.png)
From Reads to Leads is a newsletter for B2B tech founders and marketing leaders whose marketing isn’t working. It’s about positioning, messaging, content strategy, operations, and results. If this was sent to you, subscribe here so you don't miss the next email.
In today's newsletter:
After I posted some recent client results on LinkedIn showing ChatGPT as a major traffic acquisition channel, a lot of people asked how we achieved them.
I’m not ready to share our recipe yet; I want to confirm it works for other clients and give it more time.
In the meantime, I’ve been studying Ethan Smith’s (CEO of Graphite) conversation with Lenny about AEO, and I think his framework offers one of the most grounded playbooks for earning visibility in AI right now.
Here is what he says.
AI responds to questions, not search terms. To understand what questions you want to show up in AI answers for, take your competitor's paid search data – their money terms (use SpyFu or SEMrush), and organic search data – their SEO keywords.
Then literally paste them into ChatGPT with this prompt:
"Turn these keywords into questions people would ask."
It does 80% of the work for you.
Identifying the questions your customers are really asking (not just what everyone else is writing about) is easy if you’re specific about what you do, who you serve, and how you're different.
Too bad most teams never stop to ask what their customers actually care about.
Check out my newsletter issues on the Context Engine where I explain how to write content that builds a brand and works for AI search.
AI gives different answers each time you ask. You need to track your share of voice across multiple runs.
To do that, pick any of the answer tracking tools available. They're all basically the same. Choose the cheapest one that tracks across ChatGPT, AI overviews, Claude, and Perplexity.
Track your questions.
You need to understand what types of content and URLs AI cites. Different types of sites dominate different verticals.
Look at who shows up repeatedly. These URLs tell you where to focus your efforts.
AI pulls from pages that answer the main question and every possible follow-up. One page addressing 50 related questions beats 50 pages with one answer each.
Structure pages with clear H2s for each sub-question. Include "does it work with...", "can I use it for...", "what's the difference between..." sections.
AI cares about volume of mentions across sources. Here are some possibilities you can use:
Most "best practices" in SEO/AEO are wrong because nobody tests properly. You need to know what actually works for YOUR business.
To set up a proper experiment, you need to take 200 questions from your tracking list. This gives you enough data for statistical significance.
Your control group will have 100 questions. Do nothing and just track them.
Your test group will have another 100 questions. Start answering these questions on the platforms you want to test. For example:
Wait for 2 months and measure impact. If your share of voice for the test group improves while the control group stays the same, your intervention has worked.
SEO people are good at on-site optimization but terrible at community engagement. You need different skills for different channels.
Keep your SEO team focused on landing pages and technical optimization. Hire a content creator for Reddit/YouTube.
People ask AI extremely specific product questions. Your help center probably answers 20% of them. The other 80% is opportunity.
Here is what you can do:
Hope it helps. Watch the full video here.
But I can't just leave you with that. My mission has always been to make the internet a more interesting place for content, not more spammy. So here’s a warning for those chasing visibility at all costs:
No amount of AI exposure can compensate for a brand people don’t trust.
Currently, it will suck up whatever content exists without distinguishing between expertise and SEO-optimized garbage. And that’s how spam wins: when people care more about being cited than being credible. It’s the same game the SEO industry played for two decades.
But my guess is ↓
The system will eat itself.
When AI-generated content becomes dominant (and it already is - there's more AI content than human content online now), we get what researchers call "model collapse."
The logic:
His example: Ask "What's the best ice cream flavor?" enough times in this loop, and eventually every answer becomes "vanilla."
When that happens, AI tools will have to become selective.
They'll have to develop their own version of EEAT (Expertise, Experience, Authority, Trust). They'll have to start caring about source quality.
But until then? We're stuck in the spam acceleration phase.
Now is exactly when quality matters most: keep your standards high and invest in authority. When AI citations start favoring true expertise, you’ll already be ahead.
What's your favorite ice cream flavor?
Kateryna
P.S. If we aren't connected already, follow me on LinkedIn and Instagram. If you like this newsletter, please refer your friends.
P.P.S. Need help with quality content? Zmistify your content with Zmist & Copy
What questions should you ask a content writer at a job interview to reveal their skills and experience? These are my top 10 content writer interview questions!
Statistics make your arguments more persuasive, but only if used correctly. Learn more about how to use numbers to give your arguments a firm foundation.
Subscribe to From Reads to Leads for real-life stories, marketing wisdom, and career advice delivered to your inbox every Friday.
