TL;DR: A new study found AI Overview citations get clicked ~1% of the time. Most SEOs will panic. The smart ones will pivot. Citation isn't a traffic play anymore — it's an influence play. Generative AI has fundamentally changed how buyers discover brands. The brands that win in 2027 won't be the ones chasing clicks. They'll be the ones entering the buyer's consideration set before the website visit ever happens. I call this the 1% Click Problem — and it rewrites the entire GEO playbook.
James here, CEO of Mercury Technology Solutions.
From my office in Wanchai, looking across the HKCEC— I keep thinking about how search is undergoing its own phase transition. Not gradual. Sudden. Like water becoming steam.
You probably saw the study. Researchers tracked one month of browsing behavior from 900 U.S. adults. When Google served an AI Overview:
- Citation links were clicked in ~1% of visits
- Users clicked fewer organic results overall
- They were more likely to end the browsing session entirely
One percent.
The SEO industry's first instinct will be to declare AI Overviews "bad for SEO." I've already seen the LinkedIn posts. The Twitter threads. The agency newsletters with subject lines like "Is This the End of Organic Search?"
They're asking the wrong question.
The right question is: What happens when someone sees your brand but never clicks?
Because that — the invisible impression, the unlogged exposure, the zero-click discovery — is becoming the new battleground. And most companies have no idea how to fight there.
The Traffic Trap
For twenty years, SEO trained us to think in a straight line:
Impression → Click → Conversion
Every tool, every dashboard, every agency report optimized for that middle arrow. Click-through rate. Cost per click. Landing page conversion. We built entire industries around measuring and maximizing that click.
AI search breaks that chain. Not weakens it — breaks it.
The new journey looks more like this:
Search → AI Answer → Brand Exposure → Consideration → Later Search → Conversion
Notice what's missing? The click from the AI Overview. The middle step — the one we spent two decades optimizing — disappeared from analytics. But it didn't disappear from the buyer's mind.
This is what I call The 1% Click Problem: the gap between being seen and being clicked isn't a failure of GEO. It's a fundamental shift in how discovery works. And if you're still measuring success by traffic from AI citations, you're optimizing for a metric that no longer correlates with business outcomes.
Think of it like this. In the Battle of Amritsar Starzone, Yang Wen-li didn't need to destroy every Alliance battleship to win. He just needed to make their commanders believe they couldn't win. The battle was over before the first shot was fired.
The same thing is happening in search. The sale is increasingly being decided in the AI Overview — before the user ever reaches your website.
Citation vs. Mention: The Hierarchy of Influence
Here's where most GEO strategies fall apart. They optimize for citation frequency — getting their URLs referenced as sources — without asking a more important question:
Did the AI actually recommend your brand in the answer?
Your article could be cited three times as evidence for Google's response while your competitor gets named as the solution. That's a catastrophic trade. Your URL isn't what buyers remember. Your brand is.
I see this constantly in our audits. A company will celebrate: "We're cited in 40% of AI responses for our target keywords!" Then I ask: "How many of those citations mention your brand name in the answer text?"
Crickets.
Citation is becoming a vanity metric. It's the new backlink count — impressive in a report, meaningless in isolation. What matters is whether the AI's generated answer positions you as a relevant solution, a credible option, a brand worth considering.
At Mercury, we've started measuring what we call Narrative Share of Voice — not just whether you're mentioned, but how you're mentioned.
Compare:
"Acme is project management software." (Weak. Commodity positioning.)
Versus:
"Acme is particularly suited to distributed engineering teams managing complex microservices architectures." (Powerful. Specific. Memorable.)
Same mention count. Completely different commercial value. One makes you a checkbox. The other makes you a shortlist.
The brands that understand this distinction will dominate 2027. The ones that don't will keep chasing citation counts while their competitors capture the buyer's consideration set.
The New Measurement Model: C → M → N → S → P (The LLM SEO Funnel)
So if citation isn't the goal, what is?
At Mercury, we use a five-layer framework for measuring GEO as an influence channel rather than a traffic channel. Think of it as a funnel, but one that measures mental state rather than page views.
C — Citation
Did AI use your content as evidence? Track frequency, cited URLs, query context, citation position, competitors cited beside you.
Useful. Necessary. But insufficient. Because ~99% of AIO visits don't produce a citation click.
M — Mention
Were you actually included in the generated answer? Not as a source — as a solution. This is where brand recall begins.
For commercial queries, I'd rather have: Brand mentioned + no citation than Citation + competitor mentioned.
Your URL isn't what buyers remember. Your brand is.
N — Narrative
What did the AI say about you? This is where most GEO audits stop measuring and start guessing.
We track Narrative Share of Voice — the semantic positioning of your brand across AI responses. Are you described as "project management software" or "the platform remote teams actually want to use"? Are you "an option" or "the recommended solution for X use case"?
