PASSIONFRUIT PULSE
You're reading Passionfruit Pulse, where we squeeze the week's biggest search shifts into one short email. Google changed something. AI search moved the goalposts. Again. We'll tell you what happened, why it matters, and the one thing you can actually do about it this week. Three minutes, coffee in hand, let's go.
Blog of the week
Google quietly turned Search into a task layer
While the GEO industry was arguing about fan-out, Google shipped three features that change the job entirely: hotel price tracking on individual properties, agentic calling that lets AI Mode call local stores to check stock, and a Canvas tool that builds full travel itineraries inside Search.
Sundar Pichai has called this a "task-based" vision. The features make it real.
The downstream consequence for SEO is that being ranked is no longer the same as being useful. If your business cannot be acted on by an agent if the inventory feed is stale, if the booking page takes five clicks, if the LocalBusiness schema is half-implemented the agent skips you regardless of position.
Three shifts that matter most:
Schema stops being a ranking signal. It becomes the interface: Sloppy markup was acceptable when the goal was a rich snippet. It's not acceptable when an agent is choosing which restaurant to call or which hotel to alert a user about. Partial schema is now a liability, not a missed opportunity.
Data freshness becomes a visibility problem, not a conversion one: Inventory that's wrong when the agent calls doesn't just lose the sale it teaches Google your data is unreliable. SEO is now downstream of operations and inventory pipelines in ways most teams aren't structured for.
The commodity content era is ending faster than expected: Google's own examples are sharper than anything the industry has produced. Compare "7 Tips for First-Time Homebuyers" (commodity, will be summarized into a paragraph nobody clicks) with "Why We Waived the Inspection and Saved $15k: A Look Inside the Sewer Line" (non-commodity, gets cited because no one else could have written it). The first is dying. The second is the new bar.
Research you should read
The fan-out study everyone in GEO needs to read this week
Our literature review of 30+ publications isn't just one more study. It's the first one with enough scale and methodological rigor to actually settle an argument the industry has been having since GPT-5 shipped.
The five things worth internalizing:
Retrieval rank is still the single biggest predictor: 58% citation rate at position 1. 14% at position 10. The thing AI search vendors keep telling you doesn't matter anymore is, by the data, the thing that matters most.
Heading-query cosine similarity beats fan-out coverage by a wide margin: 41% citation rate at 0.90+ match. 30% below 0.50. Headings that directly answer the user's likely question outperform headings designed to cover sub-topics.
Fan-out queries themselves are statistical noise: Only 27% of fan-out keywords stay consistent across repeated searches. 66% appear once and never again. Tracking them as KPIs is tracking the weather, not the climate.
The "ultimate guide" strategy actively underperforms: Focused content (500–2,000 words) beats comprehensive content (5,000+) at every measured citation rate. Wikipedia is the only exception, and Wikipedia is not comparable. You don't have 4,383 words of edited prose, 31 lists, and 6.6 tables per page across 6.8 million entries.
The "proof" the industry has been citing is a four-article experiment: Semrush's widely shared fan-out case study went from 2 to 5 citations across four updated posts in one month. No control group. The lead author's own honest summary: "You're playing on unstable ground." That's the strongest commercial evidence in support of the entire paradigm.
INDUSTRY SIGNAL
Shortcuts in GEO are dying for the same reason they died in SEO
Ann Smarty's piece this week is the cold-water moment the GEO discourse needed.
Her point is pattern-recognition, not contrarianism. Listicles worked for LLMs because the format made entity extraction easy. Brands and publications piled in. Sponsored mentions everywhere. Then Google noticed.
ClickUp went heavy on the listicle play and lost an estimated 7M in organic traffic over the last 6 months.
But the harder data point is underneath the listicle conversation: Ahrefs reports LLM traffic is currently 0.26% of total inbound traffic for digital brands. Smarty's own consumer research points the same direction — AI adoption inside the tech bubble is running well ahead of how consumers actually behave.
This is the uncomfortable subtext under every GEO conversation right now: the channel everyone is building tools and agencies and conference talks around is, in raw traffic terms, still a rounding error. Not for long, probably. But long enough that trading fundamental SEO visibility for an LLM optimization shortcut is almost always the wrong trade.
The teams that will own AI search in 2027 are not the ones running the cleverest tactics in 2026. They're the ones building real category presence in the surfaces LLMs already pull from Google, Reddit, YouTube, Maps, owned content with proprietary data and being known for what they do.
If your GEO strategy looks like a clever trick, it has a shelf life. If it looks like brand building, it compounds.
TOOLS
Product Feature of the Week: Top Queries Where You Are Not Mentioned
The standard AI visibility question is "Where am I being cited?" It's the wrong question. The bigger lever is the inverse: "Where is my entire category being cited and I'm completely absent?"
That's the gap our team uses to prioritize content briefs, and it's now a view inside the product. Top Queries Where You Are Not Mentioned surfaces, for every query in your tracked set where your brand doesn't appear, the citation count, the brands AI is recommending instead, and which platforms the query is firing on.
A real example from a recent audit: a supplements brand we work with was completely absent from "healthy hydration powders" — a query firing 2,127 times across competitor mentions, with 8 different brands eating that consideration set. They had no content addressing it. They had a product that fit it perfectly. The brief wrote itself.
That's the actual unlock: not measuring what you're winning but measuring what your category is winning that you're not. Most teams have never seen the second list. It changes what you build next.
Our Passionfruit Playbook
One question to audit: Which 5 queries in my category am I completely absent from, and who's getting cited instead?
One quick win: Pick your single highest-traffic page. Rewrite the first 50 words as a direct, definitive answer to the query it ranks for. Front-loaded answers ("the ski ramp") account for 44.2% of ChatGPT citations. This is a one-hour fix with one of the highest ROIs in the dataset.
One thing to stop doing: Reporting fan-out coverage as a KPI. Replace it with retrieval rank for your top 25 pages and heading-query cosine similarity. The first metric is theater. The second two are the actual mechanics.
Creator prompt: What can I publish this week that contains at least one specific number, one named example, or one original test that no competitor in my category could have written?
Until next week,
Passionfruit Team