Short answer
GEO (Generative Engine Optimization) is the work of getting a brand found, understood and cited in the answers of artificial intelligence tools such as ChatGPT, Gemini, Perplexity, Claude and Google's AI Mode.
SEO competes for a position in a list of links. GEO competes for a place inside the answer the AI writes. It doesn't replace SEO: it uses the same technical foundation and adds four fronts, which are easy-to-cite content, a consistent brand identity across the internet, authority in sources AI consults and monitoring what each AI says about you.
What GEO is, explained without jargon
When someone asks ChatGPT “what's the best industrial maintenance company in Campinas?”, the AI doesn't return ten blue links. It reads several sources, compares, summarizes and delivers a ready-made recommendation, sometimes with two or three names and the links it pulled the information from.
GEO is everything a company does to be in that recommendation, and to be described correctly when it shows up. The acronym stands for Generative Engine Optimization: optimizing for systems that generate a new answer instead of just listing pages.
In practice, GEO answers three questions about your brand:
- Visibility: when someone asks about what you sell, does the AI mention you?
- Perception: when it does, what does it say? Are the facts right? Is the tone positive?
- Position: do you appear as the first recommendation or as the fourth option, after your competitors?
You'll also come across names like AI SEO, LLMO (Large Language Model Optimization), AIO (AI Optimization) and AEO (Answer Engine Optimization). They describe the same territory with different emphases, and I explain the difference further down.
Where the term GEO came from
The term took off with the research paper “GEO: Generative Engine Optimization”, by researchers at Princeton University and the Indian Institute of Technology Delhi. The first version came out in November 2023 and the work was presented at KDD 2024, one of the world's leading data science conferences, in Barcelona.1
The authors created GEO-bench, a set of 10,000 questions across 25 domains, and tested nine ways of rewriting content to see which increased its presence in generated answers. What they found became the market's foundation:
The finding that stands out most to me is in the details: sites in lower search positions gained much more from GEO than the leaders. In the test, citing sources raised the visibility of a site in 5th place by 115%, while the 1st-place site actually lost ground. For mid-sized companies, which will never beat the industry's biggest portal in traditional SEO, that opens a door.
I started studying GEO because of this paper, and what convinced me was the finding about smaller sites. For years, the answer to “how do I beat the leader on Google” was “you don't”. In GEO, content quality and clarity carry real weight again.
A caveat: the study tested text changes in a controlled environment. Today's commercial AIs use many more signals, such as reputation, mentions on other sites and location data. The paper proves that content form matters, not that rewriting text is enough.
Why GEO matters now
Until recently, the buying journey started on Google and went through a list of links. In 2026, a growing part of it starts and ends inside a conversation with an AI. The numbers help size the shift:
In Brazil, the milestone was September 8, 2025, when Google launched AI Mode in Portuguese. It works as a search tab and in the Google app, and the questions asked there are two to three times longer than a regular search.3 In other words, people describe the whole problem and expect a complete answer, not a list.
There are two practical effects for a company:
- Fewer clicks, more decisions made before the click. When AI summarizes, compares and recommends, many people reach your site already knowing what they want, or never reach it, because they chose the competitor that was cited.
- Visitors who arrive via AI tend to convert better. Traffic from chatbots is still small, around 0.28% of all web visits in March 2026, versus 28% from Google.7 But in a case tracked by Seer Interactive, ChatGPT visitors converted at 15.9%, versus 1.76% for Google organic search.8 It's a single case, not a rule, but the pattern repeats in several industry reports.
The question is no longer “am I on the first page?” but “what does AI answer when someone asks about what I sell?”
What the research shows: the numbers that matter
I've gathered here the most solid studies published through November 2026, most of them done in the United States, where AI search arrived first. For each number, I note who did it, with what sample and when. Keep in mind that much of it comes from companies that sell SEO and GEO tools, so I treat the data as direction, not law.
When Google shows an AI summary, people click a traditional result in 8% of visits. Without the summary, in 15%. Links cited inside the summary get clicked in only 1% of visits, and 26% of people end their browsing there, versus 16% without a summary.
