What AIO is, and why it matters more than regular SEO
A buyer used to open search, skim ten links, and call two or three of them. Now they ask an AI assistant and get one answer with one or two names. If you are in that answer, you get the call. If you are not, a competitor got it for free.
AIO (AI Search Optimization) means preparing the site and the business listings so language models can recommend the business from facts. Facts, structure, and trust, instead of swapping the meaning for keywords. Below: how AIO differs from SEO, what actually affects answers in Alice and ChatGPT, and where owners break their own domain.
SEO and AIO do different jobs
SEO is built for a search crawler: indexing, rank, a click on the snippet. AIO is built for the model's answer: were you named, in what wording, in what context.
In SEO you can sit in fifth place and still get traffic. In AIO there is no fifth place. There is "named" and "not named". For a local business that difference is sharp. The person does not compare fifteen sites. They trust the assistant's voice or paragraph.
Classic SEO is still running. If you only optimize for Google and Yandex in the mode of "ten blue links", you are preparing for the last channel. In parallel you need a layer for AI: llms.txt, schema, an FAQ built for commercial questions, matching NAP data (name, address, phone), and open bots where that is safe.
What models actually look at
OpenAI and Yandex have not published a checklist for "how to get recommended". The same signals keep showing up in audits of hundreds of sites:
- A clear description of the service and the place. Not "quality service", but "implants in Kazan, from 45,000 ₽, same-day booking".
- An FAQ with real client questions, not SEO paragraphs.
- Schema.org markup: Organization, LocalBusiness, FAQPage, and Product where it fits.
- An llms.txt file with correct URLs, not a paste from someone else's domain.
- A robots.txt that does not accidentally ban YandexBot, GPTBot, or ClaudeBot if you want to be visible there.
- Consistency: the site, 2GIS, Yandex Business, Avito, and reviews all say the same thing.
- Mentions outside the site: industry sites, press, case studies, reviews that were not bought.
The model does not read a brand's soul. It builds an answer from what it can check. Empty or contradictory, and it picks the competitor with the cleaner signal.
E-E-A-T in AIO is not an acronym for a report. Experience: real cases, photos of the work, timelines. Expertise: who does the work, licenses. Authoritativeness: mentions beyond your own domain. Trust: the same phone, the same address, reviews without sudden jumps. On a beauty salon site that is "stylist Anna, 8 years, link to a portfolio", not "the best specialists in the city".
A common question: should you pay for ads inside AI products. For now, most commercial recommendations are built on organic signals. Ads in classic search do not replace the AIO layer. Yandex Direct without a site that bots can read is a double spend: the click is paid, and Alice's answer still sends people to the businesses the model knows better.
Why "we already have SEO" does not save you
The usual picture: search positions are tolerable, leads are thin, Alice names other companies. The SEO agency reports ranks. The owner does not see that the channel moved.
The reasons are ordinary. The copy was written for keywords, not for questions. The site has no prices and no timelines. Bots were blocked "so they would not steal". Schema is crooked, or it sits only on the homepage. The FAQ is three paragraphs of marketing. llms.txt was copied from a SaaS template with someone else's paths.
SEO and AIO overlap. They do not duplicate each other. You can rank at the top of search and be missing from AI answers if the structure for models was never assembled. The reverse happens too: a strong AIO layer sometimes pulls classic search along, because the site becomes easier to read for any consumer of the content.
Online stores lose in AI answers differently from service businesses. Stock, price, delivery, returns, and a comparison with alternatives matter there. Marketplaces often take the answer if your site has no clear product pages and no Product schema. B2B stalls on cases, timelines, and licenses. A generic page "about AI assistants" is not enough. You need facts that match how the deal actually happens.
A myth: "AI makes things up anyway, so optimization is pointless". Models do invent. Commercial recommendations are usually built on sources that repeat. If no source carries matching facts about you, the model is not inventing a version of you. It names someone else. That is an empty file on your side.
The queries where businesses lose most often: "who would you recommend", "where is it cheaper", "who is closer", "who has good reviews". That is not research traffic. The person is ready to buy. Missing one such answer costs more than a hundred fewer impressions on a long-tail keyword.
