SEO and GEO
GEO without the magic. How to actually show up in ChatGPT, Gemini and Perplexity answers, and what no one can honestly guarantee
Agencies promise “a presence in AI answers.” We explain what’s true about that, what’s a sales pitch, and what you can realistically do and then measure.
For two years, a new category of services has been growing: GEO, or generative engine optimization. The promise is tempting. Instead of typing a phrase into Google, more and more people ask ChatGPT, Gemini or Perplexity: “what’s the best vacuum for dog hair under PLN 1,500?” Whoever shows up in the answer makes the sale. Whoever doesn’t, doesn’t exist. Hence the offers: “we’ll get your brand into AI answers.”
We’ll say it plainly: nobody can honestly guarantee that. Model answers change from day to day, between versions, and even between two identical queries. Google officially says you don’t need to do anything special for AI Overviews beyond good SEO. Tracking “whether the chatbot recommends us” is an interesting observation but a terrible measure of success. If someone is selling you GEO as a citation counter, they’re selling you a lottery ticket.
What actually makes a difference
The good news is that what drives visibility in AI answers is fairly down-to-earth, and it largely overlaps with what we’ve called solid SEO and solid data for years.
- Product data machines can read: name, features, dimensions, material, price, availability, EAN codes, schema.org structured data. A model won’t guess that a dog bed has a cover you can wash at 60°C if that information isn’t in the data.
- A site that’s fast and accessible to crawlers. It sounds trivial, yet we still see stores that block crawlers from product pages or load content only through scripts.
- Content that answers questions instead of just describing products. Guides, comparisons, answers to common customer doubts, reviews and real-life use stories. Those are exactly the fragments models build their answers from.
- Credibility beyond your own site: reviews, mentions, citations on industry websites, consistent brand information across different places. A model that sees the same information in several sources is more likely to repeat it.
- Brand consistency. If the same product has three different names in three channels, to a model those are three different products, and none of them is certain enough to recommend.
Notice there are no tricks on this list. No hidden tags that “fool” the model, no special files just for AI. Just product knowledge, well described and easy to find. Exactly what the customer needs, too.
How to measure it, if not by citation count
We measure what can be counted and what translates into money. First, traffic and queries coming from AI assistants: most of them leave a trace in analytics as a referral source, and some show up in server logs. Second, changes in organic traffic and in the number of keywords the store ranks for, because the same work improves both at once. Third, conversion and margin from that traffic, because a visit that starts with “which product should I choose” is usually closer to a decision than one that starts with a generic keyword.
We treat watching chatbot answers as a qualitative signal: what the model says about the category, who it mentions, which product features it considers important. That’s valuable input for content planning. But we don’t report it as a result, because a month from now it could look different through no fault or merit of ours.
Who GEO makes the most sense for
Today the biggest advantage goes to specialist stores: boutiques with a curated range, stores that offer advice, brands with their own product expertise. Their problem is that this expertise lives in people’s heads and conversations, not on the website, so the model has nothing to cite. Writing it down as guides, comparisons and customer interviews is enough to start appearing in answers that used to go to bigger players with more text.
For large catalogs, GEO is above all about clean data, because with thousands of products no human can describe each one individually. That’s where AI comes in on the store’s side: generating descriptions and attributes according to the brand’s rules, with quality control before publishing.
A short checklist for today
- Do your product pages have structured data, and is it correct? (Google Search Console will show you the errors.)
- Can crawlers access your product pages and images?
- Does your store have at least one piece of content that answers a customer’s question instead of just describing a product?
- Is the product name the same in your store, in Merchant Center and on Instagram?
- Can you identify traffic from AI assistants in your analytics?
If the answer to three of the five is “I don’t know,” you don’t need GEO. You need order. GEO will follow.