Pew Research Center, browsing data from 900 US adults, 68,879 searches, March 2025.15
Brand mentions on the web correlate three times more strongly with appearing in AI Overviews than backlinks (0.664 vs. 0.218). The second strongest factor is links with the brand name in the anchor text (0.527).
Ahrefs, 75K brands, 2025, revised in 2026.16
of pages cited in AI Overviews rank in the organic top 10 for the same search. In July 2025, it was 76%. The rest come from positions 11 to 100 (31%) and beyond position 100 (31%).
Ahrefs via Search Engine Journal, 863K keywords, March 2026.17
of pages cited by ChatGPT search ranked 21st or worse on Google for the same search. Being first on Google guarantees nothing on ChatGPT.
Semrush, AI search study, July 2025.19
of pages cited by ChatGPT, Gemini and Perplexity had been updated in the past year. On Gemini, 78%. Comparisons and reviews are the “freshest”.
Seer Interactive, 7,683 pages and 47K citations, July 2026.23
Ranking 1st on a heavily cited comparison list (“the best X”) raises the chance of the brand being mentioned by up to 16.5 percentage points, and makes it appear earlier in the answer.
Peec AI, 200K answers from 8 AIs, May 2026.21
is how much a visitor from AI search is worth, on average, compared with a traditional organic visitor, measured by conversion rate. Semrush projects that AI traffic could surpass traditional search by 2028.
Semrush, July 2025 (company projection).19
of ChatGPT answers add the word “best” to searches the user never typed. “Review”, “top”, “comparison”, “vs” and the current year are also added frequently.
Peec AI, 5 million fan-out searches, April 2026.22
Of the thousand pages ChatGPT cites most, only about a third are something a brand can directly influence. The rest are Wikipedia, dictionaries, app stores and corporate pages of big brands. Distribution by page type:
Ahrefs, top 1,000 pages cited by ChatGPT, September 2025.18
The numbers tell the same story along different paths. Clicks are shrinking, Google's top 10 no longer guarantees a citation and what weighs most is what the internet says about your brand, not just what your site says. For a mid-sized company, that's good news: mentions, reviews and well-made content are within reach, while earning a link from a major portal almost never is.
How AI decides who to cite
Each platform has its quirks, but AIs that search the web follow a similar path from question to answer. Understanding that path is what separates GEO from guesswork.
- The person asks a long question“I need a preventive maintenance company with 24/7 service for a factory in Campinas. Who's reliable?”
- The AI splits the question into several searchesThe technique is called query fan-out. Google describes it as breaking the question into subtopics and firing several queries at once: maintenance companies in Campinas, reviews, contract pricing, 24/7 service, certifications.This is where most brands disappear
- It looks for sources for each queryIt uses a search index (Google's, for AI Overviews and AI Mode; other indexes and its own browsing, for ChatGPT, Perplexity and Claude) and reads the most relevant pages.
- It picks the passages that answer bestIt prefers clear, specific passages with data and sources, from sites it can access and that look trustworthy. Opinions repeated across several independent sources weigh more.
- It writes a new answer and cites a few sourcesThe answer blends what it read with what the model already knew. Some sources appear as links. Brands can be recommended even without a link, when several sources speak well of them.
This fan-out varies a lot between AIs. In a survey of 5 million searches, Peec AI found an average of 1.4 searches per question on Perplexity, 2.1 on ChatGPT and 6.8 on Grok. And the AI adds words the user didn't type, such as “best”, “review”, “comparison” and the current year.22 In practice, your page competes for searches nobody typed, so it also needs to answer “best X in 2026”, “X vs Y” and “is X worth it?”.
That's why there are two layers of presence, and good GEO work covers both:
What the model already “knows”
The knowledge learned in training. It changes slowly, with each new model version, and reflects what the internet said about your brand up to that date. Reputation, mentions and consistency over time matter here.
What the AI searches in the moment
The pages read at the time of the question. It changes fast, in days or weeks. Indexing, crawler access, content clarity and pages that answer the fan-out questions matter here.