For a local business, AIO often matters more than classic SEO for leads. People do not scroll fifteen sites. They trust one voice or one paragraph in Yandex Neuro. Getting there is harder than buying a click in Yandex Direct, and over time it is cheaper if the first attempt is clean. A mistake costs more to repair than starting from an audit.
Where owners break it themselves
After a couple of webinars about "AI search", people start editing the site with no snapshot. The same mistakes follow.
Generating "unique" copy
Ten pages about "the best company in the city", with no numbers and no cases. Models cut that. Worse: trust in the whole domain drops.
Blocking every bot
"Why would we want ChatGPT on the site." They closed it, and took out a crawler they needed. Visibility on one platform goes to zero. The competitor is open the whole time.
Schema to tick a box
The wrong type, a fake rating, an office address that does not exist. It contradicts the maps. The AI assistant picks the safer source.
One llms.txt and done
The file helps. It does not replace content, listings, and reviews. More in the article on llms.txt: why one file is not enough, and how to avoid pointing a bot at the wrong pages.
Ignoring the mobile version
On mobile the site hides the price block or the FAQ inside an accordion with no text in the HTML. The bot sees an empty page. Check more than desktop, or half the AIO layer is dead.
How visibility breaks without anyone meaning it
You can do it. In practice it looks like this:
- Sudden bot blocks after the bots had already been visiting.
- Different content for people and for bots on the same page.
- Mass deletion of services with no redirects. For the model, you disappeared.
- Fake reviews next to three stars on the maps. The AI assistant picks the competitor.
- Edits with no GEO Score measurement. A month later you cannot tell what broke the answers.
The model does not take offense. It stops quoting you. Clients ask assistants and hear other names.
Measure it. Stop guessing.
AIO without numbers is a belief. You need a snapshot by platform: Alice, Yandex Neuro, ChatGPT, GigaChat, Perplexity. In Leadsy that is GEO Score and a breakdown across 56 parameters in the express audit.
A minimum ritual once a month: run the URL through the audit, ask assistants three typical queries for the niche, compare with the competitor in the report. If you are missing three times, the problem is not the "mood of the algorithm". Write the answers down word for word. The wording is a signal, not only the fact that you were mentioned.
Then a full audit with an action plan for the developer, or done-for-you implementation if there is no team. Without priorities you fix headlines while robots.txt is still blocking the site.
A workable order
If the goal is leads from AI answers, not a report that says "positions in the top":
- A free express audit: GEO Score, holes, bot access. From the homepage.
- The critical items, in order: robots.txt, NAP, FAQ, a price in HTML.
- Schema and llms.txt as a second stage, not as a substitute for content.
- A full audit with an action plan, or done-for-you implementation if you have no team.
On your own, only the first step makes sense: diagnosis. After that, without a map, you treat the symptom.
Pricing: the pricing page. Questions: FAQ.
Where to start today
Leave robots.txt alone. Do not copy someone else's llms.txt. Do not upload "AI copy" with no numbers.
Start with a snapshot. Paste the URL into the audit below. In a couple of minutes you will see what Alice, Yandex Neuro, ChatGPT, and GigaChat see, and where you lose to competitors.
A pile of critical errors is useful news: you can see why the leads were missing. The other side: every day without the fixes, the leads go to the businesses AI already recommends.
After the audit, ask out loud: "who would you recommend for [service] in [city]". Three times in a row, other brands: you are not in the recommendations. That is measurable.
"We have a site" and "we get recommended" are different things. A site without AIO is a storefront in a closed mall: it exists, and it has no clients.
Six platforms in one Leadsy report: Alice, Yandex Neuro, ChatGPT, GigaChat, Perplexity, Gemini. Block one bot and you lose a channel.
A short checklist before any edits: one phone number everywhere, prices on the site match the listing, an FAQ built from ten real calls, robots.txt does not cut YandexBot and GPTBot, schema has no fake rating. Fifteen minutes, and you can tell whether you would show the site to a colleague.
If GEO Score is green, compare it with a competitor in the niche. The gap is what matters. You at 62, the leader at 81: the answer will name them.
Alice, separately: how to get into the recommendations. When an agency promises AI promotion, ask about robots.txt, schema, and measurements before and after. Blog posts alone are the old SEO in a new wrapper.
A free audit does not force you to buy the full review. It does stop the guessing. Guessing in AIO costs more than any plan.