChatGPT story anymore. For most of the past year, “AI visibility” basically meant ChatGPT. In June, Google’s AI surfaces (AI Overviews, AI Mode, and Gemini) were the highest-visibility engines for most of the brands we track, and one client’s entire June visibility gain came from Gemini.
The flip side is click erosion in classic search. On one account, AI Overviews were present on 56% of the queries losing clicks, and queries lost clicks even when rankings improved. The 2026 pattern is rankings holding while clicks fall.
It cuts both ways, though. Another client watched a term jump from position 50 to #1 and land inside the AI Overview itself. The same surface that takes clicks away hands them to whoever gets cited.
(Google also ran its June spam update, an unusually short two-day rollout that was a non-event for nearly every account in our book. After four Google updates in four months, “verify before reacting” continues to be the right first move.)
The piece to listen to
On a recent Kitchen Side: The Hidden Drivers of AI Search, David, Alex, and I unpack what actually determines who gets cited in AI answers.
The idea we keep coming back to from that conversation: AI visibility behaves less like a channel and more like a brand recall survey. The answer engines are a mirror; they reflect whatever story exists about you across your site, reviews, communities, and press. You can’t optimize the mirror. You can only change what it reflects.
That reframe changes the work. It’s less about securing and defending a visibility score, and more about orchestrating the most accurate, compelling story about your product and brand everywhere AI forms its consensus.
What we saw across the portfolio in June
Conversions kept outrunning traffic. Organic conversions grew for 56% of the accounts with comparable figures, even though organic sessions fell for roughly six in ten. Efficiency over volume, again.
In-answer AI visibility contracted mildly on one platform we track (typically 2 to 6 points), but rankings barely moved. Seven of our clients hold the #1 AI-visibility spot in their category, and our two independent visibility platforms agree on standing where they overlap.
Refreshed and updated pages did the heavy lifting again: one account moved a page from #16 to #1, another took three separately updated pages to #1, and comparison pages jumped from #8 to #1 and #19 to #4 elsewhere in the book.
The takeaway: the AI referral line is a volatile input; the conversion rate of AI-referred visitors is the durable one. June is the third straight month that signal has held.
Since we’re at the halfway mark, the H1 scorecard: four Google updates, two major ChatGPT behavior swings, and through all of it the category leaders in our book held 60–80%+ visibility inside AI answers. One client climbed from 73% to 80% between January and June and held #1 every single month. One fintech client more than doubled its non-branded ranking keywords (2,108 to 4,357) by concentrating on BOFU. Traffic was the weather; visibility was the climate, and the climate rewarded brands that kept investing.
Three things that are working right now
Refreshing beats shipping net-new, fifth month running. Refresh pipelines prioritized by conversion potential, intent, and decay are recovering rankings faster than net-new content at the same budget. The honest caveat: a handful of pages updated in the same window slipped this month, which is why refreshes are a continuous engine, not a one-and-done project. The number we’re watching now is how well these positions hold.
Correcting what AI gets wrong about you. One client’s buyers kept seeing pricing misinformation in AI answers, so we shipped a fact-dense pricing page built to set the record straight; an AI Overview picked it up within a day. On another account, we caught an LLM citing an inaccurate third-party site and sending prospects to demos with wrong information, so brand-reclamation outreach is now a standing line item. AI engines will assemble your story from whatever’s available. Publish the authoritative version before someone else’s stale one wins.
Off-site validation. AI engines are shifting more citations to third-party sources; community threads, video, and review sites all gained citation share across accounts this month. We’re treating it as a recommendation gap, not a content gap, and the fix is off-site rather than another blog post. It’s also increasingly a cross-channel fight: we watched competitors bid up a client’s own brand terms while organic clicks thinned.
One more peek behind the curtain: we’ve formalized an internal experimentation program to turn AI-search folklore into controlled tests. Two are live now, one testing whether proprietary research data in the top third of a post lifts LLM citation rates (treatment vs. control, measured at four weeks), and one testing whether hyper-specific pages get retrieved more reliably than broad ones. I’ll share what holds up.
Our prediction to revisit next month: the share of AI visibility and AI-referred traffic coming from non-ChatGPT engines keeps climbing, and AI-referred conversion rates hold even if session volume wobbles again. Watch conversions over sessions.
The whole point of an agency view across 30+ B2B accounts is that we see the patterns before any one brand would. Reach out for a strategy call, and we’ll dig in.
Want more insights like this? Subscribe to Field Notes
