AI SEO & GEO Services
Your buyers ask an assistant before they open a search engine. I build the layer that gets ChatGPT, Perplexity, Claude and Google AI Overviews to name your product and quote it correctly — in your codebase, not in a monthly slide deck.
AI SEO — also called generative engine optimization, or GEO — is the work of getting a product found and cited inside AI answers: ChatGPT, Perplexity, Claude and Google AI Overviews. Oleg Sotnikov builds that layer as an engineer: llms.txt and markdown endpoints for agents, structured data wired into one entity graph, pages that open with a direct answer, crawler access, and a repeatable way to measure who gets cited. Classic search keeps working, because the same structure that makes a page quotable is what makes it rank. This site is the reference implementation — open /llms.txt or read the source.
What the AI Visibility Layer Is Made Of
Six pieces of work, in the order they usually pay off.
The technical GEO layer
llms.txt and llms-full.txt give agents a curated map of the site. Structured data is wired into one entity graph with stable @id anchors instead of scattered snippets. Every commercial page opens with a paragraph an engine can lift whole. All three ship on this site.
Content pipelines that scale
One good page will not move an engine. I build the publishing pipeline — sourcing, drafting, review, second language, sitemaps, RSS, IndexNow — so answer-shaped pages ship every week without a content team. I wrote the Generated.app pipelines running across 19 production sites.
Crawler access and indexing hygiene
GPTBot, ClaudeBot, PerplexityBot, Google-Extended and the rest have to reach you first. Bot-fight rules at the CDN quietly hand them a 403, and a 404 pinned in a cache outlives the bug that caused it. So I check the response the crawlers actually get.
Measuring AI citations
The point of the work is a number you can watch. I set up a repeatable check that replays your buyers' questions across the engines, scores who gets named, and traces each quote back to the page it came from. Crawler hits in the server logs and assistant referrals in analytics fill in the rest.
Classic SEO stays intact
I don't trade anything away for AI visibility. Crawl budget, Core Web Vitals, canonicals and hreflang, internal linking: these still decide your Google results. A slow site also gets its fetch budget cut by the AI crawlers.
Built for SaaS and product teams
When someone asks an assistant which tool to use, it reaches for comparisons, integration and alternative pages, docs and changelogs. For a SaaS, what matters is being the product named in that answer, so I shape the site around the questions that end in a signup.
How I Work
Visibility audit
I ask the engines the questions your buyers ask and record who gets cited today. In parallel I check what the crawlers receive: robots rules, CDN behavior, server-rendered HTML, schema, speed. You get the gaps ranked by what they cost you.
Technical layer
llms.txt, markdown endpoints, one connected entity graph, a direct answer at the top of each commercial page, crawler access unblocked. This part is code, and it ships in days, not quarters.
Content engine and measurement
Then the recurring work: a publishing pipeline aimed at the questions that end in revenue, and a scheduled citation check so you watch a trend instead of guessing.
Why Me
- I built the Generated.app publishing pipelines, now running across 19 production sites — content for AI search at volume is my daily work, not a service line I bought and resold
- This site is the working demo: /llms.txt, entity-graph JSON-LD, a markdown endpoint on every article, an answer block on every commercial page. Read the source before you hire me
- 25+ years in IT and 1,000+ projects behind me, plus an AI platform of my own that processes 11B+ tokens a month — I work on both sides of this problem: the page, and the model reading it
AI Visibility Monitoring
AI brand visibility is easy to assume and hard to check: answers are composed fresh each time, and no rank tracker covers them. So I measure it by putting the same questions to the engines on a schedule.
AI visibility tracking
A fixed list of your buyers' questions, put to ChatGPT, Perplexity and Google AI Overviews on a schedule. For each answer I log whether you were named, what got quoted, and from which page. Change the list between runs and the trend means nothing.
Share of model
Of the answers where a product like yours could be named, what fraction names you and who takes the rest. Share of model is the same idea as share of voice, counted in AI answers instead of media coverage, and it moves slowly enough that a real gain stands out from the noise.
AEO tools
AEO tools — answer engine optimization tools — automate the polling and draw the chart. The loop is what pays: pick the queries, measure, fix the pages that lost, measure again. I run it across the Generated.app pipelines on 19 production sites.
Frequently Asked Questions
What is AI SEO?
AI SEO, also called generative engine optimization (GEO), is the work of getting a product named and quoted inside AI answers — ChatGPT, Perplexity, Claude, Google AI Overviews — rather than only in a list of blue links. It combines machine-readable surfaces (llms.txt, structured data, clean markdown) with page structure that puts a direct answer first and content that covers the questions buyers actually type. Oleg Sotnikov builds that layer as an engineer, inside the client's codebase.
How is AI SEO different from traditional SEO?
Traditional SEO competes for a position in a list of results. AI SEO competes to be the source a model quotes while it composes one answer. Citations matter more than rankings there, buyers type whole questions instead of keywords, and the reader is a machine that never runs your JavaScript. The foundations overlap — crawlability, speed, structure, real expertise — which is why this work strengthens Google results rather than replacing them.
How do you measure whether AI engines cite us?
By running a fixed set of your buyers' questions across the engines on a schedule and recording who gets named, what text is quoted, and which URL it came from. Server logs add the other half: which AI crawlers fetch you, how often, and what they receive. Analytics then shows the sessions arriving from ChatGPT, Perplexity and the rest. None of this is as tidy as a rank tracker, and anyone quoting you a precise share-of-voice number is inventing it.
How long until we see results?
The technical layer ships in days, and crawlers pick it up on their next fetch cycle. Being cited takes longer, usually weeks to months: models rebuild their picture of you from a corpus that refreshes slowly, and a citation still depends on having something worth quoting. I don't promise a date, because nobody can honestly give you one for this channel yet.
Do you replace our SEO agency or work with it?
Either way works. Teams searching for an AI SEO agency — or a SaaS SEO agency with GEO on the menu — mostly find classic agencies with a new label; I work as an AI SEO consultant and engineer instead. Most often I build the technical layer and the measurement while your agency keeps doing content and links, which finally turns their pages into citable sources. If there is no agency, I can run the whole thing — though I would rather hand your team a pipeline they own than sell an open-ended retainer.
How do I track my brand's visibility in AI answers?
Fix a list of the questions your buyers actually ask, put it to ChatGPT, Perplexity and Google AI Overviews on a schedule, and record whether you were named and which page got quoted. Roll that up into share of model: the fraction of relevant answers that mention you, read next to competitors. AEO tools automate the polling; the number moves when you rewrite the pages that lost and re-measure the same list.
Find Out What the Engines Say About You
Thirty minutes. Before the call I run a few of your buyers' questions through ChatGPT and Perplexity and tell you what comes back.
Scoped per engagement. No long-term retainer required to start.
Related reading
AI search, structured data, and content pipelines that hold up.
