AI search visibility

What has changed

Classic search handed you ten blue links and left the choice to the user. An AI assistant makes the choice before the user does: it reads several sources, writes the answer itself and usually cites only a handful of pages. Instead of ten slots there are now three or four. That makes visibility more competitive — but it also means a well-structured smaller site can be cited alongside a large brand, because the assistant is not measuring your ad budget, it is measuring how clearly and credibly you answer a specific question.

The second difference: an assistant answers a question, not a keyword. In classic SEO you optimise a page for the phrase “website maintenance price”. In AI search the page needs a clear, readable answer to “how much does website maintenance cost and what drives the price”. Same topic, a completely different shape of content. Stuffing old pages with keywords does not help — the content has to be rewritten so the answer is right there at the top.

We do not claim how big this channel is in percentage terms — nobody knows that reliably and the numbers shift month to month. What we do see with clients is simpler: enquiries that start with “ChatGPT recommended you” tend to have a shorter sales cycle, because the person arrives already holding a recommendation. That alone makes the channel worth working on, even when its exact size cannot be measured.

Why most sites are invisible to AI

The reasons are almost always technical or editorial, not the fault of some algorithm. Most of them can be fixed within a single project, and they repeat from site to site so consistently that we have turned checking them into a fixed list in our audit.

• Content is rendered only by JavaScript. Google usually copes, but many AI crawlers do not execute JavaScript at all — they see an empty page.
• There is no structured data. Without Organization, Service and FAQPage markup a machine has to guess the company facts out of prose.
• The pages contain no concrete answers. When every page opens with a slogan, there is nothing to quote.
• Company facts are inconsistent across the web: a different legal name, an old address, a dead phone number, a service list that does not match the site.
• robots.txt blocks AI crawlers — often by accident, because somebody once copied another site’s file.
• Content hides in PDFs, images or behind a login where no crawler can read it.

What we do

This is a list of real work, not a service package on paper. We start with an audit that looks at the site through a machine’s eyes: what a crawler actually sees, which facts are structured, which questions go unanswered. Then we make the fixes in an agreed order and show what changed at each step.

• Crawler readability: we move content to server-rendered or prerendered form so the text exists without JavaScript. On most sites this is the single biggest win.
• llms.txt: we publish a machine-readable summary of the company, its services and its common questions, so an assistant has one clear source.
• Structured data: Organization, Service, FAQPage and Article markup, validated and consistent with the actual content.
• Answer-first pages: we rewrite key pages so the direct answer sits in the first two sentences and the reasoning follows.
• Consistent facts: name, address, phone, services and descriptions are aligned across the site, the business register, directories and profiles.
• robots.txt cleanup: we allow the relevant AI crawlers and exclude the ones the client does not want — a deliberate decision, not an accident.
• Baseline mapping: we ask the assistants how they currently describe your company and your competitors, and we write the answers down.

How we measure it

Before the work starts we pick 10–20 questions together with the client — the questions their customer would genuinely ask. For example “who maintains WordPress in Tallinn”, “what is a good website maintenance service for an online shop”, “who can fix an old website”. We run them through ChatGPT, Perplexity and Google AI Overviews and record the answers and the sources cited.

The same measurement is repeated every month. You can see in black and white whether the company is being named, whether the description has become more accurate, and whether the source is your own site or somebody else writing about you. The result is a shift, not a score — which is exactly why the baseline has to be recorded before the work begins.

We do not promise a position or a guaranteed mention. The assistants change models and source selection constantly, and no honest provider can promise a slot in an answer they do not control. We promise properly executed work and a measurable baseline, so you can decide for yourself whether to continue.

Paid placements inside AI assistants

Paid advertising alongside AI assistant answers is still emerging: what is available differs by market and platform and changes quickly. We do not sell it as a separate product and we do not promise outcomes. Where such a placement exists in a client’s market and fits their goals, we set it up as part of the wider visibility work and measure it with the same logic as every other channel.

How this relates to classic SEO

AI visibility does not replace SEO, it stands on top of it. Assistants draw largely on the same indexes and sources as ordinary search: a fast, indexable, well-structured site is a prerequisite for both. If the technical SEO is missing, even the best-written answer will not help — the crawler never reaches the page.

In practice we usually run the two together: the technical base and keyword work under the SEO service, the answer and structured-data side from this page. If the site is technically slow or broken we start with website optimisation, because otherwise the work is wasted.

If you also want AI on the site itself — a chatbot, smarter search or a content assistant — that is a separate service: see AI services for your website. In short: this page is about being found in AI, that page is about what AI does on your site.