GEO for Business: How to Make ChatGPT and Perplexity Recommend You

What is GEO and how is it different from regular SEO

Classic SEO answers the question "how to rank in the top 10 of Google search results." GEO (Generative Engine Optimization) answers a different question: "how to make AI systems (ChatGPT, Perplexity, Gemini, Google AI Overviews) cite my brand in their answers."

The difference is fundamental. In SEO you compete for a position in a list of links. In GEO you compete for a mention in the answer itself, after which the user often doesn't click any links at all. According to Gartner forecasts, by the end of 2026 traditional search queries will drop by 25% — replaced by queries to AI systems. This means a business not optimized for AI is effectively losing a quarter of its potential organic traffic.

How AI decides who to cite in an answer

To optimize for AI, you need to understand how AI chooses sources at all. Simplified, the mechanism is this: when a user asks a question, the large language model doesn't invent the answer from scratch — it reaches out in real time to a knowledge base (this is called retrieval-augmented generation, RAG). From this base, the model extracts the most relevant fragments from various sites and synthesizes the answer.

Sources that get mentioned are those with specific characteristics: clear structure, factual density, schema markup, brand reputation across the web. AI isn't looking for "the prettiest text" — it's looking for "the easiest text to extract from." A site with structured facts, clear headings, tables and concrete numbers gets cited dozens of times more often than a site with beautiful but vague marketing copy.

Why 2026 is the tipping point for AI search

To understand the seriousness, look at the numbers. ChatGPT processes over a billion queries per week. Perplexity AI grows 40% per quarter. Google AI Overviews now appear on roughly 30–40% of queries — and that share is rising fast. Users have gotten used to getting a ready answer in 5 seconds instead of clicking through ten blue links.

What does this mean for business in practice? If clients used to go through 3–5 sites and compare in their search funnel, now they read one generated answer and make a decision based on it. Whoever is mentioned in that answer wins. Whoever isn't doesn't exist.

Most markets are still 6–12 months behind on adapting to this shift. That's both bad news (most agencies and businesses don't understand GEO yet) and good — you have a window to take positions while competitors are still asleep. In a year that window closes, and catching up will cost many times more.

Seven GEO optimization steps you can start tomorrow

1. Check whether AI sees you at all

This is the very first and most important step that almost no one does. Open ChatGPT, Perplexity and Google AI Overviews one by one. Ask 8–10 questions your business should be answering: "best web development agencies," "where to order a turnkey website in Kyiv," "Shopify vs Хорошоп for small business," "top digital agencies 2026." Note whether they mention your brand, in what context, and who's mentioned instead.

If they don't mention you — you have not one but three problems: AI doesn't know you, your competitors are already there, and clients are going to them. If they do mention you but in a negative or neutral tone — that's also a signal, because tone affects user decisions.

Build a table: query, system you checked, whether you were mentioned, which competitors were mentioned, what tone. This is your starting benchmark, against which you'll measure progress at 3 and 6 months. Without this table, you won't understand whether GEO is working at all.

2. Implement schema markup on all key pages

Schema is structured JSON-LD markup in your page code that explains to AI what's on the page: company, product, service, article, review, price, contact. Without it, AI sees your site as continuous text and has to guess. With schema — as a structured database from which it's easy to extract facts and use in answers.

Minimum to implement for a business site: Organization (who you are), LocalBusiness (if you have a physical location), Service (on service pages), Article (on blog articles), FAQPage (on FAQ blocks), BreadcrumbList (breadcrumbs), Review or AggregateRating (reviews). For e-commerce, add Product, Offer, ItemList.

You can verify implementation through Google's official Rich Results Test or Schema Markup Validator. An error in one field — and the entire markup is ignored. This isn't work where you can improvise: do it right or don't do it at all.

3. Rewrite content in "AI-friendly" format

Imagine your text is being read not by a person but by a machine looking for short factual answers to specific questions. How do you write for it? Clear H2 and H3 subheadings that frame exactly the question users ask. Specific numbers instead of general words — write "raised conversion by 34% in 8 weeks," not "significantly improved results." Short direct answers in the first sentences of each paragraph — AI often pulls exactly the first sentence.

