How to Get Into ChatGPT and AI Engine Answers
Picture this: a potential customer asks ChatGPT "which dentist should I choose in my area" or "where can I order a kitchen, affordable and good quality". The AI returns a list of three or four companies. Links to their sites are there — but the user already has the answer and probably won't click through. They just note the name and look up the contacts.
Your company isn't on that list. Even though your classic SEO is excellent — top-5 rankings, keywords, meta tags. ChatGPT simply doesn't know you exist.
This isn't hypothetical — it's the reality of June 2026. According to Similarweb, AI services already send over a billion referrals to websites per month, growing by hundreds of percent year over year. And AI Overviews research shows Google now displays an AI answer for a large share of queries. Classic search traffic is falling, AI traffic is rising. The old SEO rules only work partway in this new reality.
Below is a concrete plan. No fluff, no "hire us". Only what actually works and what you can start doing today. This one focuses on ChatGPT and similar engines; for the broader picture, see how to get into AI answers in general.
Why Your Site Doesn't Appear in AI Answers — Even If It Ranks #1
A typical situation: the site ranks top-3 for every key query, meta tags are fine, content is unique. Yet ChatGPT, Perplexity and AI search skip it and cite competitors.
The reason isn't that AI is "biased". It's that classic SEO and AI optimization are two different processes. They overlap, but they're not the same.
A search engine evaluates a page for relevance: keywords, a good title, load speed, backlinks. An AI engine looks differently — it needs to extract a fact. It scans the content and asks: can I take a ready answer from here, is there a clear structure, are the author, organization and date marked, can this source be trusted?
If the page reads like a promotional pitch listing advantages — the AI skips it. Like a reference article with no author or date — skips it. Like a wall of text with no subheadings — skips it.
GEO and AEO: Two Tools That Solve This
The SEO industry has talked a lot about GEO and AEO for two years. Without the marketing fog, the difference is this:
| Approach | What it optimizes for | Goal |
|---|---|---|
| GEO (Generative Engine Optimization) | ChatGPT, Perplexity, Gemini, Claude, AI search | Appear in answers the AI generates |
| AEO (Answer Engine Optimization) | Voice assistants (Siri, Alexa), featured snippets, "People also ask" | Become the source of short, direct answers |
In practice they go together: if you've done GEO, AEO usually follows. See our dedicated guide to AEO optimization for more. So from here on we talk about GEO as the general approach.
How AI Decides Whom to Cite
AI engines have no public "ranking algorithm" like search engines do. But two years of observation reveal patterns. An AI engine works with two sources:
| Source | How it works |
|---|---|
| Training data | The huge body of internet text the model learned from. If your site existed before 2024 and was indexed, it's somewhere in the model's "memory". But the model remembers patterns, not specific sites — so that alone isn't enough. |
| Real-time search | ChatGPT Search, Perplexity and AI search connect to search engines. They scan several sources and look for consensus: if three independent sites describe a service similarly, that becomes a strong signal. |
The practical takeaway: ranking #1 is half the job. The other half is making sure that once the AI reaches your page, it can read it and extract an answer — and that requires preparing the page differently than for a search engine.
The Step-by-Step Plan: 7 Actions That Actually Work
Steps in descending order of importance. The first three are the foundation; the rest amplify the result.
Step 1. Add Schema.org Structured Data
The single most impactful action. Structured data is still used by far from every site — per Web Almanac, a large share of pages have none or incomplete markup. Adding it puts you ahead of most competitors automatically.
Schema.org is the "language" you use to tell the AI what's on the page: not just "text about dentistry", but "a medical organization — here's the name, address, phone, hours, services with prices, the doctor, and FAQs with answers".
The minimum every page needs
- Organization — name, logo, contacts, social links. The anchor: the AI understands who you are.
- WebPage — title, description, language, publication date, breadcrumbs.
- FAQPage — questions and answers. The most powerful format: the AI sees the question → sees a ready answer → cites it.
Markup for specific page types
| Page type | Suitable schema |
|---|---|
| Products | Product — price, availability, description |
| Services | Service — service area, price |
| Articles | Article — author, date, image |
| Guides | HowTo — step-by-step process |
| Contacts | ContactPage |
The format is JSON-LD, placed in the page head as a script of type application/ld+json. Don't hand-write the code — use generators. After adding it, always validate the markup with the Schema.org validator: one missing comma and the AI ignores the whole schema.
Step 2. Add E-E-A-T Signals
E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) isn't just a buzzword from Google guidelines — for AI it's literally the "can this source be trusted" test. What to do:
- An About page with real information: who you are, what you do, since when, who founded it. Not "a dynamically growing company", but "operating since 2016, a team of 12, custom kitchens in the region".
- Contacts: phone, email, physical address — visible on the page, not only in schema.
- Article authors: every post signed by a real person with a short bio ("Peter Ivanov, CEO, 9 years in SEO").
- Publication and update dates: a page with no date, or a three-year-old one, earns less trust.
- Links to authoritative external sources: cite a study — link to the original. AI engines see the link graph.
Step 3. Rewrite Content for Fact Extraction
The hardest and most underrated step. Most commercial copy is written as "we're the best, here are 10 reasons to choose us". That works for humans, not for AI: it looks for facts it can extract and drop into an answer, not emotions. Look at your page: what concrete fact could the AI take from it?
| Text variant | What the AI extracts |
|---|---|
| "A wide range of dental services at affordable prices. Experienced doctors, modern equipment, individual approach. Book today!" | Nothing. Zero facts. |
| "DentPro clinic. Operating since 2014. Open 9:00–21:00 daily. Cavity treatment from $60, implants from $470, cleaning $48. Insurance accepted. Book online or by phone." | Name, hours, three services with prices, insurance, contacts — almost a ready answer. |
Concrete techniques
- Every paragraph should contain at least one extractable fact: a number, date, price, address, name.