Each AI has different preferences
No single source dominates every AI. An analysis of 30 million cited sources, done by Peec AI in the United States in March 2026, showed Reddit, YouTube, LinkedIn and Wikipedia among the most cited domains overall, but with very different profiles:5
- ChatGPT pulls more editorial and encyclopedic sources, with Wikipedia at the top.
- Google AI Overviews and AI Mode pull more YouTube, Reddit, Facebook and LinkedIn.
- Perplexity cites Reddit heavily and also review platforms like G2.
A report by Tinuiti with Profound, from January 2026, reached the same conclusion by another route: the most cited source changes with the platform, the industry and the intent of the question.6 The same report shows AI Mode cites 143% more distinct domains than AI Overviews. For a brand, that means ranking well on Google doesn't guarantee presence on ChatGPT, and vice versa. You have to measure AI by AI.
GEO, SEO, AEO and LLMO: what's the difference
The terms overlap and the market hasn't settled on a vocabulary yet. This is the most useful way I've found to tell them apart:
| SEO | AEO | GEO | |
|---|---|---|---|
| Goal | Rank pages in the results list | Be the direct answer to a question | Be cited and recommended inside the generated answer |
| Where it shows up | Traditional Google and Bing | Featured snippets, voice assistants, “people also ask” | ChatGPT, Gemini, Perplexity, Claude, Copilot, AI Overviews, AI Mode |
| Unit of success | Position and click | The chosen answer | Mention, citation, position in the recommendation and how the brand is described |
| What weighs most | Relevance, links, page experience | Objective, well-structured answer | Everything in SEO, plus a consistent entity, citable content and reputation in external sources |
| How it's measured | Search Console, rankings, traffic | Presence in answer snippets | Questions monitored on each AI, visibility, sentiment, cited sources and traffic from AI |
LLMO (Large Language Model Optimization) and AI SEO are used as synonyms for GEO. Google, for its part, makes a point of saying in its 2026 guide that, from its point of view, AEO and GEO “are still SEO”.4 That makes sense for Google, because its answers come from the same index and the same search quality systems. But ChatGPT, Claude and Perplexity aren't Google, and the work on perception and presence in external sources goes beyond classic SEO.
At Mind, the rule is simple: GEO first is not GEO only. We start with presence in AI answers because that's where the decision is happening, and we organize SEO, content, media and the site to reinforce that presence.
GEO platform by platform
Each AI searches, reads and cites in its own way. This is a practical summary of what's known about each one, with the crawler that needs to be allowed on your site.
ChatGPT
OpenAI- How it searches
- It uses its own search index and reads pages at question time. It splits the question into about two searches and often adds “best”, “review” and the year.
- Crawlers
OAI-SearchBotputs the site in ChatGPT search.GPTBotis for training and can be blocked without removing you from search.ChatGPT-Useropens pages at the user's request.- Preferred sources
- Wikipedia, Reddit, Forbes, TechRadar and LinkedIn. Comparison lists are almost a fifth of cited URLs, but have been losing ground since late 2025.
In practice: allow OAI-SearchBot, have very clear “about” and services pages and work on presence in third-party lists and comparisons.
Google AI Overviews and AI Mode
Google- How it searches
- It uses Google Search's own index and quality systems, with query fan-out. AI Mode cites many more distinct domains than AI Overviews.
- Crawlers
Googlebotitself. There's no special crawler or markup.Google-Extendedcontrols use in training and in the Gemini app, not presence in search.- Preferred sources
- YouTube, Reddit, Facebook, LinkedIn, Yelp and Medium. In AI Mode, category pages and official brand sources stand out.
In practice: solid SEO, a complete Google Business Profile, YouTube videos and category and service pages that answer the fan-out questions.
Gemini
Google- How it searches
- It uses Google Search as a base to answer about the present.
- Preferred sources
- Reddit, YouTube, Wikipedia and Medium. It's the AI most sensitive to fresh content: 78% of the pages it cites were updated in the past year.
In practice: keep dates, prices and information current and show the update date on the page.
Perplexity
Perplexity AI- How it searches
- It always searches the web and shows sources very visibly. It does fewer fan-outs per question than the others.
