GEO and AEO, how I get a product found and cited inside AI answers.

I'm Eduard Dziak, Head of Marketing at ZIK Analytics. I own SEO and GEO/AEO there, getting a SaaS product found in search and cited inside ChatGPT, Perplexity, Gemini and Google's AI answers, as part of running growth end to end. ZIK is based in Lisbon, and I work remotely from Kuala Lumpur.

I have done SEO hands-on since 2018, across enterprise B2B and B2B SaaS before ZIK, so I came to AI search from the search side.

What I own at ZIK Analytics, as of October 2026

I have run marketing at ZIK Analytics since April 2025, and I lead a team of eleven, three of them managers, across paid, SEO, AI search, email and the website. By self-attribution, generative AI is now our fourth-largest acquisition channel.

At ZIK I rebuilt the tracking first, and the paid account then turned profitable. Google Ads has since reached roughly 250% ROI, measured on the revenue the ads drove against the ad spend. I wrote up how I run that side on my paid acquisition page.

I built the first framework at ZIK that measures AI search next to classic SEO. It tracks AI mentions, citations, AI referral clicks and page performance, maps our priority prompts to the pages that should win them, and feeds a reporting agent the CEO can ask what is working and where we are weak. I validate it with repeated prompt tracking and source checks, and I check visibility trends against the revenue downstream.

I cannot point to one initiative that did it, because I planned SEO and AEO as one program across channels. So below is how I approach it, in the order I wrote it down when a journalist asked me.

How I get a product cited

Publish data nobody else has

AI systems and search engines cite information that exists nowhere else, and original data is the easiest way to have some. I had watched several interviews with engineers from OpenAI and Anthropic, and they all said the same thing, that they can improve the algorithm and scale the computing, but the data are limited. Google’s own guide for AI search asks for the same thing in plainer words, “non-commodity content”.

So I publish reports built on the company’s own data, and then use those figures across the articles and, where they fit, the marketing pages. The key is consistency, and it is one of the real challenges. Whenever I cite our data, I make sure it is the same data everywhere.

Write headlines as questions

I moved from keyword-only headlines to question-based and statement-based ones, because a question is the shape of what an answer engine is matching.

Ask the engines, then check your own site

To appear, you need to know what information the engines want. So I run the prompts where we want to appear and look at exactly what they give back, the features, the data, the history and so on. AI always surfaces broader information. Then I check that our site says those things. It does not all need to be on one page. It needs to be on the relevant pages.

Keep the product documented, and current

I believe the more an AI knows about a product, its features and what it can do, the more it can recommend it. So I document every feature update, big or small, and answer as many real customer questions as I can, while making sure the help content does not cannibalise the SEO pages.

Outdated information is dangerous. Everything has to stay current and never contradict the rest of the site. I also saw somewhere, and honestly I don’t remember where, that a changelog is great for getting cited, because it is where the latest information lives.

Use the same topics on every channel

I align the content plan across every channel, so the same topic gets covered from every side and the engines get a strong signal. YouTube is one of the sources AI systems read, so it belongs in that plan as much as the blog does.

Make it cheap to read

The cheaper you make it for AI to get your content, the better. I keep the HTML easy to read, especially for data reports, where tables and charts matter, and keep the internal links and page speed in order and the page structure clean, so the main content is obvious. On this site, every page also has a plain markdown version, linked at the top.

Build the brand

Probably the most powerful lever for getting cited is branding. Engines repeat the brands that other people talk about, so mentions in the press, on podcasts and from partners count for as much as links.

This site is where I test it

This site runs on the same ideas, and the evidence is public.

What actually reads this site shows which AI crawlers and search engines fetch these pages, and it checks requests that claim to be AI crawlers against the addresses their operators publish. Its first report showed how easily a chart like that can be poisoned. On 22 August 2026, 0 of 140 requests claiming to be ChatGPT-User came from OpenAI’s published addresses, and that one hour lifted the site’s apparent AI share from 3.5% to 16.4%. That is why every figure on that page carries its window and method.

What the web actually marks up is my monthly measure of which Schema.org terms websites really use, recalculated without the rows that make most published figures wrong.

Every page here also has a markdown version, and /llms.txt lists them.

Where GEO fits in growth

Discovery has moved to AI, but evaluation still goes to search engines, where people go to verify what they were just told. The funnel from informational to commercial to transactional intent still works. What changed is where discovery starts, and that is why I run GEO inside growth, next to SEO, paid, lifecycle and the website, instead of as a separate project.

Questions

What are GEO and AEO, and are they just SEO?

Generative engine optimization (GEO) and answer engine optimization (AEO) are two names for the work of getting an AI assistant to describe your product correctly and cite your pages when someone asks a question you can answer. Google’s view, in its guide to optimizing for generative AI search, is that it is “optimizing for the search experience, and thus still SEO”. My own view is blunter. Most AI-first content is entity SEO with a new name.

Can AI tools do this for you?

They can do a lot of it. I run a content pipeline in Claude Code that goes from keyword research to a published, internally linked page. But I would rather write something by hand than publish AI slop, so a person still decides what is true and what is worth publishing.