- AI reads lists better than solid text — they're structured by definition.
- Tables are even better: a comparison in a table can be cited almost verbatim.
- Answers to concrete questions ("How much?", "How long?", "Where are you?", "How to book?") belong at the top of the page, not the bottom.
Step 4. Sort Out the Technical Side
AI engines pull data through search engines and their own crawlers, so technical accessibility for bots is critical.
robots.txt
Make sure the file doesn't block AI crawlers. If it disallows GPTBot or PerplexityBot, you've shut the door yourself. Add explicit allow rules for them.
SSL certificate
Must be active and valid. AI engines, like search engines, lower trust in sites without HTTPS.
llms.txt file
The new llms.txt standard proposed in 2025 is the robots.txt equivalent for language models. It lists your key pages with short descriptions, giving the AI a "map" of your content. Not yet mandatory, but early adopters already gain an edge.
sitemap.xml
Should contain accurate lastmod dates in valid ISO 8601 format. AI engines use the sitemap to understand which pages are the freshest.
Step 5. Structure Headings by Meaning, Not Keywords
AI reads H1–H6 headings as a table of contents. An H2 like "Benefits of working with us" is flagged as marketing and likely skipped. An H2 like "How much does a dental implant cost in 2026" is a concrete question with a likely concrete answer — and it gets read.
Hierarchy matters too: H1 → H2 → H3 should unfold the topic logically, like chapters in a book. Don't skip levels.
Step 6. Build the Brand's Information Footprint
AI looks for consensus across sources. If only your own site talks about your company, the signal is weak. If industry media cite you, you're in relevant directories, and your founder comments on market news — the signal multiplies. What works:
- Publications in industry media and on relevant platforms.
- An up-to-date company listing on maps and local directories.
- A verified business profile.
- Conference talks — even small ones — that get mentioned online.
- Customer reviews on independent platforms.
Step 7. Check the Result — More Than Once
Once everything is set up, verify whether the AI sees your site. Not by eye — with concrete tools.
Free GEO / AEO audit
It scans the page, checks Schema.org, E-E-A-T, OpenGraph, FAQPage and HowTo — and shows what's wrong and what to fix first. 30 seconds, no sign-up. You immediately see the gaps from the AI's point of view, not classic SEO.
Check whether AI can see your site
A free GEO / AEO audit — Schema.org, E-E-A-T, FAQPage, HowTo. No sign-up, results in 30 seconds.
Manual check
Open ChatGPT Search or Perplexity and ask the questions your customers actually ask: "Which [service] should I choose in [city]", "How much does [product] cost in 2026". See whether your company is mentioned.
Regular monitoring
Do this monthly. AI engines update their sources constantly: a site that wasn't cited yesterday can appear in answers today — if you did everything right.
What Doesn't Work — Don't Waste Time On It
Three myths I hear regularly:
Myth 1. "Just generate lots of content with AI"
AI engines recognize templated AI text and lower their trust in it. They look for unique facts, and generated content by definition contains nothing unique.
Myth 2. "GEO is just SEO plus schema.org"
Not enough. Schema is the foundation, but without rewriting content for fact extraction (step 3) and E-E-A-T signals (step 2), the schema stands on empty ground.
Myth 3. "Just buy backlinks"
AI engines don't use the link graph the way search engines do. Consensus across sources and intrinsic content quality matter more.
What to Do Today
A short checklist for this evening (a full point-by-point self-check is in a separate checklist):
- Run a GEO / AEO audit of your site — see which Schema.org markup is missing.
- Check robots.txt — make sure AI crawlers aren't blocked.
- Open your homepage and ask: what single concrete fact can the AI extract from the first paragraph? If "none" — rewrite it.
- Add an About page with real information and contacts if you don't have one.
- Manually ask ChatGPT about your service in your city — see who's in the answers now.
This isn't a six-month marathon. The first four steps take one evening. And the result — appearing in AI answers — pays them back many times over.
Frequently asked questions
Why doesn't a top-ranking site appear in ChatGPT answers?
Classic SEO and AI optimization are different processes. Search engines judge relevance; AI extracts facts and checks structure, authorship and source trust. Without Schema.org markup, E-E-A-T signals and content written for fact extraction, ChatGPT doesn't treat the page as a usable answer.
What's the difference between GEO and AEO?
GEO (Generative Engine Optimization) optimizes a site for generative AI — ChatGPT, Perplexity, Gemini. AEO (Answer Engine Optimization) targets voice assistants and featured snippets. In practice they're done together: GEO usually brings AEO along.
What markup is needed to appear in AI answers?
At minimum Organization, WebPage and FAQPage in JSON-LD. Add Product for products, Service for services, Article for articles, HowTo for guides. After implementing, always validate with the Schema.org validator.
How long until AI starts citing a site?
AI engines update their sources constantly. After adding markup and rewriting content, first mentions usually appear within a few weeks — which is why you re-check monthly.
Does AI-generated content help you get into answers?
No. Templated AI text lowers trust: AI looks for unique facts, and generated content doesn't contain them. Structure, specifics and E-E-A-T signals matter more.
Check Your Site Right Now
Free GEO / AEO audit: JSON-LD, E-E-A-T, schema.org — no sign-up, takes 30 seconds.