How ChatGPT and Gemini “see” your brand: an AEO / GEO guide

What AEO and GEO are

Both terms come from the marketing industry and are not official standards. Still, they describe a real shift: search engines no longer only return links, they write an answer to the question itself.

AEO — Answer Engine Optimization

AEO means structuring content so that a search engine or AI assistant can easily take the answer from your page. This covers featured snippets, "People also ask", voice answers and AI Overviews.

The core AEO question: does our page contain a clear, accurate one- or two-sentence answer to the user's specific question that can be read on its own? If the answer is buried in a long introduction, systems struggle to find it.

GEO — Generative Engine Optimization

GEO is broader. The goal is for generative systems such as ChatGPT, Gemini, Perplexity and Copilot to cite your brand as a source or mention it in the right context.

A generative model synthesizes an answer from several sources. So not only your website but also what other sites write about your brand affects the result. In GEO, content, technical SEO and PR work together.

How it relates to SEO

Google's official documentation says no special extra optimization is needed to appear in AI features and that core SEO best practices still apply. See Google Search Central's page on AI features.

Which queries trigger AI answers most

AI answers appear most for informational queries: "what is", "how to", "what is the difference", "which to choose". For transactional queries such as "buy CASCO" or "Google Ads agency Baku", users still often go straight to a site or a map.

This shapes the content plan. TOFU and MOFU guides, comparisons and FAQ pages are the main way to appear in AI answers, while BOFU pages — services, pricing, products — must turn that visitor into a conversion.

In practice AEO/GEO does not replace weak SEO. If a page is not indexed, loads slowly or covers the topic superficially, its chance of appearing in an AI answer is low too.

How an AI answer picks its sources

The internal algorithms of AI systems are not public and each one works differently. But the general mechanism is known, and it is enough to make practical decisions.

Step 1: the page must be found and read

The source of an AI answer usually comes from a search index. Google's AI Overviews rely on the Google index, while ChatGPT search finds pages through its own crawlers and search partners.

So the first step is a technical check: the page is indexed, robots.txt does not block it and the main text is in the HTML without depending on JavaScript. The full checklist is in our technical SEO article.

Step 2: allow AI crawlers

Many sites block all bots for security or to save bandwidth, and AI systems then cannot see them. Decide consciously which bots you allow in robots.txt.

• Googlebot — the main crawler for Google Search and AI Overviews; do not block it.
• Google-Extended — controls use of content for Gemini models; it does not affect appearing in Google Search.
• OAI-SearchBot — used to show sites in ChatGPT search.
• GPTBot — for training OpenAI models; blocking it does not block ChatGPT search.
• PerplexityBot — for being cited in Perplexity answers.

This is a business decision. If you do not want your content used for model training, you can block training bots and keep search bots open.

Step 3: picking the passage that fits the query

A system may split the user's question into several sub-queries and look for matching passages for each. A specific paragraph or table is evaluated more than the page as a whole.

If your page answers "how is CASCO insurance priced in Baku" clearly in one paragraph, that paragraph is more likely to be picked. If the answer is spread across five paragraphs, the system may prefer a clearer source.

Step 4: trust signals

When a model sees conflicting information it leans towards the source that looks more trustworthy. Trust signals are close to E-E-A-T in classic SEO: experience, expertise, authoritativeness and trustworthiness.

• It is clear who the author is and why they can speak on the topic.
• Sources and dates are stated.
• The brand is mentioned in the same context on other reputable sites.
• The site has contact details, an about page and a privacy policy.

Content structure: question + direct answer

The most practical change for appearing in AI answers is content structure. The goal is not to write for robots but to answer the reader's question in the first seconds. This works for both people and systems.

The answer block: the first 40–60 words

Tie each main H2 section to a question and answer it directly in the first paragraph. Details, examples and exceptions come next. Journalists call this the inverted pyramid.

Example: a "What is Performance Max?" section should start with "Performance Max is an automated Google Ads campaign type that combines Search, YouTube, Display, Gmail and Maps inventory in one campaign." History and nuances come later.

Definition blocks

"What is X" queries are very common in AI answers. Define the term in one sentence, then explain why it matters and give a practical example.

Lists, steps and tables

Comparisons, step-by-step instructions and selection criteria read better as tables or numbered lists. Models also use structured information more easily than information buried in prose.

• A table for comparisons: "the difference between A and B".
• A numbered list for processes: "how to set it up".
• A short list for criteria: "what to look for".

Azerbaijani-language queries

Azerbaijani users often type in mixed ways: "e" instead of "ə", or with an English term. The same intent may be written as "kaska sigortasi", "KASKO sığortası" or "kasko qiyməti". Use the main spelling in headings and text and mention alternatives naturally once, for example in the FAQ.

This is not keyword stuffing. The goal is not to repeat a word dozens of times but to show the terms people actually use in a readable context.

Checklist

1. Is the page's main question answered in the first paragraph?
2. Does each H2 match a real reader question?
3. Are numbers and claims sourced?
4. Is the text presented with an author and a date?
5. Does the page have at least one comparison table or step list?

We describe the workflow for producing content to this standard in our AI content quality control article.

Schema: FAQ, HowTo, Organization

Schema markup (structured data) describes the information on a page in machine-readable form: this page is an article, this is an organization, this is the product's price. Google's introduction to structured data recommends JSON-LD.

