How is GEO different from SEO?

SEO fights for a position and a click on a link. GEO fights for inclusion in the neural net's answer itself - which retells sources and often keeps the user from leaving for your site. It's not a replacement for SEO but its continuation: the foundation is the same (index, content, links), with a new goal added - to be cited.

Will llms.txt get me cited in AI?

On its own - no. llms.txt helps the machine understand your structure and facts faster, but if you're not in the index and nobody links to you, there's nothing to cite. On my fresh site llms.txt was in place and citations were zero. Base first - index and links - then fine-tuning.

How long until GEO shows results?

About as long as SEO - it's the same base. Pages get indexed, the link profile builds, then AI starts pulling you into sources. There are no instant buttons. IndexNow speeds new pages into the index but doesn't replace depth and links.

Do I need a separate budget for GEO?

Usually no. For services and small business GEO is hygiene on top of normal SEO: factual pages, Schema, reviews, links, regular checks of AI visibility. A pricey 'GEO tool' is needed less often than it's sold.

GEO by the numbers: 40 citations vs 0, and why

Q: How do I measure whether AI cites me?

Since February 2026 Bing Webmaster Tools has an AI Performance report: it shows when your site lands in the sources of Copilot answers and Bing AI summaries, and for which phrases. Plus ask ChatGPT, Perplexity and your local search AI about your topic and see if you make the answer.

SEO isn’t dead. The results page just keeps becoming a ready AI answer instead of a list of links. By Semrush’s 2025 measurements, 43% of searches with a Google AI Overview end without a single click to a website (Semrush via Position Digital). The person read the answer and closed the tab. The question “how do I get into that answer” is GEO - generative engine optimization, optimizing to be cited in AI answers.

Tons of articles have been written about GEO, and almost all of them are generalities. I have something more concrete: two numbers from my own projects that explain the mechanics better than any guide. 40 citations versus 0.

What GEO is and why now

GEO is the work of getting a neural net to cite your site when it answers a person. AI Overviews in Google, Copilot in Bing, answers from ChatGPT and Perplexity - they all pull fragments from the index and retell them in their own words.

Why the topic sharpened right now shows in two recent stories. First, an AI answer is now legally the platform’s own content: a German court held Google liable for what its AI made up (breakdown here). Second, the neural net increasingly builds the short list a person picks a contractor from (more on that). In both cases the winner is whoever the AI takes into its sources. Everyone else simply isn’t in the conversation.

GEO used to be measured by eye: manual prompts, third-party services, guesswork from logs. Since February 2026 there’s a scoreboard. Bing Webmaster Tools launched the AI Performance report: it shows when a site landed in the sources of Copilot answers and Bing AI summaries, and for which phrases. For the first time you can look at citations in numbers, not promises.

40 vs 0: my two numbers

I opened that report for two of my projects .

A cleaning project I run marketing for: a domain with history, systemic SEO, 245 URLs in the sitemap. In six weeks - 40 citations in Copilot answers, peaking at 8 in a day. And the trend is rising: the last two weeks gave 19 citations against 8 in the previous two.

A personal site on a fresh domain: 74 pages, launched in March. Over the same period - 0 citations and 3 impressions in Bing. To the search engine the site is nearly invisible, and Copilot simply mirrors that.

One person, the same principles, llms.txt on both. The difference is the index. Whoever Bing sees well is who Copilot pulls into answers. Whoever it doesn’t see isn’t there.

What actually drives citations

Not magic and not one magic file. Three things.

Index depth. Copilot cites what already sits well in Bing search. No pages in the index, nothing to cite. The cleaning project has 245 URLs; the personal site has crumbs in the index.
A mass of useful pages. Not five “about us” pages, but real volume that answers people’s questions. That’s what the neural net assembles an answer from.
A link profile. The only recommendation Bing gave my personal site: “not enough inbound links from quality domains.” Links matter here too.

llms.txt and Schema are fine-tuning on top of the base. They help the machine understand faster who you are and what you do. But with no base, there’s nothing to tune.

A GEO checklist

What to do, in order, without the esoterica.

Open your scoreboard. Bing Webmaster Tools - AI Performance. See how many citations you have and how many pages Bing even sees. It’s free and sobering.
Give the machine a factual minimum. Schema markup (Organization, services), clear pages, an llms.txt. So the AI knows who you are without guessing.
Grow the index and the links. Useful pages for real questions plus links from normal domains. It’s slow and dull, but it’s exactly what moves citations.
Speed up indexing with IndexNow. It instantly pings Bing and Yandex about new and changed pages. On my sites it fires automatically on every deploy.
Measure monthly. Not once. The AI Performance numbers plus a manual check in ChatGPT, Perplexity and local search AI. A one-off check is a snapshot; a regular one is a trend you can show a client.

What not to do

Don’t buy “GEO magic.” If someone sells you AI citations as a pricey standalone service that bypasses the index, it doesn’t work. Citations grow from the same base as regular search.

Don’t pray to llms.txt. The file is useful, but it won’t replace pages or links. I tested that on myself: the file was there, citations were zero.

Don’t treat GEO as a separate channel. It’s hygiene on top of SEO, not a second marketing system. One foundation, with a new goal added - being in the answer.

Bottom line

GEO sounds like a new discipline, but it’s really an honest continuation of SEO for a new results page. A neural net cites not the best but the legible: the one it sees well in the index and the one others link to. My 40 vs 0 is about that, not about secret tricks.

So the plan is down to earth. Base first: index, useful pages, links. Then fine-tuning: Schema, llms.txt. And regular measurement through AI Performance, so you see numbers instead of taking it on faith.

Want to know whether AI search sees you and how many citations you have - message me on Telegram @dimik90 or order an audit. I’ll look at your Bing scoreboard and where your base is thin.

How this foundation is built from scratch is in the cleaning case: SEO from zero in 50 days. And why all this matters now is in the breakdowns on Google’s liability for AI and marketing to AI agents.