← Back to Editorial

GEO: how to get ChatGPT, Perplexity and Google to cite you

Short answer: GEO (Generative Engine Optimization) is the work of preparing your website and your brand so AI search tools like ChatGPT, Perplexity, Gemini and Google's AI Overviews use you as a source and mention you in their answers. It builds on classic SEO but changes the goal: ranking in a list of ten links isn't enough; you have to be one of the few sources the AI chooses to cite.

More and more buyers ask an AI before they search Google: which agency to hire, which software to pick, how much something costs. If your company isn't in that answer, it isn't in the comparison. This guide explains how those answers work and the six steps we applied on promarketing.com.co. If you'd rather we do it for you, it's part of our SEO services: technical, content and AI search.

What GEO is and how it differs from SEO

SEO aims to get your page ranked well in a search engine's results. GEO aims to get an AI model to read your content, understand it and cite it when it answers. They share the same technical base: if a search engine can't crawl and index your site, an AI won't cite it either.

What changes is the format that wins. An AI assembles answers from fragments of several sources, so it rewards content that answers directly, with verifiable data, and brands that show up consistently across many trustworthy sites.

How AI search tools choose their sources

When you ask a question that needs current information, most of these tools search a web index, retrieve a handful of pages, extract the most useful passages and write an answer with links to their sources. Three things decide whether your page makes that handful:

  • Access. The tool's crawler can read your page, and the page is indexed.

  • Passage relevance. A paragraph on your page answers the question completely, without depending on the rest of the text.

  • Trust. Your brand and your facts are corroborated elsewhere: reviews, directories, press, official profiles.

Step 1: let the right crawlers in

Many companies blocked every AI bot in their robots.txt as a precaution and, without realizing it, removed themselves from the answers. Each AI company runs several crawlers with different jobs, and it's worth deciding on each one.

  • OAI-SearchBot is what OpenAI uses for ChatGPT search. Block it and you don't appear in those answers. OpenAI notes that a robots.txt change can take about 24 hours to take effect.

  • GPTBot is OpenAI's training crawler. Blocking it opts you out of training, not out of ChatGPT search.

  • ChatGPT-User visits pages when a person asks for it inside ChatGPT. According to OpenAI's documentation, because these actions are user-initiated, robots.txt may not apply.

  • PerplexityBot, ClaudeBot and Claude-SearchBot do similar jobs for Perplexity and Claude.

  • Google-Extended controls whether Google uses your content to train Gemini. For AI Overviews, what matters is being properly indexed by Googlebot, Google Search's regular crawler.

In our robots.txt we let search and assistant crawlers through and only slow down the ones that add load without adding visibility.

Step 2: make sure Google and Bing index you

Before thinking about AI, confirm the basics: Google and Bing have your pages in their index. AI Overviews come from Google Search's index, and several assistants rely on existing search indexes to find current information.

  • Verify your site in Google Search Console and Bing Webmaster Tools, and submit your sitemap to both.

  • Check the indexing report to see which pages were left out and why.

  • If your site is bilingual, use hreflang tags so each language is indexed in its market.

  • Fix speed and crawl errors: a slow or broken page gets crawled less.

Step 3: write so you can be quoted

An AI quotes passages, not whole pages. So every section of your content should open with a direct answer of 40 to 60 words that makes sense on its own, then go into detail.

  • One question per section. Use the question your customer would ask as the subheading and answer it in the first paragraph.

  • Data with a source. Your own numbers, dates and specific names. "We migrated 4,000 products" gets quoted; "we have lots of experience" doesn't.

  • Definitions and comparisons. "What is", "how much does it cost" and "X or Y" questions are the ones that trigger the most answers.

  • Real FAQs. The ones your sales team actually hears, with short, complete answers.

  • A visible update date. Fast-moving topics reward recent content.

Step 4: structured data and a consistent entity

To an AI, your company is an entity: a name, an address, a founder, a set of services and profiles that confirm each other. If every site shows a different detail, trust drops.