FAQ and HowTo: what changed

A few years ago FAQ and HowTo schema produced large rich results. In 2023 Google limited them: FAQ rich results are mostly kept for well-known government and health sites, and HowTo rich results are no longer shown.

Still, FAQPage markup that describes a question-and-answer block actually visible on the page does no harm and clarifies the structure. The rule is simple: schema must only describe content the user can see.

Organization: the brand's "passport"

Organization schema lists the company's name, website, logo, contact details and official social profiles in one place. It helps AI systems tell your brand apart from similarly named companies. See Google's Organization documentation.

• name and url — exact and identical across the site.
• logo — a raster image in a standard size.
• sameAs — official LinkedIn, Instagram and Facebook profiles.
• contactPoint — phone and email, if shared publicly.

Article, Product, BreadcrumbList

Article for blog posts, Product and Offer for online stores and BreadcrumbList for all inner pages are useful. They help the information be understood in the right context.

Adding ratings, reviews or prices that are not on the page violates Google's guidelines. Placeholder values must not go into schema either.

Brand-mention strategy

A generative model does not learn about your brand only from your website. It also considers news sites, industry directories, forums, comparison articles and social platforms. How your brand is mentioned online is an integral part of GEO.

Links versus mentions

In classic SEO a backlink is the key signal. In GEO an unlinked mention also has value: for example, the brand named with an accurate description in an article about agencies doing e-commerce audits in Baku.

Where to be mentioned

• Industry media and business news portals.
• Professional event, conference and webinar pages.
• Directories and listings (only real, moderated ones).
• Expert commentary: Q&A for journalists, podcasts.
• Case study or partnership pages on client and partner sites.

Consistency matters

The brand's name, description, field of activity and contact details must be the same everywhere. In local SEO this is called NAP (name, address, phone) consistency. Conflicting information confuses models and may lead to a wrong description.

What not to do

Fake reviews, paid link networks, the same text spread across dozens of sites and artificial "best agencies" lists create long-term risk. They violate Google's policies and damage brand trust.

A practical start

1. Ask ChatGPT, Gemini and Perplexity 10–20 typical questions about your brand and note the answers.
2. List the sources cited in those answers.
3. Identify where the brand is described wrongly or out of date.
4. Create pages on your site that answer those questions clearly.
5. Repeat the check monthly and compare the changes.

Measuring AI traffic in GA4

What is not measured cannot be managed. Traffic from AI answers usually mixes with other sites in GA4's Referral channel. A few simple steps let you see it separately.

Recognizing AI sources

When ChatGPT sends a user to a site it often adds utm_source=chatgpt.com to the link. Other systems usually appear by referral domain: perplexity.ai, gemini.google.com, copilot.microsoft.com and so on.

A custom channel group

In GA4 go to Admin → Data display → Channel groups, create a new channel group and add a channel called "AI Assistants". Set the condition that the source matches the regex below.

• chatgpt\.com|openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai

Place this channel above Referral, otherwise the traffic will still fall into referral. Update the list regularly as new AI services appear.

Linking it to revenue

Session counts are not enough. Tie AI traffic to key events — lead forms, calls, purchases — and to e-commerce revenue. We explain how the attribution model affects these numbers in our data-driven vs last click article.

A real example

In our AI search case study content and structure were rebuilt for Google's generative AI features. AI Overviews impressions in Search Console grew from about zero to 500 a day, with 19.1K impressions over four months.

In GA4 the chatgpt.com source was tracked separately and brought $978.08 in revenue in that period — 12.68% of total website revenue. This is not a promise every site will repeat, but it shows why measurement matters.

Watch Search Console too

Track impressions for brand and product queries in Google Search Console. Brands mentioned more often in AI answers often see growth in direct brand searches too. Show this relationship together in the monthly report.

A 90-day AEO / GEO plan

Doing everything at once is hard. The plan below splits the work into three phases. Timings are approximate and depend on site size and team resources.

Days 1–30: diagnostics

• Check indexing and Core Web Vitals reports in Search Console.
• Decide your robots.txt policy for AI crawlers.
• Record how ChatGPT, Gemini and Perplexity answer brand questions.
• Create the "AI Assistants" channel group in GA4 to get a baseline.

Days 31–60: content and structure

• Rework the 10–20 pages with the most traffic or revenue into question + direct answer format.
• Add author, last-updated date and sources to each page.
• Set up Organization, Article, BreadcrumbList and, where relevant, Product schema.
• Prepare content briefs for new informational articles.

Days 61–90: brand and measurement

• Plan expert commentary and case studies for industry media.
• Align the brand description and contact details across platforms.
• Add AI traffic with key events and revenue to the monthly report.
• Repeat the AI answer check and compare it with month one.

This plan is not a guarantee of results. AI systems change quickly, so measure monthly and update the plan.

Frequently asked questions

Conclusion

AI answers are a new layer of search, but an old truth sits underneath: clear, accurate and trustworthy content wins. Technical basics, question-and-answer structure, correct schema and consistent brand mentions work together.

The most important step is measurement. If you do not see AI traffic as a separate channel in GA4, you cannot evaluate the results of your work in this area. Complete the content side with our content marketing strategy and the SEO foundation with our what is SEO guide.