- Crawlers
PerplexityBotfor indexing.- Preferred sources
- Reddit leads, with about a quarter of citations in January 2026, followed by YouTube, LinkedIn, Wikipedia and review platforms like G2.
In practice: be present where your customers discuss the problem and have reviews on industry platforms.
Claude
Anthropic- How it searches
- It searches the web when the question calls for current information.
- Crawlers
- Three separate crawlers since 2026:
Claude-SearchBotindexes for search,Claude-Useropens pages at the user's request andClaudeBotcollects training data. Anthropic warns that blocking Claude-SearchBot may reduce visibility.
In practice: check that your robots.txt and CDN don't block Claude-SearchBot and Claude-User.
Copilot and Bing
Microsoft- How it searches
- It uses the Bing index. Since February 2026, Bing's official guidelines mention GEO by name and cover how content is used in Copilot answers.
- What Microsoft asks for
- Headings that work as chapters, question-and-answer pairs, lists and tables, JSON-LD structured data and facts stated directly. Avoid answers hidden in tabs, PDFs or images.
- Watch out
- The
NOARCHIVEdirective prevents the page from being used in Copilot answers. The guidelines also started penalizing “artificially fabricated language” and prompt injection aimed at AI.
In practice: register the site in Bing Webmaster Tools and review your robots meta tags.
When a client asks me which AI to prioritize, the answer depends on where their audience is, but I almost always start with ChatGPT and Google. ChatGPT because that's where the largest user base is. Google because AI Mode and AI Overviews show up for people who are already searching. The mistake is optimizing for one and assuming the others will follow: the data shows different sources, crawlers and preferences on each.
What works in GEO, with evidence
I've split the practices into six fronts. For each one, I note where the evidence comes from, because the GEO market is full of recipes without proof.
1. Let AI in
If the page isn't indexed, or the AI's crawler is blocked, it can't be read or cited. Check robots.txt and your CDN (such as Cloudflare) so you don't block search crawlers like OAI-SearchBot (ChatGPT), PerplexityBot and Google's and Anthropic's crawlers. Keep important content in text, not just in images or heavy PDFs.
Evidence: a technical requirement documented by Google for AI Overviews and AI Mode and by the platforms themselves.
2. Answer first, explain later
Open every page and every section with a direct two- or three-sentence answer, then go deeper. Use headings that are real customer questions. A good page answers the main question and the questions the AI will fan out from it: price, timeline, comparison, who it's for, who it's not for.
Evidence: the query fan-out mechanism described by Google and the gain from clear, fluent language in the Princeton study.
3. Bring data, sources and expert quotes
Your own numbers (average response time, how many clients, measured results), sourced statistics and quotes from people with a name and title make content easier to cite. Generic content, like everyone else's, is rarely chosen.
Evidence: quotes, statistics and source references yielded 30% to 41% more visibility in the Princeton study. Google's guide asks for original, non-commodity content.
4. Be the same company everywhere
AI needs to be sure who you are. Name, description, address, phone, services and differentiators should match on your site, Google Business Profile, LinkedIn, Instagram, directories and marketplaces. A complete “about” page and Organization structured data, with links to official profiles, help connect the dots.
Evidence: Google asks that structured data match the visible text and says there's no special schema for AI. Contradictory information is the most common reason AI gets a brand's facts wrong.
5. Build authority beyond your site
AI trusts what several independent sources say about you more than what you say about yourself. Reviews on Google and industry platforms, press coverage, inclusion in lists and comparisons, YouTube videos, community discussions and well-kept LinkedIn profiles feed the answers.
Evidence: Reddit, YouTube, LinkedIn, Wikipedia and review platforms are among the domains AI cites most (Peec AI, 30 million sources, 2026).
6. Keep it alive and monitor
Update dates, prices, team and cases. AI answers change frequently, and a competitor can take your place within weeks. Track a fixed set of questions on each AI and adjust content based on what they answer.
Evidence: the variation in sources across platforms and over the months, documented in the 2026 Tinuiti and Profound report.
