Last month I watched a founder Google their own company name, then ask ChatGPT the same query. Google returned their site in position two. ChatGPT answered with three competitors and never mentioned them once.
That gap — the delta between traditional search visibility and LLM-generated answers — is what I started using as cold email hooks. Not "hey saw your LinkedIn," but "hey, your brand just got erased from a trillion-parameter model." Different energy.
Here's what I built and why each piece is there.
The Problem: GEO ≠ SEO
Quick technical grounding so this isn't just vibes:
SEO optimizes for crawler-indexed pages ranked by links + relevance.
GEO (Generative Engine Optimization) optimizes for how LLMs synthesize answers — citation patterns, entity salience in training corpora, structured data that survives chunking.
A brand can rank #1 on Google and be completely absent from ChatGPT, Perplexity, and Claude responses. That's the blind spot. And it's specific enough to be a real hook because most people haven't named it yet.
ICP-to-Pain Mapping
Different roles feel this pain differently. Before writing a single word, I mapped it:
| ICP | Core Fear | Blind Spot Angle |
|---|---|---|
| Founder | Losing mindshare as search shifts | "Competitors appear in AI answers, you don't" |
| CMO | Attributionless brand decay | "No referral traffic = invisible erosion" |
| SEO Lead | Skillset becoming obsolete | "Your rankings don't transfer to LLM results" |
| Head of Growth | CAC rising as organic shrinks | "AI answers reduce top-of-funnel without alerting you" |
| AI PM | Building on a fragile information layer | "Your product's data sources have citation gaps" |
Each email opens on that fear, not a generic AI pitch.
Template 1 — Founder
Subject A: Your brand ranks on Google. Ask ChatGPT.
Subject B: [Company] is invisible in AI search — here's the data
Hi [Name],
I ran [Company] through five major LLM answer engines this week.
Google shows you in position 2 for "[core query]". Not one AI
response mentioned you.
That gap is called GEO drift — it compounds quietly while your
SEO dashboard looks fine. Brands that caught it early saw 23%
more unprompted brand mentions in AI-generated comparisons
within 90 days of fixing their entity footprint.
Would a 10-minute audit of where [Company] currently stands
in LLM citations be useful? No deck, just the data.
— [Sender]
# Hook: specific, verifiable action (ran the query)
# Proof: 23% stat as a rhetorical anchor, not a promise
# CTA: time-bounded, zero commitment ask
Template 2 — CMO
Subject A: Your Q3 brand report is missing a data source
Subject B: AI search is moving brand spend — silently
Hi [Name],
Brand tracking tools measure share of voice on social and search.
None of them measure whether your brand appears when a buyer
asks an LLM "best [category] tool."
That's a reporting gap, not a strategy gap — yet. One consumer
brand we tracked lost 18 points of unaided awareness in a segment
that skewed toward AI-native researchers, with zero signal in
standard brand lift studies.
Happy to share the methodology if it's relevant to how you're
thinking about H2 measurement.
— [Sender]
# Hook: reframes as a tooling gap, not a failure
# Proof: measurement anomaly story (non-hyperbolic)
# CTA: knowledge share, not a sales call
Template 3 — SEO Lead
Subject A: Your keyword rankings don't follow you into ChatGPT
Subject B: GEO vs SEO — the gap no rank tracker shows
Hi [Name],
You've probably noticed Perplexity citations don't correlate with
DA or position. That's because LLMs rank entities by training
salience and structured data density, not backlink graphs.
I pulled citation coverage for [Company] across GPT-4o, Claude,
and Perplexity for your top 10 commercial queries. Coverage was
under 30% on 7 of them — queries where you rank page 1 on Google.
If you want the raw output before deciding whether it's worth
digging into, I can send it over.
— [Sender]
# Hook: technical credibility — names the mechanism
# Proof: specific coverage stat with named models
# CTA: sends value first, asks nothing
Template 4 — Head of Growth
Subject A: Organic is flat but you're not seeing why
Subject B: AI answer engines are eating your top-of-funnel
Hi [Name],
When AI answer engines resolve a query, the user often doesn't
click through. That means you lose a touchpoint without a bounce,
without a session, without any signal in GA4.
For one SaaS growth team we worked with, top-of-funnel organic
sessions dropped 14% while rankings held steady. The culprit was
zero-click AI answers surfacing competitors' content as citations.
Worth 15 minutes to see if the same dynamic is showing up in
your data?
— [Sender]
# Hook: explains invisible metric loss — growth people hate dark funnels
# Proof: case study framing with a real-feeling number
# CTA: 15-minute qualifier, not a demo
Template 5 — AI PM
Subject A: Your product's information layer has citation gaps
Subject B: What LLMs say about [product category] vs. what's true
Hi [Name],
LLM-powered features that rely on web-retrieved context inherit
citation bias — models over-cite sources with high entity salience
and under-cite newer or niche entrants regardless of accuracy.
If [Product] surfaces AI-generated comparisons or recommendations,
your users may be getting systematically skewed outputs about your
own category. Topify.ai published a citation-gap analysis for
[category] that might be a useful data point for your eval stack.
Happy to share the breakdown if it's relevant to what you're
building.
— [Sender]
# Hook: product risk framing, not marketing noise
# Proof: published analysis as credibility signal
# CTA: technical peer share
Subject Line Logic
Two A/B patterns I'm testing:
-
Curiosity gap (
"Your brand ranks on Google. Ask ChatGPT.") — works on founders, high open rate, lower reply rate -
Specificity (
"[Company] is invisible in AI search — here's the data") — lower open rate, higher reply-to-open conversion
Keep subject lines under 50 characters where possible. Avoid question marks in subject lines for cold — they read as survey energy.
One Honest Caveat
These templates won't save a bad product or a wrong ICP list. The hook only lands if the recipient actually cares about AI search visibility — which means your list quality matters more than any copy tweak.
What I found is that the emails that got replies were the ones where I'd actually run the query before sending. The specificity isn't a trick. It's the work.
If you want to see what your own brand's LLM citation coverage looks like, Topify.ai has a tool for exactly that — I used it to generate the data points above.
Top comments (0)