GEO Services: Get Cited by ChatGPT and Perplexity

Generative Engine Optimization (GEO) is the work of getting a brand named and linked inside answers written by AI assistants such as ChatGPT, Perplexity, Gemini, Claude and Google’s AI Overviews. Brilad does this by fixing how AI systems reach your site, how they identify your brand, and whether your pages contain passages worth quoting.

Where the term comes from

The name was introduced in a 2023 research paper by researchers from Princeton University, Georgia Tech, the Allen Institute for AI and IIT Delhi (“GEO: Generative Engine Optimization”, Aggarwal et al., later published at KDD 2024). They built a benchmark of search queries and tested how changes to a web page affected its visibility in AI-generated answers.

Two results from that paper shape how we work. Adding cited sources, direct quotations and statistics to a page raised its visibility in generated answers by up to about 40% in their tests. Keyword stuffing, the old SEO habit, did not help and sometimes hurt. The paper also found that the effect of each method varied by subject area, which is why we test changes on your own queries instead of applying a fixed checklist.

How AI assistants pick the pages they cite

Most AI assistants that show citations use retrieval-augmented generation (RAG). The assistant rewrites your question into one or more search queries, pulls candidate pages from a search index, selects passages from those pages, and writes an answer based on them. Google calls the query-splitting step “query fan-out” in its AI Mode documentation.

That process creates four points where a brand can drop out:

The crawler is blocked, or the page only shows its content after JavaScript runs.
The page is in the index but is not retrieved, because its wording does not match the sub-queries the assistant generates.
The page is retrieved but no single passage answers the question on its own, so another source is quoted.
The passage is used, but the brand is described inconsistently across the web, so the assistant is unsure what the brand is and leaves the name out.

Our GEO work addresses each of these in order.

Crawler access

AI companies run separate crawlers for model training, for their search index, and for fetching a page when a user asks about it. Blocking the wrong one can remove you from AI answers entirely.

Company	Training	Search index	User-requested fetch
OpenAI	GPTBot	OAI-SearchBot	ChatGPT-User
Anthropic	ClaudeBot	Claude-SearchBot	Claude-User
Perplexity	(none listed)	PerplexityBot	Perplexity-User
Google	Google-Extended (a robots.txt control token that uses the Googlebot crawl)	Googlebot	Google uses separate fetchers for some user-triggered features

Google-Extended controls whether your content is used for Gemini model training and grounding. It does not affect Google Search, and AI Overviews use the normal Googlebot crawl. The only way to keep a page out of AI Overviews is to restrict snippets with nosnippet, which also removes the normal search snippet. Most businesses should not do that.

The second access problem is JavaScript. Vercel’s 2024 analysis of AI crawler traffic found that the major AI crawlers it measured, apart from Google’s, download JavaScript files but do not execute them. A Next.js or React site that builds its content in the browser can look almost empty to those crawlers. We check what each crawler actually receives and move important content into server-rendered HTML where needed.

Entity clarity

An AI assistant can only recommend a company it can identify with confidence. We make the brand consistent in the places models learn from and retrieve from:

Organization, Person and Service schema on your site, linked into a single JSON-LD graph with stable @id values.
sameAs links from that schema to your LinkedIn page, Wikidata item, Crunchbase profile and other official profiles.
The same company name, address and phone number on your site, Google Business Profile and directories.
Author pages for the people who write your content, with their actual role and experience.

Passages worth citing

An assistant quotes a passage of a few sentences, so the passage is the unit that has to work. We rewrite important pages so each section opens with a direct answer to one question and can be understood without reading the rest of the page. Then we add material the assistant cannot get from ten other sources: your own data, pricing ranges, process details, named examples and first-hand test results. Pages that only restate what everyone else says give the assistant no reason to pick them.

About llms.txt

llms.txt is a proposal published by Jeremy Howard of Answer.AI in September 2024: a Markdown file at the root of a site that lists its most useful pages for language models. It is cheap to add, and we add it. We do not sell it as a ranking factor. As of 2026 no major AI provider has publicly committed to using llms.txt to choose citations, and Google has said its systems do not use it. Anyone promising results from llms.txt alone is overselling it.

Measuring AI visibility

AI answers change between runs, between models and between signed-in users, so a single screenshot proves very little. We build a fixed set of prompts that match how your buyers ask questions, run each prompt several times on each assistant on a schedule, and record:

whether your brand is mentioned, and in what position
which of your URLs are cited
which competitors are named instead
whether the description of your company is accurate

The last point matters more than most teams expect. Wrong descriptions (an old service, a closed office, a competitor’s product attributed to you) usually trace back to a specific outdated page on your site or elsewhere, and that page can be fixed.

Frequently asked questions

Why doesn’t ChatGPT mention my company?

Usually one of seven causes: ChatGPT is not searching for that question, OAI-SearchBot cannot reach your site, you do not appear in search for the sub-questions, no page answers the question directly, your brand is confused with another, few other sources mention you, or it describes you wrongly. Our 7-point diagnostic checklist has a test for each.

Can anyone guarantee that ChatGPT will recommend my company?

No. Nobody outside OpenAI controls ChatGPT’s answers, and the same is true for every other assistant. GEO increases the chance of being cited and measures the change over time. A guaranteed placement offer is a red flag.

Should I block GPTBot and other training crawlers?

That is a business decision about whether your content may be used to train models. It is separate from AI search visibility. You can block GPTBot and still allow OAI-SearchBot, which is the crawler that decides whether you appear in ChatGPT search results.

How is GEO different from SEO?

GEO builds on SEO: Google’s AI features use its search index, and ChatGPT search draws on search results. GEO adds access for non-Google crawlers, passage-level answers, brand mentions across the web and tracking of AI answers. Our article GEO vs SEO: what actually changes goes through each difference with sources.

How is GEO different from AEO?

AEO targets answers shown on a search results page, such as featured snippets. GEO targets answers written by AI assistants. The content techniques overlap, but GEO depends much more on entity clarity and on which crawlers can reach you.

How long before changes show up in AI answers?

Assistants that search the web live can reflect a change once the relevant crawler revisits the page, often within days or weeks. Changes to what a model knows without searching only arrive when the model is retrained, which can take months and is outside anyone’s control.

GEO Services (Generative Engine Optimization)