Noah Greenberg recently shared something that caught my attention: imagine talking to marketers in 1999 and telling them, "google only makes up 2% of your revenue, dont spend too much time SEO or SEM".

That line is a useful gut check for where we are right now with AI-driven discovery. Many teams look at today’s AI referrals and conclude it is too small to prioritize. Noah’s point is that the wrong comparison is current percentage. The right comparison is trajectory and competitive leverage.

In the early Google era, companies that learned SEO before it was mainstream gained durable advantages. Noah called out Zillow and NerdWallet as examples of brands that rode organic search to leapfrog incumbents. The marketers who waited until SEO was already a massive channel for competitors ended up, as he put it, "old marketers, filled with regret, waiting to rank".

The modern parallel is GEO (Generative Engine Optimization) and AI Search Optimization. The label itself is new. As Noah noted, "Fortunately, the term "GEO" didn't exist 12 months ago" and "AI Search Optimization" only really went mainstream recently. Translation: you are not late yet. But you can become late quickly.

The SEO-to-GEO shift: what actually changed?

Google SEO trained us to think in clicks, rankings, and sessions. LLM-driven discovery (ChatGPT, Claude, Gemini, and increasingly AI features inside traditional search) forces a different mental model:

The interface is conversational. Users ask longer, messier questions.
The output is synthesized. The model may cite sources, summarize, or recommend vendors without sending a click.
The competitive set can widen. A user who would have searched "best expense software" might ask, "What should a 200-person fintech use for expense management and why?".

This does not replace classic SEO. It expands the surface area where brand perception is formed. If an LLM is shaping consideration, your marketing job is to influence what it knows, what it trusts, and what it cites.

A DIY crash course (expanding on Noah’s playbook)

Noah laid out a simple path: get smart, run tests, measure the right thing, and then bring leadership along. I would structure it as a 6-month program any marketing team can run.

1) Get smart (fast) so you stop guessing

Noah’s first step was blunt: "Get smart." The fastest way to lose time in a new channel is to debate it without shared vocabulary.

Start with two tracks of learning:

Concepts and mechanics. Understand how LLM answers are generated, what citations mean (and when they do not appear), and how retrieval works.
Practical patterns. Learn what types of pages and content formats tend to get referenced (definitions, comparisons, original data, FAQs, clear entity descriptions, and reputable third-party coverage).

Noah recommended:

Muck Rack’s free 60-minute Fundamentals of GEO course
Ethan Smith with Lenny Rachitsky’s Ultimate Guide to AEO podcast

He also suggested following practitioners who publish timely updates and analysis:

Aleyda Solís for rapid changes and news
Kevin Indig for strategic frameworks
Josh Blyskal for data-driven insights into citations

My add-on: pick one internal owner to maintain a living memo (one page) of learnings, definitions, and experiments. The goal is not mastery. It is a shared baseline so the team can move.

2) Run your own tests (play before you over-plan)

Noah’s advice here is the most important: this is changing too quickly to wait for a perfect playbook.

A lightweight daily habit can outperform a quarterly strategy deck:

Spend 15 minutes a day asking LLMs the questions your buyers ask.
Use multiple models (ChatGPT, Claude, Gemini) because results vary.
Track whether your brand is mentioned, how it is described, which competitors show up, and what sources are cited.

Then move from observation to intervention:

Launch a few targeted pages (or improve existing ones) aimed at high-intent, high-confusion questions.
Clarify your product category, use cases, and differentiators in plain language.
Publish proof points that a model can summarize easily: benchmarks, clear pricing principles, integration lists, security posture, and common objections.

Two practical testing tips:

Test one variable at a time. Update a page’s structure (for example: add a comparison section, an FAQ, or a glossary definition), then re-check model outputs over time.
Build a prompt library. Save the exact queries you use so you can re-run them consistently and detect changes.

Noah mentioned that changes can happen overnight and that tools exist to track this. Even without sophisticated software, you can create a simple spreadsheet: prompt, model, date, brand mention (Y/N), sentiment, cited sources, and notes.

3) Update your metrics: the short-term goal is not clicks

This is the part many teams miss because we are trained to worship attribution.

Noah’s warning is clear: "Unlike Google, LLM's send a lot less direct traffic." If you judge GEO by sessions alone, you will likely under-invest.

So what should you measure in the first six months?

Share of voice in AI answers. For a fixed set of prompts, how often are you recommended or included in shortlists?
Message pull-through. When you are mentioned, do the reasons match your positioning, or is the model guessing?
Citation footprint. Which third-party sources does the model cite when discussing your category, and are you present in those sources?
Down-funnel lifts. Look for correlated movement in branded search, direct traffic, demo requests, or sales conversations that mention "I saw you recommended".

A practical compromise is to treat GEO like PR plus SEO: part awareness, part credibility, part discoverability. Not everything will be neatly attributable, but it can still drive real pipeline.

4) Bring leadership along (and make yourself the owner)

Noah advised: "Present to your leadership team. Insert yourself as the authority." That is not about ego. It is about making sure your company has an internal champion who can coordinate experiments, align content, and communicate results.

If you want buy-in, do three things:

Show evidence, not hype. Bring screenshots of real model outputs for your key prompts: where you appear, where competitors appear, and what sources are shaping the narrative.
Define a 90-day plan. A small list of experiments, owners, and reporting cadence beats a vague promise to "do GEO."
Translate to business risk. The message is simple: if customers outsource research to AI, then AI becomes a gatekeeper for consideration.

Every CMO is trying to decide who leads this. If you consistently deliver small wins and clear reporting, you become the default owner.

The overlooked advantage: build a GEO study group

Noah ended with a recommendation I love: create a study group. Find three or four peers at other companies and meet biweekly to compare notes.

Why this matters:

The space changes fast. External feedback loops help you adapt.
You get pattern recognition: which content types earn citations, which prompts matter, and which tactics are just noise.
You avoid reinventing the wheel alone.

Structure it like a book club:

10 minutes: what changed in model behavior since last meeting?
20 minutes: show one experiment result each
20 minutes: share one new resource (course, tool, case study)
10 minutes: pick next experiments

A simple 6-month GEO plan you can start this week

If I condense Noah Greenberg’s message into an execution checklist, it looks like this:

Learn the basics of GEO and AEO so your team shares language.
Build a prompt set that mirrors your buyer journey.
Audit current AI answers for brand presence and positioning accuracy.
Publish and update content that is easy to cite and hard to misinterpret.
Track share of voice, message pull-through, and down-funnel signals.
Report wins to leadership and make GEO an owned function.
Join forces with peers to keep learning.

You are not behind. But as Noah implied with his 1999 analogy, the window where early movers can compound advantages does not stay open forever.

This blog post expands on a viral LinkedIn post by Noah Greenberg, CEO at Stacker. View the original LinkedIn post →