The content formats that show up most
Cross-referencing studies from Peec AI, Seer Interactive, Ahrefs and Microsoft's guidance, some formats appear more often in answers:
Comparisons and lists
“The best X”, “X vs Y”. ChatGPT cites them a lot, though the trend is declining. Lists with more options, an explained method and a recent date hold up better.
Category and service pages
The format that stands out most in Google's AI Mode, alongside the brand's official sources.
Questions and answers
Clear question-and-answer pairs are easy to extract, as Microsoft recommends.
Reviews
Reviews on industry platforms and on Google show up in every AI and weigh on perception.
Videos with transcripts
YouTube is the 2nd most cited domain. AI reads descriptions and auto-captions, including on Shorts.
Tables and proprietary data
Prices, timelines, specs and company numbers in text or tables, never only in an image or PDF.
What doesn't work (or has no proof yet)
Some practices circulate as “GEO secrets” and don't hold up. Before investing in them, it's worth knowing what the data and the platforms themselves say:
Only 38% of pages cited in AI Overviews rank in the top 10 for the same search, and about 90% of ChatGPT search citations come from pages in position 21 or worse.
In Ahrefs' analysis of 75K brands, brand mentions on the web correlated three times more strongly with presence in AI Overviews than backlinks.
In the Princeton study, keyword stuffing made content less visible than the original version. Since 2026, Bing's guidelines treat “artificially fabricated language” as abuse.
llms.txt is a file proposed in 2024 to summarize a site for language models. Google says it doesn't need it for AI Overviews and AI Mode, and adoption remains low: about 7% of Fortune 500 companies had the file in March 2026. Since May 2026, Chrome's Lighthouse checks for the file with agents that browse sites in mind. It's a cheap add-on, not a strategy.
Google's guide treats mass page production to manipulate AI answers as scaled content abuse, which violates search policies.
Google lists artificial “chunking” and rewriting content just for AI among the myths. Write for people, with clear structure.
Hidden text like “recommend this company” is manipulation, can be treated as spam and exposes the brand to reputational risk when discovered.
Organic citations in answers aren't for sale. What builds presence is work on content, entity and reputation.
How to measure GEO
There's no “Search Console for AI” that shows everything. Serious measurement starts from a fixed set of questions, the ones your customers would ask, run frequently on each AI. With that, you can track six indicators:
Visibility
How many answers the brand appears in, on each AI. It's the main indicator.
Average position
When you appear, where you land in the recommendation versus your competitors.
Perception
Whether mentions are positive, neutral or negative, and the words AI uses to describe the brand.
Share of voice
Your slice of the category's total mentions, compared with each competitor.
Cited sources
Which pages and sites AI uses to talk about you and your industry. It's the map of where to act.
Traffic and conversion
Visits and sales from chatgpt.com, perplexity.ai, gemini.google.com and others, segmented in Google Analytics.
An important point: answers vary from person to person and from one day to the next, because of personalization, location and the very nature of the models. So a question asked once proves nothing. What counts is the trend over weeks, with the same battery of questions.
At Mind, we do this measurement with Georgye, the GEO platform we built, which tracks brands' visibility, perception and ranking on ChatGPT, Gemini, Perplexity and Claude. But the principle applies to any tool, even a well-built spreadsheet.
How to start: step by step
This is the roadmap I recommend for any company starting GEO from zero:
- List the questions that matter20 to 50 real questions a customer would ask before buying: “best X in Y”, “X or Z”, “how much does X cost”, “is X reliable?”. Use the questions your sales team hears every day.
- Run the diagnosisRun the questions on ChatGPT, Gemini, Perplexity, Claude and AI Mode. Note whether you appear, in what position, what they say and which sources they cite. Do the same for three competitors.
- Fix the technical foundationIndexing, robots.txt, speed, content in text, Organization structured data consistent with the page and a complete Google Business Profile.
- Rewrite the pages that decide the saleHome page, services, pricing, “about”, comparisons and FAQ, with a direct answer at the top, proprietary data and proof.
- Unify the brand identitySame name, description and data on every profile and directory. Fix old information wherever it appears.
