How AI Is Changing Search — and What It Means for Your Business

A Revolution Most Companies Are Missing

Let's say this plainly before the rest of the article: for us, this isn't a conference-slide theory. This entire blog — including the article you're reading right now — is built so that ARDURA exists as a recognizable entity in AI answers, not just another result in Google. It's our own strategy, before we ever pitch it to anyone else.

In 2024, ChatGPT surpassed 2 billion monthly visits. Perplexity — 100 million. Gemini is built into every Android phone. Microsoft Copilot — into every Windows.

People no longer type "best SEO agency Warsaw" into Google. They say to their phone: "Recommend me an agency that will build a website and handle SEO in Poland." And they get an answer — without clicking a single link.

If your company isn't in those answers, you're losing customers. Every day. And you don't even know it, because you won't see it in Google Analytics — which is one reason the first step in our GEO audits is always manually asking AI about a given company, not staring at dashboards that never see that traffic at all.

---

The Old Model: 10 Blue Links

This is what search looked like for 25 years:

1. User types a query
2. Google returns a list of links
3. User clicks on 3–5 links
4. Compares websites
5. Contacts the chosen company

In this model, SEO = fighting for the top 3 positions. Simple. Measurable. Predictable.

The New Model: AI Answers Directly

This is what search looks like in 2026:

1. User asks AI (by voice or text)
2. AI analyzes hundreds of sources in a fraction of a second
3. AI generates a response — one, specific answer
4. User gets a recommendation without clicking any links
5. Or clicks a link to one, selected source

The key change: The user doesn't browse 10 results. They get 1–3 recommendations. Either you're in those 1–3, or you don't exist. That line sounds harsh, but it's exactly why we look at a weak Google ranking differently in our audits than we did a couple of years ago — back then it meant weak but measurable traffic; today it means zero chance of AI even considering the company at all.

---

What Questions Do People Ask AI?

Purchase questions (bottom of funnel)

• "Which SEO agency in Poland is the best?"
• "Recommend a company that builds fast websites in React"
• "How much does website positioning cost?"
• "Which technology is better — WordPress or Next.js?"

Educational questions (top of funnel)

• "What is GEO?"
• "How does SEO work?"
• "What's the difference between SEO and SEM?"
• "How do I check if my website is optimized?"

Comparative questions (middle of funnel)

• "Compare Next.js vs WordPress for SEO"
• "What are the pros and cons of SEO?"
• "GEO or SEO — which is better for a B2B company?"

Note: AI prefers answering specific questions. A page with a clear Q&A structure has a higher chance of being cited than a page with continuous text — which is why every article on this blog, this one included, carries an FAQ section built from real questions, not bolted on for the sake of the format.

---

How Does AI Choose Whom to Cite?

This is the key question — and it's exactly the list we run through ourselves when we build ARDURA's own visibility, before we ever pitch it to a client. There's no public "AI algorithm" like PageRank for Google. But from research and observations, we know what works.

1. Source authority

AI prefers sites with high authority — lots of backlinks, citations, presence on credible platforms. This overlaps with SEO, but in our approach authority doesn't end at links — it also means being present as a recognizable entity in the places AI trusts.

What to do: Build links, publish on industry portals, be cited by media.

2. Content quality and depth

AI cites content that exhaustively covers a topic. Not a 300-word article with generalities. AI looks for articles with specific numbers, research, comparisons, tables.

What to do: Write articles of 2,000+ words. Use data, statistics, case studies. Not "it's good to have a website" — but "a website converts 2.3% more with every 100ms improvement in LCP (Google study 2024)." We hold ourselves to that same rule on this blog — hence the tables and specific numbers instead of generalities in every article, not just this one.

3. Content structure

AI models parse text structurally. Headings, lists, tables, definitions — these are elements that AI easily "understands" and cites.

What to do:

• Headings as questions (H2: "How much does SEO cost?")
• Answer in the first sentence after the heading
• Numbered and bulleted lists
• Comparison tables
• Definitions (bold term + explanation)

4. Structured data (Schema.org)

Schema.org helps AI understand page context. FAQPage, Article, HowTo, Organization — these are signals that AI uses for information extraction.

What to do: Implement Schema.org on every page. FAQ schema is a must-have — AI loves questions with answers in JSON-LD format. We have this implemented consistently across every page type on our own site — it's one of the first things we check on a client's site too.

5. Freshness

AI prefers up-to-date content. An article from 2021 about "SEO trends" won't be cited in 2026. Publication date and regular updates matter.

What to do: Update articles every 3–6 months. Add dates to frontmatter. Remove outdated information. This article carries its own update date for the same reason — GEO isn't something you "implement once."

6. Uniqueness

AI has access to millions of articles. If yours says the same thing as 1,000 others — why would it cite yours specifically? This is, at bottom, the core of our approach to entity: it's not about repeating the definition of GEO the same way as the tenth other agency, it's about giving AI a reason to point at you specifically.

What to do: Add unique perspectives, your own data, original case studies. "Based on our 200 projects, the most common SEO problem is..." > "SEO is important for business."

