GEO: How to Get Recommended by ChatGPT Without Paying

Crawlers, answer-first pages, structure, sources, a consistent brand description, reviews and measurement: a GEO guide grounded in what OpenAI, Google and research actually document.

By Hadrien Bonato-Pape · Published on September 28, 2026 · Strategy

GEO (generative engine optimization) means making your business easy for AI assistants like ChatGPT to find, understand and cite. You can't buy it, because ads don't influence answers. You build it: a site crawlers can reach, pages that answer up front, structured and sourced content, a brand description that matches everywhere, and genuine reviews.

This guide pulls together what the platforms themselves document (OpenAI and Google pages read on September 30, 2026) and what research has actually measured, with no miracle percentages. If you're weighing this work against advertising, our explainer ChatGPT Ads: what they are and how they work shows why one doesn't replace the other.

What exactly is GEO?

The term comes from a research paper, "GEO: Generative Engine Optimization" (Aggarwal et al., presented at KDD 2024). The authors call tools that answer with written text drawn from several sources, rather than a list of links, "generative engines." GEO covers everything a website can do to show up in those answers and get cited.

In practice that mostly means ChatGPT (with web search), Google's AI Overviews and AI Mode, and other assistants that cite their sources. GEO doesn't replace SEO; it builds on it.

Can you pay to be recommended by ChatGPT?

No. OpenAI repeats it on its ads help page: ads run on systems separate from the model, are labeled as sponsored and appear below the answer. Advertisers can't shape, rank or alter what ChatGPT says. A ChatGPT Ads campaign buys you a paid slot under the answer, never a place inside it. GEO is the only route to the latter.

How does ChatGPT find information about your business?

When a question needs current information, ChatGPT can search the web and cite its sources, OpenAI explains in Searching the web with ChatGPT. It often rewrites the question into one or more targeted queries, which it may send to partner search providers (OpenAI points to Microsoft's privacy statement among others). It then reads the pages it finds and writes an answer with citations.

Two OpenAI crawlers matter here, per its crawler overview:

• OAI-SearchBot is for search. Sites that opt out of it won't be shown in ChatGPT search answers.
• GPTBot is for model training. You can block it and still appear in search.

Google's guidance, AI features and your website, is refreshingly direct: there are no extra requirements or special optimizations for AI Overviews and AI Mode, standard SEO best practices apply, and a page must be indexed and eligible for a snippet to appear as a supporting link.

Step 1: let the right crawlers in

This is the fastest check and the one most often missed. Open your site's robots.txt and your hosting or CDN settings:

1. Allow OAI-SearchBot if you want to appear in ChatGPT search. OpenAI notes that robots.txt changes can take about 24 hours to take effect.
2. Allow Googlebot for Google Search and its AI features.
3. Make sure your firewall or CDN isn't blocking these crawlers by default. OpenAI publishes their IP ranges so you can allow them.
4. Keep important information in text, not only inside images or videos. Google recommends this explicitly.

Step 2: answer first, then elaborate

An assistant is looking for an answer to a question. Give it in the first lines of the page, in one or two sentences that stand on their own, then go into detail. A "Services" page that opens with three paragraphs of company history loses to one that starts with "Nord Coffee is a specialty café in Portland, open seven days a week, roasting its beans on site."

Think in the questions customers actually ask: "how much does… cost," "what's the difference between…," "is it right for…". Every important question deserves a clear answer somewhere on your site.

Step 3: structure for machine readers

• Headings that say what's below, ideally phrased the way customers ask.
• One idea per section, so a passage can be quoted on its own without losing meaning.
• Tables for comparisons (plans, prices, hours) and numbered lists for steps.
• An FAQ at the bottom of key pages.
• Structured data if you use it, as long as it matches the visible text. Google notes that no special schema.org markup is required for its AI features.

These principles mirror a good content strategy: a page that helps a busy human helps an assistant too.

Step 4: source your claims

This is the most telling finding in the GEO paper. Of the nine techniques the authors tested, citing sources, adding quotations from relevant sources and adding statistics were among the ones that improved visibility in generated answers the most, along with more fluent writing. Keyword stuffing, a leftover from old-school SEO, offered little to no improvement.

In practice: when you state a fact, say where it comes from and link the source. A number with no origin helps neither the reader nor the assistant. And never make things up; a false claim repeated by an assistant costs you more than not being mentioned at all.

Step 5: describe your brand the same way everywhere

Assistants cross-check what they read. If your site says "small-batch roaster," your Google profile says "tea room" and your Instagram says "coffee shop," the question "what is this place?" gets three different answers. Write a two-sentence reference description (what you do, for whom, where) and reuse it:

• on your homepage and About page;
• on your Google Business Profile and your industry's directories;
• in every social media bio;
• in press releases and partner pages.

Name, address, phone number, hours and offer should match everywhere. Google asks for the same thing in local search, and keeping your Business Profile up to date is on its list of best practices for AI features.

Step 6: get talked about beyond your own site

Your site says what you think of yourself; reviews, articles and mentions say what others think. For local search, Google says a business's prominence is based in part on how many websites link to it and on its reviews, and that more reviews and positive ratings can help its ranking (Google Business Profile Help).

Ask happy customers for reviews, reply to all of them including the negative ones, and aim for mentions in media and specialist sites in your field. One hard rule: Google prohibits offering incentives (discounts, freebies) in exchange for reviews. Reviews have to reflect a genuine experience.