Structure matters more than text beauty. AI loves FAQ blocks at the end of pages — these are ready-made question-answer pairs perfect for citation. Lists and tables for comparisons work better than prose: when AI sees a structured list, it cites it 3–4 times more often.

Small but important detail: write so that your sentence makes sense out of context. AI often takes one sentence and inserts it into an answer — and if your sentence starts with "This means that..." without explaining what "this" is, no one will use it.

4. Create expert content with real data

AI hates "general fluff" — text that repeats what's already been said on a hundred other sites. This is called "information gain" — how much new information your page adds compared to existing ones. If your article "10 tips for choosing an agency" repeats what 100 others say — it simply won't be noticed.

What AI loves: your cases with specific numbers ("client X, niche Y, was 12 leads/month, became 47 in 90 days"), your internal market research, original analytics, expert opinions with real author names. One case with real numbers gets cited more often than twenty articles in "top 10 tips" format.

Practical approach: once a quarter publish "data from the field." Collected stats on 50 launched projects — publish insights. Surveyed 200 clients — publish results. Noticed mobile conversion in Shopify e-commerce averages 1.8% — publish that fact. This creates unique content that becomes a reference source for AI for years.

5. Build brand presence beyond your own site

AI evaluates not only what's on your site but what's said about you across the web. This is the "corporate echo" principle: the more mentions from various independent sources — the higher the probability AI recognizes you as an authority. Simply put: one site isn't enough.

What goes into the citation ecosystem: optimized Google Business profile, registration in industry catalogs, regular guest publications on industry media, LinkedIn activity with posts from real employees, conference talks with mentions in programs, client reviews on third-party platforms.

Especially important: AI trusts "three-dimensional" mentions much more — when your brand is mentioned in context of something specific. Not just "Lúmi is a good agency" but "Lúmi launched a site for brand X that increased conversion by 40%." The first is empty mention, the second is factual mention. Their value differs by orders of magnitude.

6. Optimize for E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness)

This isn't a new concept — Google has been using E-E-A-T for years to assess content quality. But in 2026 it became critical for AI: models actively check who exactly the content author is and whether they have the right to speak on the topic. Anonymous articles in 2026 are practically never cited — AI can't verify the source and refuses to reference it.

What to do: every key page should have a real author with real experience. That means a separate author page with bio, LinkedIn link, list of projects, publications, talks. It should be clearly visible how many years the person has been in the field, what projects they've worked on, what certifications they hold.

For an agency this means the "About" page is no longer a formality. It's a full document with team photos, bios, roles, specific experience. Each team member who writes for the blog has their own author page with all their articles. This isn't cosmetic — it's a factor AI evaluates mechanically through structured data.

7. Track AI visibility metrics

Classic metrics like Google rankings still matter, but they don't show your AI visibility. This is a new parallel reality requiring new metrics. What to measure: citation frequency (how often you're mentioned in AI answers per period), share of voice (your share among all niche mentions), sentiment (mention tone), recommendation rate (how often AI specifically recommends rather than just mentions you).

Monitoring tools: Profound, Otterly, Semrush AIO, Ahrefs Brand Radar, Brandwatch. Budget — from $50 to $500 per month depending on scope.

No budget for paid tools? Do manual audits weekly. Write down 10 key niche queries, run them through ChatGPT, Perplexity and Gemini, log in Google Sheets. After 3 months you'll have dynamics, after 6 — full picture of what works.

Real example: what a GEO-optimized page looks like

To not leave everything at the theory level, let's break down how a GEO page differs from a regular marketing page. Imagine you have a service page for "Online Store Development."

The marketing version usually starts with a beautiful banner, then vague phrases like "we create powerful solutions for your business," "individual approach," "team of professionals." Ends with an "Order" button. AI won't extract a single fact from such a page because there are none.

The GEO-optimized version of the same section: clear H1 "Turnkey Online Store Development — 2026," then H2 "How Much It Costs" followed by a table with three packages and concrete dollar prices. Then H2 "What Platform To Build On" with a comparison of Shopify, Хорошоп, OpenCart and facts on each. Then H2 "How Long Development Takes" with concrete timelines per stage. At the bottom, FAQ with 8 questions and direct answers. Schema markup for Service, Offer, FAQPage. Page author — a real person with a LinkedIn link.

