AI search in 2026: how a small business gets found in ChatGPT, Google AI and Perplexity

Search now answers for itself

Search used to hand you a list of links and you competed for the click. Now both Google and ChatGPT increasingly give a finished answer first and tuck the links underneath. Pew Research Center tracked real search sessions in 2025: when an AI summary sits on top, people click a regular result link about 8% of the time versus 15% without one, and click a source inside the summary only about 1% of the time.

The lesson isn't 'SEO is dead' — it's that the rules moved. Some customers get their answer without opening a site. So the job is no longer only to win the click, but to land inside the answer as a source the AI trusts.

Why it matters even for a small company

AI search looks like a big-brand story, but the traffic is already real. By Semrush's count, referral visits from ChatGPT grew about 206% year over year (January 2025 to January 2026). People aren't only chatting with AI — they act on its answers when they buy.

For a local business that's an opening. When someone asks AI 'where can I get custom embroidery in Tallinn' or 'a good personal trainer nearby', a small but clearly described company can land in the answer — if its site is readable to both the machine and the human.

How AI search picks who to cite

Nobody publishes the exact formula, but ChatGPT, Perplexity and Google AI reward the same things:

• A direct answer near the top. A page that answers 'how much does a website cost' in its first paragraph is easier to quote than one that buries it halfway down.
• Clear structure: one H1, logical subheadings, short paragraphs. The machine finds the right chunk faster.
• Specifics and numbers. Facts, price ranges, timelines and sources get cited more readily than vague claims.
• Reputation beyond your site. AI checks whether others mention you — reviews, directories, articles. The more consistent the picture, the more confidently it names you.
• Freshness. A current 'last updated' date and recent data beat a 2023 page on the same topic.

What to fix on your site now

The good news: most of this is ordinary site hygiene done properly. The concrete steps:

• Answer the page's main question right away, in the first lines, then expand into detail.
• Tidy your headings: one H1, meaningful subheadings for sub-questions, and a short FAQ block with crisp answers.
• Add structured data (schema): Article, FAQPage, Organization — it helps the machine understand the page.
• Don't block AI crawlers in robots.txt (OAI-SearchBot for ChatGPT, PerplexityBot, Claude-SearchBot) and stay indexable in Google — otherwise you're simply invisible.
• Speed and mobile. A slow, heavy page gets indexed and cited less often.
• Show an update date and keep the facts current.

The Estonian edge — three languages

In Estonia customers search in Estonian, Russian and English. AI search answers in the language of the question and looks for a source per language. A site that answers clearly in all three can become the answer where a competitor only has an Estonian version.

That's the underrated move for a local business: not one site 'for everyone', but three clear versions, each able to land in the AI answer for its audience.

A real example: Apex Steel

For steel-structures maker Apex Steel we rebuilt the corporate site: a clear structure from the first screen, SEO-ready, with a 0.9-second first-screen load. The result — top 3 in niche search, +90% time on site and ×2.5 requests from the site.

The same things that lifted it in ordinary search — fast loading, clear structure, real numbers in the project pages — are exactly what AI search likes. A solid foundation works on both fronts at once.

What not to expect (honest caveats)

You won't get full control. You don't decide what the AI shows, and no one guarantees the same result across every platform. This is influence, not a control panel.

Clicks from informational queries will drop — some people get the answer and don't visit. But those who do click are usually closer to buying: they've read the answer and come to check you specifically. So measure differently — not only clicks, but brand mentions in AI answers and the quality of incoming enquiries.

Schema alone isn't enough without off-site reputation. Reviews, directories, mentions in articles are how AI cross-checks whether to trust you. That's long-term work, not a one-time setting. If you first want to understand the wider AI shift happening in 2026, start with our breakdown of agentic AI (link below).