Still Posting AI Agent Content?

Your team has a content calendar with "AI agent" briefs sitting in it for the next four weeks. The creators are booked, the hooks are half-written, and someone on the account side is waiting for sign-off before outreach goes out.

Here's the problem. That calendar was built off a narrative that peaked back in March. Nobody told you because nothing dramatic happened. No crash, no announcement, no reason to notice. The conversation just quietly kept fracturing while the brief stayed the same (we've all shipped a campaign into a narrative that had already moved on, and it is a genuinely expensive way to learn this lesson).

This isn't an argument to drop AI content. It's a map of where it actually went, so the next brief lands somewhere the smart accounts still are, not somewhere they've already left behind.

Stage 1: The Peak

AI agent conversation didn't peak once. It peaked twice, a month apart, and the two peaks measure different things. Absolute attention (signal) peaked the week of 2 February. Relative dominance (mindshare, the narrative's share of all crypto conversation) peaked a month later, the week of 2 March.

That gap isn't noise. Total market-wide conversation contracted sharply between those two weeks, so AI agent's share kept climbing even as its own volume was already easing off. Anyone reading mindshare alone in early March would have called it the top of the market. It was already past that.

By the week of 15 June, signal had fallen 64.5% from its February peak, and mindshare had fallen 54.4% from its March peak. Both lines point the same direction now. There's no ambiguity left in this one.

AI agent weekly signal vs mindshare, December 2025 to June 2026

Stage 2: The Fracture

A peaked narrative doesn't die, it splits. We looked at eight sub-narratives that branched out of AI agent conversation over the following months: DePIN, ZK, Robotics, Oracles, Reputation Layer, Bittensor, and DeFAI.

Most of them are shrinking in line with the parent narrative. DeFAI is the cautionary one. It crashed from real early signal down to a residual trickle, and the accounts still posting about it skew heavily toward low-credibility, high-volume behaviour rather than genuine interest. That's what a sub-narrative looks like when it doesn't survive the split.

DePIN is the exception worth noting, not because it's loud, but because it's quietly gaining relative ground while overall volume falls. Its share of total conversation is up over a third since January, even though its raw signal is lower than it was then. That's a narrative holding its footing while everything around it deflates, which is a different (and more durable) kind of signal than a volume spike.

Stage 3: Where the Smart Kols Actually Went

Here's the part that should change how you brief this quarter. Total AI agent posting volume dropped 64% from peak to now. The rate of credible accounts (Smart KOLs) participating in that conversation dropped by roughly a fifth in the same window. Credible voices didn't run for the exit. They just posted less often.

That distinction matters because it tells you the audience quality hasn't collapsed, only the noise has thinned out. And when we looked at where those same credible accounts are spending their other posts, a pattern showed up that a volume chart alone would never surface.

Solana's and Base's ecosystem accounts are now splitting their AI-related posting almost evenly between AI agent content and DePIN content, not as an occasional crossover but as a consistent, near 1:1 pattern. That's an infrastructure account choosing to spend attention on physical infrastructure narratives alongside agent narratives, which is a fairly deliberate signal from accounts whose job is to know where the ecosystem is heading. One VC-background account we tracked has gone further still, posting more than twice as often about AI broadly as about AI agents specifically, which reads as a credible individual voice actively despecialising away from the narrower topic.

Not everyone is rotating, and that's useful too. At least one high-signal a16z-affiliated account posted about AI agents exclusively in the same window, with zero crossover into any of the infrastructure sub-narratives. The rotation isn't universal. It's selective, and it's concentrated in ecosystem and infrastructure-adjacent voices rather than agent-native ones.

Of every sub-narrative we measured, one produces more signal per post than AI agent content itself, and it isn't close. Bittensor's Smart KOL participation rate is now the highest of any sub-narrative we tracked, well above AI agent's own rate. Worth being honest about the caveat here: Bittensor's total volume is small enough that a single high-signal post can move the number, so read this as "the accounts talking about it skew credible," not "this has broad coverage yet." It's an early, narrow signal. It's still the clearest one we found.

Stage 4: This Week's Tell

AI remains the single strongest technology narrative in crypto conversation this week, ahead of every other theme by a wide margin. That part of your brief was always correct. What's changed is which half of "AI" is doing the work.

For the first time across the full period we measured, infrastructure-side conversation (DePIN, Bittensor, ZK, Oracles, Reputation Layer, Robotics combined) matched or exceeded consumer-facing agent conversation, and it's now happened twice in the last three weeks. Consumer agent content still leads overall, but the gap has roughly halved since late May. That's a trend line, not a blip.

AI infrastructure signal vs consumer agent signal, eight-week window

The clearest example of quality moving ahead of volume this week is Akash Network, tagged across both DePIN and AI. Its signal rose several times over its recent average while its post count grew only modestly, meaning the increase came from engagement quality rather than a flood of new posts. That's exactly the shape a genuinely credible narrative shift takes before it becomes obvious enough for everyone to chase it at once.

What You End Up With

You don't need to pull AI out of the plan. You need to stop treating "AI agent" as the whole plan. The credible accounts your campaigns are trying to reach are already spending a growing share of their attention on the infrastructure conversation sitting right next to it, and they made that move without announcing it anywhere you'd have caught by watching engagement counts alone.

Next brief, split the allocation instead of defaulting to agent-only creators. Your Q3 calendar can thank you later.

None of this came from watching timelines scroll by. It came from tracking mindshare, signal, and Smart KOL participation as separate, comparable numbers across every sub-narrative at once, week over week, and being able to name the specific accounts behind a rotation instead of guessing at it from vibes. That's the difference between noticing a shift after everyone else has already priced it in, and catching it while it's still quiet enough to act on.

If your team is briefing creators off what's loud on your own feed, you're working with the same blind spot every other agency has right now. Cookie Pro is how we pulled every figure in this piece, and it's built for exactly this kind of question, before it becomes obvious to everyone else at once.

Try it in Cookie Pro → cookie.fun