The first version looks great to a human but is invisible to AI. The second looks less "magical" but it's the version AI will cite in answers to "how much does an online store cost" or "Shopify vs Хорошоп."

Specific mistakes that kill your AI visibility

Mistake #1 — "Perfect" marketing site with no real substance

Pages like "we're the best, we're reliable, we're unique" are never cited by AI because they contain not a single fact. Marketing language is GEO's worst enemy. Every sentence on the site should carry either a fact, a number, or a concrete example. If a sentence can be plugged into a competitor's site without changing anything — it's empty and needs rewriting.

The "fluff test": take any paragraph from your site, replace your company name with a competitor's — if the text remains just as true, it's empty.

Mistake #2 — No blog or a blog with 300-word articles

AI cites deep materials of 1,500+ words with concrete data. Surface-level 300–500 word articles don't give the model enough context and facts to extract anything useful. If your blog is SEO filler written by a copywriter for $5/article — it's not an asset, it's a liability.

Better one 3,000-word article per week than five superficial ones in the same time. Quality > quantity, especially in the AI era where shallow content turns into "gray sludge" that AI ignores.

Mistake #3 — Site closed to AI crawlers via robots.txt

Check yourdomain.com/robots.txt — is GPTBot, ClaudeBot, PerplexityBot, Google-Extended, ChatGPT-User allowed to read your site. Many tech teams block all new bots by default "just in case" — meaning you're literally invisible to AI.

The paradox: some sites block AI bots fearing their content will be "stolen for training models," and simultaneously complain AI doesn't cite them. It's like locking the door and wondering why no one visits.

Mistake #4 — Complete absence of external mentions

If no one writes about you anywhere except your own site, AI has no reason to trust you. This is the "social verification" principle: AI trusts those confirmed by independent sources. A site is a monologue. External mentions are a dialogue where others confirm your expertise.

Mistake #5 — Stale content

AI prioritizes freshness: a 2022 article has almost no chance against a 2026 article on the same topic. This means it's not enough to write a good article once — you need to regularly update it, add fresh data, refresh examples.

Practical approach: every 6 months, go through the top 10 blog articles, update publication date, add 2–3 new paragraphs with fresh data, fix outdated information. This costs minimal time but gives huge AI visibility boost.

How to build a 90-day GEO strategy: the roadmap

Theory is good but you need a concrete action plan. Here's the real sequence we use at Lúmi for our clients.

Month 1: foundation and audit

The first 30 days are basic infrastructure without which everything else is pointless. Week 1: audit current AI visibility (10–15 key queries through ChatGPT, Perplexity, Gemini), robots analysis, AI bot accessibility check. Week 2: implement schema markup on home, service pages, contacts. Week 3: rewrite key pages in AI format — H2 for questions, FAQ blocks, concrete numbers. Week 4: create author pages, bios, team profiles for E-E-A-T.

Month 2: content and expertise

Days 30–60 — creating content AI wants to cite. Launch a blog with 2 articles per week, each minimum 1,500 words with concrete data and cases. In parallel — publish 3–4 expert pieces on third-party resources. Register in all relevant niche catalogs. Collect and publish client reviews on third-party platforms.

Month 3: scaling and measurement

Third month is when work starts paying off and you need to measure correctly. Set up AI visibility monitoring (Profound, Otterly or manual tracking). Analyze first mentions: where they appeared, in what context, what's being cited. Reinforce pages already starting to be cited. Launch guest publications on key platforms. First public case with "before/after" numbers.

Realistic 90-day expectations: appearance in Perplexity answers (the fastest system), partial appearance in Google AI Overviews, not yet in ChatGPT (their model updates slower). Stable mentions across all three systems — that's a 6–9 month horizon.

How long until GEO starts working

Realistic horizon — 3 to 6 months until first stable results. First mentions in Perplexity may appear within 2–3 weeks of optimization because that system indexes new sources quickly. ChatGPT and Google AI Overviews are slower because their models update less often.

Frequently asked questions about GEO