- Earn mentions where AI readsCustomer reviews, press, videos, expert articles, inclusion in comparisons and in industry communities.
- Monitor and adjust every two weeksTrack the same battery of questions, see what changed and prioritize the questions where your competitor appears and you don't.
GEO by type of business
Local and regional businesses
Clinics, offices, restaurants, service industries and physical stores depend heavily on the Google Business Profile, reviews and correct location data. Questions like “best dentist in Curitiba” are answered with heavy weight on those signals, especially on Google.
B2B companies
The purchase is long and researched. AI is used to build short lists of suppliers and compare options. Published cases, certifications, technical content, LinkedIn presence, inclusion in comparisons and the reputation of the company's experts all weigh in.
E-commerce and product brands
AI already recommends products and, more and more, helps people buy. Complete, consistent product information, up-to-date price and availability, reviews, how-to videos and presence on review sites and in communities all matter.
Professionals and personal brands
Doctors, lawyers, consultants and creators get cited when there are clear signs of expertise: a consistent bio, original content, interviews, publications and complete professional profiles.
What's next
- Agents that shop for you. 72% of consumers have tried or accept an AI agent shopping for them, versus 39% a year earlier, according to Global Payments' Agentic Commerce Report 2026.10 When the agent chooses, the brand needs to be easy for a machine to understand and compare.
- Increasingly conversational and multimodal search. In AI Mode, questions are already much longer than in traditional search and a meaningful share uses images and voice. Photos, videos and clear descriptions gain weight.
- More personalized answers. With memory and context for each user, AI will give different answers to different people. Measuring by sampling and by trend will become even more important.
- Fewer clicks, more weight on the brand. If people decide before visiting the site, being remembered and well described by AI is worth more than topping a list.
GEO glossary
- AI Overviews
- AI-generated summary that Google shows at the top of some searches, with links to the sources.
- AI Mode
- Google's conversational search tab, available in Portuguese in Brazil since September 2025.
- Query fan-out
- Splitting a question into several parallel searches, done by the AI to build the answer.
- Grounding
- When AI anchors its answer in pages searched in the moment, instead of answering only from what it learned in training.
- RAG
- Retrieval-Augmented Generation: a technique of retrieving documents and using them to generate the answer. It's the technical basis of most AI search.
- Citation
- When AI shows a link to your page as a source for the answer.
- Mention
- When AI names your brand in the text, with or without a link.
- Entity
- The “profile” AI and search engines build about a company, person or product: name, what it does, where it operates, how it relates to other entities.
- AI visibility
- The share of answers, within a set of monitored questions, in which the brand appears.
- Share of voice
- The brand's slice of the category's total mentions, compared with competitors.
- AI agent
- AI that performs tasks for the user, such as researching, comparing and buying, browsing sites on its own.
- llms.txt
- File proposed in 2024 to summarize a site for language models. Optional and still rarely read.
GEO FAQ
What is GEO?
GEO (Generative Engine Optimization) is the work of getting a brand found, understood and cited in the answers of AIs such as ChatGPT, Gemini, Perplexity, Claude and Google's AI Mode. SEO competes for a position in a list of links; GEO competes for a place inside the answer.
Does GEO replace SEO?
No. AIs that search the web use search indexes and well-ranked pages to build their answers, and Google itself states that SEO best practices still apply. GEO adds entity work, citable content, off-site authority and answer monitoring on top.
What's the difference between GEO and AEO?
AEO started out focused on direct answers, such as featured snippets and voice assistants. GEO is the term for generative engines, which write a new answer from several sources. In practice, the market uses both for the same work.
How long does it take to see GEO results?
Technical and content fixes on already indexed pages can show up in web-searching answers within weeks. Changing how AI describes the brand, which depends on mentions in other sources, usually takes months. That's why the work starts with a diagnosis and continues with ongoing monitoring.
Do I need to create an llms.txt file?
It isn't required. Google says it doesn't use llms.txt for AI Overviews and AI Mode, and adoption is still low. Since May 2026, Chrome's Lighthouse checks for the file with agents that browse the site in mind. It's worth it as a cheap add-on, never as a strategy.
