Search is moving into AI answers: three kinds of sites that win

By recent measurements, around 43% of commercial queries in the Russian-speaking web already come with an AI answer. The user gets a solution right on the results page and clicks links less and less. The informational traffic the industry built strategies on for a decade is shrinking in plain sight. A forecast now making the rounds in the SEO world names three kinds of sites that come out ahead. I agree with it, and I can check every point against my own site: part is built, part is honestly unfinished.

Brands

The logic is plain. When users know you, search for you by name, click you and come back, you're a legible source for the search engine. Brand demand stops being a marketing metric and becomes a survival condition in results, classic and generative alike.

Nothing to brag about here: brand demand for ipak.dev today is near zero, and that's the biggest risk in my whole content strategy. What I'm doing about it: the entity is assembled in markup (author, profiles, one canonical term), and beyond that only seeding works – publications on outside platforms that tie the term AIOS to its source. In 2026 a beautiful site with no external mentions stays invisible, whatever amount of content it holds.

Primary sources, not retellings

AI can retell the whole internet, so retellings have depreciated to zero. It will cite whoever has data of their own: measurements, cases, screenshots, conclusions from the person who actually did the work.

My blog has been built that way from the first post: 81% of requests on a local model, half a dollar a week on LLM calls, deploy incidents published as they happened, with timelines. These numbers can't be rewritten by anyone else, because they only exist inside my system.

The second half of the job is machine readability, and it's the half everyone skips. Every post on this site is served as clean markdown at /blog/post-name.md, an llms.txt digest of the whole site sits there for AI crawlers, and the definition of AIOS exists in exactly one wording – in the prose, in the DefinedTerm markup, in the FAQ. The wording lives in a single canonical file that every page reads from: a crawler has no chance of meeting two different definitions. Being a primary source means owning your data and serving it to the machine in a usable form.

Commercial landing pages

An AI answer won't deliver a couch, renovate an apartment or roll out a system into a business. For «order», «turnkey», «price with delivery» the search engine still has to send the person somewhere. Commercial traffic stays, but everyone who lost informational traffic will run there, and competition multiplies.

So the landing page has to close intent completely: timelines, who it fits and who it doesn't, cases, guarantees, a clear request form. A page that says "we provide services" is out of the game.

This is my weakest block, and I'll say it straight. The consultation page works and brings requests, but it doesn't yet close intent by the full list above – that's exactly the logic I'm reworking it in. There are no prices on the site by design: the proposal is calculated after a diagnostic of the specific process, so intent gets closed by everything else – timelines, boundaries of fit, live examples.

What a business should do with this

The check takes ten minutes. Ask ChatGPT and Alice about your category: who gets cited, are you there. Then look at what share of your traffic is informational articles. If it's large, you have a window of a year or so to rebuild: gather your own data instead of retellings, mark up your entities, and bring commercial pages to full intent closure.

On my side this work is carried by the system: the content pipeline, the markup, the AI-answer tracking are part of the same AIOS that sorts my mail and reconciles the bank. How it's built is in what is an AIOS. To try it on your business – book an online consultation.