These five AEO myths are keeping your product out of ChatGPT.
I'm Andrei, and I run Literally.dev. We measure what the frontier AI tools say about our clients' products, from ChatGPT to coding agents like Claude Code. This year, we've analyzed more than 100,000 AI conversations for over 80 companies.
I hear these myths on almost every sales call:
- AEO is just SEO.
- AEO has nothing to do with SEO.
- AI never gives the same answer twice.
- Prompts vary too much to measure.
- Any AI ranking is a scam.
All five are wrong. Let me explain why.
Myth 1: AEO Is Just SEO
AEO is not just SEO. SEO is focused on your website. AEO is focused on the AI recommending you.
If you're launching a brand-new website and trying to SEO your way into an AI recommendation, the chance of that happening is close to zero. You won't SEO your way into an answer, because SEO is about ranking your website in the search results.
That's how the old internet worked. You put a lot of time and effort into keyword research, link building and site structure, and you hope Google likes you enough to put your link near the top.
AEO is different, because you don't even need to own a website. If enough trustworthy, unrelated sources say you're the top choice for a question, the AI will recommend you.
Myth 2: AEO Has Nothing to Do With SEO
AEO is not SEO, but it has a lot to do with it.
Without search, an AI can only answer from the data it was trained on. That data is usually about a year out of date, and getting into it takes a lot of time and effort. It's a nice side effect, but unless you have truckloads of money, it shouldn't even be on your radar.
So your next best option is search. Every major AI assistant needs to search the open internet. If you can influence the search results, you can influence the AI's answer. That's why it pays to understand how search works.
And note that the page recommending you doesn't have to be on your own website.
Myth 3: AI Never Gives the Same Answer Twice
The language model itself is a next-word predictor, and it's 100% deterministic. The same words in the same order lead to the same words in the same order, every single time.
I know that sounds wrong if you've asked ChatGPT the same question twice. But I'm talking about the model, not the chatbot built around it. When we build a chatbot, we force variety on purpose. Instead of always picking the most likely next word, we pick at random from the top candidates.
That randomness changes the style much more than the substance. The AI could say "Consider Nike and Adidas" or "Adidas and Nike are popular options." If two brands are interchangeable, they're likely to swap places.
But if one brand clearly dominates, it's mentioned first close to 100% of the time. Ask AI for the top online payment processors, and Stripe is very often number one. It doesn't share first place with anyone.
Myth 4: Prompts Vary Too Much to Measure
Everyone words their prompts differently, so the myth goes that AEO can't be measured.
It's true that people type different words. But if the exact words mattered that much, you'd need perfect grammar and punctuation to get a good answer, and that's not the case. You can type broken English, skip the spaces, or ask in one language and get the answer in another. If the exact words and their order were that important, most of modern AI simply couldn't exist.
AI works on meaning. That's what makes it so powerful. Ask the same question in different words, and it understands it as the same question. If you're technically curious, look up word embeddings. I think you'll find the topic fascinating.
Myth 5: Any AI Ranking Is a Scam
The myth: any tool that gives you a ranking position in AI is a scam.
AEO tools don't work like SEO tools. A ranking in the SEO sense doesn't exist, because the mechanism is different. What AEO tools measure instead is frequency. There are two core numbers:
- Mention rate: how often the AI mentions your brand.
- Citation rate: how often it links to you.
So ranking positions turn into percentages. If the AI mentions you in 3 out of 10 conversations, that's a 30% mention rate. If it links to you once in those 10, that's a 10% citation rate.
The catch is that you have to measure at scale. If you've done performance marketing, you know you can't measure a 2% conversion rate with 10 visitors. If one of them converts, that's already 10%. And most likely none of them will, so your measurement shows 0%, which is false. 0% means you don't have a business, while 2% can mean you're the best-converting business in your vertical.
So you measure AEO by probing AI assistants at scale. You ask them over and over, and you analyze their answers.
That's not to say there are no scams in AEO measurement. My spidey senses turn on when an AEO tool is suspiciously cheap. Probing frontier AI at scale is really expensive. The API prices are public, so you can run the numbers yourself. If one proper measurement costs $50 in tokens, and a tool promises daily measurements for $100 a month, ask yourself how the math works out.
What Real AEO Measurement Looks Like
Here's what a real measurement looks like: our sample analysis of PostHog, from our own pipeline. On product analytics questions, PostHog has 88% share of voice in AI assistants and 94% in coding agents.
You don't have to take my word for it. Our AEO tooling is open source. You can run it yourself and only pay the token prices. Or let us run it for you: tech companies can apply for a free AEO analysis.
AI recommendations aren't random. They can be measured, and what can be measured can be improved.
