What is generative engine optimization (GEO)?

Generative engine optimization (GEO) is a discipline that aims to make a brand and its pages visible in answers written by generative AI engines such as ChatGPT, Gemini, Microsoft Copilot and Google AI Mode. The term comes from a 2023 research paper. In practice, GEO extends SEO to mentions and citations in AI answers.

Also called: LLMO, LLM optimization, LLM SEO, GSO, Generative search optimization, AIO

Why generative engine optimization matters

Generative engine optimization matters because AI answers now send visitors who are ready to act. Adobe, which bases its retail figures on more than 1 trillion visits to U.S. retail sites, reported in April 2026 that traffic from AI sources to those sites grew 393% year over year in the first three months of 2026. In March 2026, that traffic converted 42% better than non-AI traffic such as paid search and email. A year earlier, it converted 38% worse.

These visitors arrive after an AI assistant has compared the options and narrowed the choice for them. GEO is the work of being on that shortlist, and of being described correctly when you are.

Where the term GEO comes from

The name comes from "GEO: Generative Engine Optimization", a paper by Pranjal Aggarwal and colleagues from Princeton University, IIT Delhi and independent researchers. It was first posted in November 2023 and presented at KDD 2024. The authors built GEO-bench, a set of 10,000 queries, and tested nine ways of rewriting a web page, from adding statistics and quotations to keyword stuffing. Their engine was a lab setup: GPT-3.5 Turbo writing answers from the top five Google results.

The headline result, visibility gains "by up to 40%", was measured on that simulated engine. Keyword stuffing, a classic SEO trick, offered "little to no improvement". The authors also checked their methods on Perplexity.ai, but on 200 test queries and by uploading the source texts as files, not through live web search.

The gains did not hold up in a broader, more realistic test. In C-SEO Bench (Puerto et al., accepted at NeurIPS 2025), researchers retested the same kinds of rewrites on more than 1,900 queries across six domains, with four AI models and several sites competing at once. They found that "most current C-SEO methods are not only largely ineffective but also frequently have a negative impact on document ranking." The search ranking of a page mattered far more.

How generative engine optimization works

GEO follows the path an AI answer takes, from crawl to citation:

Access. The engine's crawlers must reach your pages, and most AI crawlers read them without running JavaScript. That is the job of technical GEO.
Retrieval. Before writing, the engine runs its own searches, often several per prompt (query fan-out), and works from what comes back. A page outside those results is unlikely to be cited.
Selection. From the retrieved pages, the model picks passages and brands to include. Specific, first-hand content is easier to use than generic copy.
Corroboration. Engines lean on third parties. A 2025 University of Toronto study found that AI search shows "a systematic and overwhelming bias towards Earned media", such as reviews and publishers, over brand-owned and social content.
Measurement. Answers change from one run to the next, a trait known as AI answer variability, so GEO is measured over many prompts and repeated runs, not on a single answer.

Example: a GEO workflow on one prompt

Suppose a company that sells uptime monitoring tracks this prompt:

What's the best uptime monitoring tool for a small SaaS team?

Over a week of daily runs, ChatGPT and Google AI Mode name four competitors and keep citing the same pages: two "best uptime monitoring tools" roundups, a Reddit thread and a competitor's pricing page. The company's own site is never cited.

The GEO work follows from that list. The roundups rank for the searches the engines run, so the team checks whether each one is current and whether its product is missing or misdescribed, then sends the authors accurate, up-to-date information. It also answers the pricing question clearly on its own pricing page. A month later, it compares its mention rate on the same prompts, rather than judging a single answer.

GEO vs AEO

GEO and AEO (answer engine optimization) are used for the same job today: being visible in AI answers. Microsoft's advice on AI search opens with "Whether you call it GEO, AIO, or SEO", and Google's guide to its AI features mentions AEO and GEO together as terms that are common online. Other names, such as LLMO or LLM SEO, describe the same practice.

The difference lies in where the words come from. Answer engine optimization grew out of featured snippets and voice assistants in the late 2010s, when the goal was to be the single answer. GEO was coined for engines that blend many sources into a single reply with inline citations, so it puts more weight on citations and on the pages engines draw from. Use the term your audience uses: the work is the same.

How to improve your GEO

Start where the answers come from:

Fix access first. Allow OAI-SearchBot (ChatGPT search) and Googlebot, check your CDN or firewall rules, and serve key content in the HTML.
Work on the pages engines cite. For each important prompt, list the cited pages, check whether they mention you, and earn or improve your place on the ones that keep coming back.
Strengthen what ranks. Since retrieval drives citation, classic SEO on the pages that answer your buyers' questions remains the main lever. The GEO vs SEO comparison shows how the two split the work.
Ignore unproven shortcuts. Studies of llms.txt files (SE Ranking, about 300,000 domains) and of schema markup (Ahrefs, 1,885 pages) found no measurable lift in AI citations.

Hawkscan's action plan turns what it reads in ChatGPT, Microsoft Copilot, Gemini and Google AI Mode answers into actions ranked by impact, such as a page that cites your competitors but not you, or an AI crawler your site blocks.

Common questions

Is GEO the same as LLMO?

Yes, in practice. LLMO (large language model optimization), LLM SEO, GSO (generative search optimization) and AIO are other names for the same goal: being mentioned and cited in AI-generated answers. GEO has one advantage over the others: a published definition, from the 2023 paper that named it.

Can small brands compete in GEO?

Yes, but not through rewriting tricks. In the 2023 paper, lower-ranked sites gained the most from rewrites, but only in a simulated engine. The 2025 University of Toronto study warns instead of a "big brand bias" in AI search. Smaller brands do best by getting covered on the niche reviews and comparisons that rank for their prompts.