Does structured data (schema) help with GEO?
It helps make clear who the company is and what it offers, as long as it matches what's written on the page. Google states there's no special schema for appearing in AI answers.
How do you measure GEO results?
With a fixed set of questions monitored on each AI, measuring visibility, average position, perception, cited sources and share versus competitors, plus traffic and conversions from AI in Google Analytics.
Do backlinks still matter for GEO?
They do, mainly for traditional search, which feeds AI. But Ahrefs found, in an analysis of 75K brands, a three times stronger correlation between brand mentions on the web and presence in AI Overviews than between backlinks and that presence.
Which AI should I prioritize?
It depends on where your audience is, but ChatGPT and Google (AI Overviews and AI Mode) usually come first, because of the size of their user base. Each AI uses different sources and crawlers, so the ideal is to measure all of them and prioritize where competitors already appear and you don't.
Can you pay to appear in an AI answer?
Organic citations in answers aren't sold. What exists is presence work: content, entity, reputation and mentions in sources AI consults.
Sources
- Aggarwal, P. et al. GEO: Generative Engine Optimization · Princeton and IIT Delhi, KDD 2024
- Google Search Central. AI features and your website (PT) · updated Dec 2025
- Google Brasil. Brazil can now use AI Mode in Portuguese (PT) · Sep 8, 2025
- Google Search Central. Optimization guide for generative AI features in Search · 2026
- Peec AI. Top domains cited by AI search, analysis of 30 million sources · Mar 2026
- Search Engine Land. Tinuiti and Profound, AI Citation Trends Q1 2026 · 2026
- Ahrefs. AI chatbot traffic · May 2026
- Seer Interactive. How traffic from ChatGPT converts · Jun 2025
- PPC Land. Google adds llms.txt to Lighthouse · May 2026
- Mercado&Consumo. Global Payments, Agentic Commerce Report 2026 (PT) · Oct 2026
- Meio & Mensagem. VML, Future Shopper 2026 (PT) · Sep 2026
- eWeek. ChatGPT reaches 900 million weekly users · Feb 2026
- Business Wire. Ahrefs, AI Overviews and click-through rate · May 2026
- Conversion. Google AI Mode, usage data after 1 year (PT) · May 2026
- Pew Research Center. Google users are less likely to click on links when an AI summary appears · Jul 2025
- Ahrefs. An analysis of AI Overview brand visibility factors (75K brands) · May 2025, revised Apr 2026
- Search Engine Journal. Ahrefs: AI Overviews citations from the top 10 drop to 38% · Mar 2026
- Ahrefs. 67% of ChatGPT’s top 1,000 citations are off-limits to marketers · Oct 2025
- Semrush. AI search and SEO traffic study · Jul 2025
- Peec AI. Content types cited most by engine · Feb 2026
- Peec AI. Listicle rank effect on visibility · May 2026
- PPC Land. Peec AI, 5 million fan-out searches analyzed · Jun 2026
- Seer Interactive. Content recency’s impact on AI visibility in 2026 · Jul 2026
- Seer Interactive. The listicle window is closing in AI search · Feb 2026
- Peec AI. Most cited domains by engine · Mar 2026
- Peec AI. Reddit in AI search · Mar 2026
- Peec AI. YouTube in AI search · Mar 2026
- Microsoft Advertising. Optimizing your content for inclusion in AI search answers · Oct 2025
- Search Engine Journal. Bing adds GEO to official guidelines · Feb 2026
- Search Engine Journal. Anthropic’s Claude bots make robots.txt decisions more granular · Feb 2026
- OpenAI. Overview of OpenAI crawlers · official documentation
I lead marketing strategy at Mind Marketing and the agency's GEO practice, which helps companies across Brazil appear, and be well described, in AI answers. Every day I track how ChatGPT, Gemini, Perplexity and Claude recommend brands, with the help of Georgye, Mind's GEO platform.
Find out what AI says about your company.
In 30 minutes, we show how your brand appears on ChatGPT, Gemini, Perplexity and Claude, and where competitors are ahead of you. You leave with a written diagnosis.