---

What Is GEO? (Generative Engine Optimization)

GEO is the systematic optimization of content for visibility in generative AI. An analogy to SEO, but for a different "search engine." At ARDURA we don't treat it as an add-on to SEO — it's a separate axis of our work, because visibility in Google and visibility in AI answers are two different questions that need two different answers. The six pillars below aren't a theoretical checklist — the technical foundations (AI-crawler access, Schema.org, llms.txt, Q&A-formatted content) are what we implemented on our own site first, before we ever proposed them to a client; recognizable entity status in AI's eyes is something we're still actively building.

GEO in practice — 6 pillars

1. Crawlability for AI bots AI models send crawlers (GPTBot, ClaudeBot, PerplexityBot) to fetch data from the internet. If you block them in robots.txt — AI can't see you. Our own robots.txt lets these crawlers in deliberately — it's the first thing we check in a client audit too, because it turns up blocked, unbeknownst to anyone, surprisingly often.

# robots.txt — let AI bots in
User-agent: GPTBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: PerplexityBot
Allow: /

2. llms.txt A new standard — a /llms.txt file (like robots.txt, but for AI). It describes the company, services, and contact data in a format understandable to AI. We have our own llms.txt (and an extended llms-full.txt) — if you don't have one yet, it's the simplest, one-time step on this entire list.

3. Content in Q&A format AI generates answers to questions. Content in question → answer format is a natural source of citations.

4. Extended Schema.org FAQPage, HowTo, Article with complete data. AI parses JSON-LD faster than raw HTML.

5. Unique data and statistics AI looks for citable data. "Average website development time is 3–6 weeks (based on 150 of our projects)" > "Website development takes several weeks."

6. Entity building Build a recognizable "entity" — a company that AI knows. Presence on Wikipedia, LinkedIn, industry portals, business directories. The more sources mention your company, the higher the chance of citation. For us this is the starting point of the whole strategy, not an add-on at the bottom of the list — and it's this pillar we're actively building for ourselves right now, rather than claiming it's already done: if AI doesn't recognize you as an entity, no single article will reverse that.

---

How to Check If AI Cites You?

Manual test (5 minutes)

1. Open ChatGPT, Gemini, and Perplexity
2. Ask questions that your company should be the answer to:
   • "Recommend a [your service] agency in Poland"
   • "What is [your service]?"
   • "Best company for [your service]"
3. Check if your company is mentioned

Automated monitoring

• Perplexity provides sources — check if your domain appears
• Tools like Brandwatch, Mention — monitoring mentions in AI
• Google Alerts for company name + industry keywords

Typical result of the first test: 90% of companies are NOT cited by AI. This is your opportunity — while the competition sleeps. In our audits, this manual test is the first step, before we even look at Search Console — because it shows whether a company exists in AI's eyes at all.

If you want to know how to specifically optimize content for individual AI platforms, read our guides on optimizing for ChatGPT Search and positioning in Perplexity.

---

SEO + GEO Strategy: A Practical Plan

This plan isn't an abstract template — it's the same order of work we follow ourselves and propose to clients, because the foundations have to be in place before it makes sense to write content for AI.

Month 1: Foundations

• Audit of current visibility (Google + AI)
• robots.txt — let AI bots in
• llms.txt — describe your company
• Schema.org — Organization, FAQ, BreadcrumbList
• Content audit — what you already have, what to convert for GEO

Months 2–3: Content

• 8–12 articles in Q&A format (blog)
• Each article: 2,000+ words, data, tables, unique perspective
• Internal linking between articles
• Update existing content for GEO

Months 4–6: Authority

• Guest posts on industry portals
• Media citations (HARO, Help a Reporter)
• LinkedIn presence (expert posts)
• Case studies with numbers

Ongoing monitoring

• Monthly AI test (manual)
• Google Search Console — ranking tracking
• Analysis of new questions and trends in AI

---

What's Next for Google?

Google isn't dying. In 2026, it still handles 85%+ of searches. But it's changing — AI Overview (formerly SGE) displays generated answers above traditional results.

The effect: even in Google, fewer clicks go to organic results. "Zero-click searches" are rising — users get answers without clicking.

Our own defensive strategy — and the one we put in front of clients:

1. SEO — still the foundation (backlinks, technical SEO, content)
2. GEO — new growth channel (AI citations, entity building)
3. Brand — the stronger the brand, the more often AI recommends it

Companies that ignore GEO won't die tomorrow. But in 2–3 years, they'll be wondering why traffic is declining despite good SEO. We'd rather be the company that asked itself that question early — which is why it's ARDURA, not just our clients, that has the technical GEO foundations described above in place and is actively building its own entity recognition, rather than waiting for someone else to do it.

---

Want to know how your company looks in the eyes of AI? Request a free GEO analysis — we'll check whether ChatGPT, Gemini, and Perplexity cite your business, with the same manual test described above.

Read next: this article is part of our complete guide SEO, GEO and AEO in 2026 — the full picture of the new search era: hard data, GEO/AEO tactics, technical SEO and off-page.