What about Perplexity, Gemini and Claude: how do they choose their sources?

Each assistant runs its own crawlers, and each crawler has a specific job: some feed a search engine, some fetch a page when a user asks a question, some collect data to train models. The rule is the same as with OpenAI: allow the search crawlers, and make a deliberate call on the training crawlers. The details change from company to company, and blocking the wrong crawler can make you invisible in one assistant without changing anything in the others (company pages reviewed in late September 2026).

Perplexity

Perplexity describes itself as an answer engine. Its Help Center explains that it searches the internet, gathers information from articles, websites and journals, and that each answer includes numbered citations linking to the original sources. Its crawler documentation separates PerplexityBot, which surfaces and links websites in its results, respects robots.txt, isn't used to train foundation models and which Perplexity recommends allowing, from Perplexity-User, which visits a page when a user asks a question and generally ignores robots.txt. Perplexity publishes the IP ranges for both, which you can hand to your host or firewall provider.

Google: AI Overviews, AI Mode and Gemini

Google's AI features and your website page is clear: no additional requirements or special optimizations; to be a supporting link, a page must be indexed and eligible for a snippet; these features may use "query fan-out," several related searches across subtopics, so a page that nails one specific subtopic can get cited even if it doesn't target the main question; no AI text files or special schema.org markup are needed; nosnippet, data-nosnippet, max-snippet and noindex limit what can be reused. The Google-Extended token decides whether your content can train future Gemini models and ground answers in the Gemini apps; Google states it doesn't affect inclusion in Google Search and isn't a ranking signal. Blocking it won't pull you out of AI Overviews, but it can keep Gemini from using your content.

Claude

Claude, Anthropic's assistant, can search the web, and every response based on a search includes citations (help page on web search; on Team and Enterprise plans, an owner has to turn it on first). Anthropic describes three crawlers in its article on crawling the web and how to block it: ClaudeBot collects content for model training; Claude-User visits pages when a user asks a question, and blocking it may reduce your visibility; Claude-SearchBot improves search result quality, and blocking it may reduce your visibility and the accuracy of what's said about you. All three honor robots.txt, including Crawl-delay, and Anthropic publishes their IP addresses.

Check your site in fifteen minutes

1. Open your robots.txt and look for the names in the table: a "Disallow: /" line under PerplexityBot, Claude-SearchBot or Googlebot makes you invisible to that service.
2. Check the catch-all rules: "User-agent: *" followed by "Disallow: /" blocks everyone, search crawlers included.
3. Ask your host whether a firewall or "bot protection" setting blocks automated visits; if so, give them the IP lists Perplexity and Anthropic publish.
4. Check Google indexing in Search Console: a page that isn't indexed can't be a supporting link in AI Overviews.
5. Look for nosnippet tags on your key pages.

For measurement, visits from AI Overviews and AI Mode are counted in Search Console's Performance report under the "Web" search type, without being broken out; for Perplexity and Claude, the pages we reviewed don't document a tracking parameter, so look in your analytics for visits whose source is their domains, when the referrer is passed along. None of these pages describes a paid way into the answers.

What about llms.txt?

llms.txt is a proposal by Jeremy Howard, first published in September 2024: a markdown file at the root of a site that summarizes it and lists its most useful pages for language models. It has caught on mainly with technical documentation sites.

Let's be precise about what it does. Google says you don't need to create AI text files to appear in its AI features. And none of the OpenAI pages reviewed for this guide lists it as a factor in ChatGPT search. Creating one costs next to nothing and does no harm; expecting results from it would be a mistake. Do steps 1 through 6 first.

Step 7: measure what's happening

Two measurements are available today:

1. Traffic from ChatGPT. OpenAI automatically adds utm_source=chatgpt.com to links clicked from its answers, per its publisher FAQ. Filter on it in your analytics tool.
2. Google Search Console. Clicks from AI Overviews and AI Mode are counted in the "Web" search type.

Add a simple routine on top: a fixed list of questions your customers would ask, put to the same assistants every month, with a tracking sheet. We'll cover that in a dedicated article.

The recap

How Oposto helps with consistency and cadence

Two GEO levers play out over time: saying the same thing everywhere, and publishing useful content regularly. Oposto learns your brand from your website, building its content DNA (voice, products, audience), and uses it to generate posts, carousels and short videos that match your description, published to 8 networks once you approve and adapted to each one. It then measures real performance so you can reuse what works. To keep that pace without it eating your week, read our piece on a constant social presence without burnout, then try Oposto for free.

Frequently asked questions

What's the difference between GEO and SEO?

SEO aims for a spot in a list of results; GEO aims for a mention or citation inside an AI-written answer. GEO builds on SEO: Google says its standard best practices still apply to its AI features.

Do I need to allow GPTBot to appear in ChatGPT?

No. According to OpenAI, OAI-SearchBot is the crawler used for search. GPTBot is for model training, and you can block it while staying visible in ChatGPT search.

Does llms.txt improve visibility in ChatGPT?

Nothing shows that so far. Google says such files aren't needed for its AI features, and the OpenAI pages reviewed don't list it as a search factor.

Do ChatGPT ads help you get recommended?

No. OpenAI says ads don't influence ChatGPT's answers. They appear below the answer and are clearly labeled as sponsored.

How can I tell if ChatGPT sends me visitors?

Look for utm_source=chatgpt.com in your analytics tool. OpenAI adds that parameter automatically to links clicked from ChatGPT search.