  • Mark up your site with schema.org structured data: Organization, LocalBusiness, Service, FAQPage, BlogPosting and BreadcrumbList.

  • Use the sameAs property to link your official profiles: LinkedIn, Instagram, Google Business Profile and directories.

  • Keep your legal name, phone, email and address identical on your site, Google Business Profile, LinkedIn and every directory you're listed in.

  • Give the founder or the experts who sign your content their own page, with their background and profiles.

Step 5: llms.txt, without inflated expectations

llms.txt is a proposed standard, published at llmstxt.org, for giving AI models a plain-text summary of your site: what you do, key facts and links to the pages that matter. It's cheap to implement, but no major search engine has confirmed it uses the file to decide whom to cite.

Our advice: do it, because it takes an hour and forces you to organize your information, but don't sell it as the fix. What moves the needle is steps 1 to 4 and step 6.

Step 6: earn mentions beyond your own site

AI tools tend to recommend brands that other sources already recommend. Your own site says who you are; external mentions confirm it's true.

  • Real reviews on Google Business Profile and in your industry's directories. Ask for them as a process, after every project.

  • Directories and marketplaces for the platforms you work with, with a complete profile and your case studies.

  • Credits and links from your clients' websites, when they agree.

  • Press and third-party content: interviews, columns, podcasts, events.

  • Original data: a study with numbers nobody else has is exactly what media and AI models cite.

How to measure whether GEO is working

GEO doesn't have a single dashboard like Search Console yet, so you combine three sources. The key is to take a baseline today and repeat the measurement every month with the same method.

  • Test prompts. Build a list of 20 questions your ideal customer would ask (for example, "best Wix agency in Latin America") and check every month in ChatGPT, Perplexity, Gemini and Google whether you appear, in what position and with which source.

  • AI referral traffic. In your analytics, look at visits from chatgpt.com, perplexity.ai or gemini.google.com. Wix Analytics already groups this traffic into its own AI platform category.

  • Search Console. Impressions on long, question-style queries tend to rise when your content starts answering better.

What we did on promarketing.com.co

We applied this checklist to our own site before offering it to clients. This is what's in production today:

  • A robots.txt that allows the search and assistant crawlers from OpenAI, Anthropic, Perplexity, Google and Apple.

  • llms.txt and a full version, llms-full.txt, generated from the CMS with our services, case studies and FAQs.

  • Structured data on every page: Organization, LocalBusiness, Service, FAQPage, CreativeWork for case studies and BlogPosting for articles.

  • One page per service, in Spanish and English, with direct answers and FAQs.

  • Search Console verified and sitemaps submitted, with hreflang for each language.

It's still too early to show results in numbers. We'll measure with the method above and publish what we find, whether it works or not.

Common mistakes

  • Blocking every AI bot out of fear and disappearing from the answers.

  • Publishing generic AI-written content with no original data: it adds nothing the AI doesn't already have.

  • Buying reviews or links. Besides the penalty risk, AI tools cross-check sources and spot patterns.

  • Ignoring Bing and industry directories because "everyone uses Google".

  • Expecting an llms.txt file to fix what the content doesn't.

Frequently asked questions

Does GEO replace SEO? No. GEO builds on SEO: without crawling, indexing and quality content, no AI will cite you. What it adds is a new goal, which is being the source chosen in an answer.

How long does it take? Technical changes, like opening up robots.txt or fixing indexing, show up in days or weeks. The authority that comes from mentions and original content builds over months.

Can you pay to appear in ChatGPT? Organic answers can't be bought. Some platforms are testing ad formats, but those are separate placements labeled as advertising.

Does it work for B2B companies? Yes, very much so. B2B buyers research thoroughly before contacting a vendor, and more and more of them start that research by asking an AI.

Start with a diagnosis

If you want to know how you show up in AI answers today and what's missing, we'll review your site against this same checklist. Get in touch or see the details of our SEO and GEO service.