The same mention count with different narrative framing produces wildly different commercial outcomes. This is why generic GEO audits miss the point. They count citations. They don't measure meaning.
S — Search
Here's the signal I'd obsess over next.
After AI visibility increases, monitor:
- Branded search volume
- "[Your brand] reviews"
- "[Your brand] pricing"
- "[Your brand] vs [competitor]"
- Direct traffic to your homepage
- Demo request attribution
Because the new journey might be:
Google search → AI Overview → Brand discovered → NO CLICK → Brand searched later → Demo requested
Traditional attribution completely misses the first touch. The AI Overview was the catalyst. But because it produced no click, it shows up nowhere in your funnel analytics. It's invisible to every tool that measures what happened on your website.
This is why GEO teams need to look at lift metrics — the delta in branded search, direct traffic, and downstream conversion that correlates with AI visibility campaigns. Not causal in the old sense. Correlational in the new one.
P — Pipeline
This is where GEO finally becomes a business channel instead of a marketing experiment.
Add one field to your demo request form: "How did you first hear about us?"
Include:
- ChatGPT
- Google AI / AI Overview
- Perplexity
- Claude
- Traditional Google search
- Other
Then train your sales team to record phrases like:
- "ChatGPT recommended you"
- "I saw you in Google's AI answer"
- "I kept seeing your company recommended for [use case]"
Now you can connect AI visibility to revenue. Not through click attribution — which is broken for this channel — but through self-reported discovery. It's messier. It's also more accurate than pretending the old model still works.
The Strategic Implication: From Answer Assets to Consideration Assets
This reframe changes what you build.
For twenty years, SEO meant building Answer Assets — content that answered specific questions well enough to earn the click. The goal was to intercept search intent and convert it into traffic.
The new game is building Consideration Assets — content so authoritative, so specific, so deeply aligned with how AI systems construct answers that your brand becomes the default inclusion in generated responses for your category.
This isn't about keyword density. It's about entity clarity — the foundation of modern SEO and GEO strategy for enterprise growth.
Artificial intelligence systems don't "read" your content the way humans do. They extract entities, relationships, and confidence scores. They build internal graphs of "what Acme is" and "what Acme is good for." The clearer that graph, the more likely you are to be mentioned — not just cited — when relevant queries arise.
This is why our GEO methodology has two dimensions:
- Method A (Readiness): Can the AI parse, represent, and cite your content? (Our current score for Mercury: 91/100)
- Method B (Selection): Does the AI's generated answer include your brand as a recommended solution? (Our current score: 72/100)
The gap between them — 19 points — is the distance between being findable and being chosen. Between claw-ready and claw-selected.
Most GEO tools only measure Method A. They tell you whether AI can access your content. They don't tell you whether AI prefers your content when constructing answers.
That's the real game now. Not access. Preference.
What to Do Differently in 2027
If you're building a GEO strategy for next year, stop optimizing for the old funnel. Start optimizing for the new one.
Stop: Measuring GEO success by citation clicks and traffic from AI Overviews. Start: Measuring branded search lift, mention frequency, and narrative positioning.
Stop: Treating AI citations as backlinks — evidence of authority that might drive traffic. Start: Treating AI mentions as shelf space — presence in the consideration set before the buyer visits any website.
Stop: Building content that answers questions comprehensively. Start: Building content that defines categories clearly — so AI systems learn to associate your brand with specific use cases, outcomes, and customer types.
Stop: Asking "How do we get more AI traffic?" Start: Asking "How do we become the brand AI recommends when our ideal customer describes their problem?"
The ultimate metric: Are you capturing lead generation from AI-driven discovery? Because that's where digital transformation meets revenue.
The companies that figure this out first will build moats that are nearly impossible to displace. Because once an AI system learns to associate your brand with a specific solution space, that association becomes self-reinforcing. Every mention trains the model to mention you more. Every citation increases the probability of future citation.
It's the Digital Ragnarök — not an ending, but a transformation. The old gods of click-through rates and keyword rankings are dying. New gods are being born. The question is whether you'll worship at the right altar.
The Bottom Line
What is the 1% Click Problem? It's the gap between AI citation and actual brand influence. When AI Overviews show your brand but users don't click, you're still winning — if you're measuring the right metrics.
The 1% click rate isn't a bug. It's a feature of a fundamentally different discovery mechanism.
AI Overviews aren't failing to drive traffic. They're succeeding at something else entirely: pre-visit qualification. They're filtering the consideration set before the buyer ever reaches a website. They're making decisions about which brands matter — silently, invisibly, at machine speed.
Your job isn't to complain about the 1%. Your job is to make sure you're in the 1% that matters — the brands that get mentioned, remembered, and searched for later.
Because the click is dead. Long live the consideration set.
Mercury Technology Solutions: Accelerate Digitality.
