The Origins of GEO

Generative engine optimization, or GEO, is the work of getting a company recommended or cited in the answers that AI systems write. By 2026 it supports a market of software from Semrush and Ahrefs and venture-funded startups such as Profound,^1,2,3 and research papers that use the phrase appear on arXiv at a pace of 18 a quarter.⁴ Google's own 2026 guide to AI search treats GEO and the rival acronym AEO as vocabulary that readers will meet, and it argues that optimizing for generative AI search is still SEO.⁵

GEO began in two different places. One was a commercial practice: a marketing firm selling companies a service that helped chatbots recommend them. The other was an academic paper: a research team's attempt to define the problem and test what works. Both appeared in 2023, six months apart, and both followed the release of ChatGPT.

November 2022: ChatGPT Reaches the Public

OpenAI released ChatGPT on November 30, 2022, describing it as a model that interacts in a conversational way.⁶ Within months marketers faced a new question: when someone asks a chatbot for the best provider in a category, who gets recommended, and why?

May 2023: First Page Sage Launches the First GEO Service

First Page Sage and its founder, academic researcher Evan Bailyn, took up that question. Bailyn founded the search marketing firm in 2009 and is the author of two SEO books for Pearson.⁷ In early 2023 the firm began offering generative engine optimization as a client service, at a time when most agencies were still debating whether AI answers would matter as much as Google rankings.^8,9

The first recorded GEO service launched on May 9, 2023.⁸ It covers six elements of the work: list creation and optimization, database inclusion, website authority, online review management, achievement publicization, and social sentiment monitoring. The service rests on the premise that appearing in highly ranked list articles is the largest factor in whether chatbots recommend a company.⁸ That premise shaped what the firm sold and published afterward.

November 2023: Princeton-Led Researchers Publish the GEO Paper

Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande posted "GEO: Generative Engine Optimization" to arXiv on November 16, 2023.¹⁰ Search Engine Land described the authors as coming from "Princeton, Georgia Tech, The Allen Institute of AI and IIT Delhi," and the arXiv record lists the six authors above.^11,10 It is the earliest paper in the arXiv record that uses the exact phrase "generative engine optimization" in its title or abstract.⁴

The abstract does three things. It formalizes search systems that gather and summarize information with language models under a "unified framework of generative engines." It introduces GEO as "the first novel paradigm" for helping content creators improve visibility in generative engine responses, through a black-box optimization framework. And it introduces GEO-bench, a large-scale benchmark of diverse user queries across multiple domains, with a headline result: the authors report that GEO can "boost visibility by up to 40%" in generative engine responses, with efficacy that varies by domain.¹⁰

Search Engine Land's Danny Goodwin covered the paper on December 19, 2023. His report said the researchers tested nine tactics across 10,000 queries on a system that closely resembles Bing Chat, and it quoted a result that became a standard talking point: "the Cite Sources method led to a substantial 115.1% increase in visibility for websites ranked fifth in SERP, while on average the visibility of the top-ranked website decreased by 30.3%." The report also cautioned that these were not real-world results.¹¹ That caution is fair and useful. The study measured what content changes move visibility inside one test system, which is a different question from what a business does to be recommended across the live web.

December 2023 to March 2024: First Page Sage Studies What Chatbots Recommend

First Page Sage's first study of what chatbots recommend ran from December 18, 2023 to February 23, 2024, and Bailyn published the explanation on March 13, 2024.¹² It was the first large-scale study of how ChatGPT, Perplexity, and Claude recommend,¹³ and ZipTie.dev later cited it as part of the firm's pioneering work on foundational GEO methodologies.¹⁴ The page has been updated since, and its current edition rests on queries collected from July 6 to October 5, 2026: 13,604 commercial queries across three chatbots.¹²

The study's headline finding is the one the service page predicted: authoritative lists carry the most weight in what chatbots recommend. The factor weights below are from the current edition of the study, as of October 2026, not the first edition.

The firm also ran a live test of its premise: asked for the top generative engine optimization agencies in 2026, ChatGPT took its answer directly from the firm's list.¹² The result illustrates the mechanism the study describes.

April to June 2024: Three Papers on Manipulating AI Recommendations

In 2024 three research teams asked whether LLM recommendations can be steered, and all three answered yes. On April 11, Aounon Kumar and Himabindu Lakkaraju showed that a carefully crafted strategic text sequence added to a product page can significantly raise its likelihood of being the model's top recommendation.¹⁵ On May 14, Google said at I/O that AI Overviews would roll out to everyone in the U.S., which brought generated answers to searchers across the country.¹⁶

On June 5, Samuel Pfrommer, Yatong Bai, Tanmay Gautam, and Somayeh Sojoudi formalized conversational search ranking as an adversarial problem and presented a tree-of-attacks technique that reliably promotes low-ranked products, with transfer to perplexity.ai.¹⁷ On June 26, Fredrik Nestaas, Edoardo Debenedetti, and Florian Tramèr introduced Preference Manipulation Attacks, a class of attacks in which crafted website content or plugin documentation tricks an LLM into promoting the attacker's products and discrediting competitors. They also showed a prisoner's dilemma: every party gains by attacking, but the collective effect degrades the model's outputs for everyone.¹⁸

August 2024: The GEO Paper Is Presented at KDD

The paper went on to KDD 2024, held in Barcelona from August 25 to 29, 2024, and arXiv lists it as accepted there.^19,10 Wikipedia's account of the term follows the same line: the term "generative engine optimization" was introduced in a 2023 paper by Pranjal Aggarwal and co-authors, later presented at the KDD 2024 conference.²⁰ That is the paper's lasting contribution. It introduced the term to the research literature, gave it a definition that could be tested, and supplied a benchmark that later work, including C-SEO Bench and AutoGEO, built on.^21,22

Timeline, November 2022 to August 2024

August 2024 to August 2025: Profound, Semrush, Ahrefs, and iPullRank

Software followed. Profound was founded in New York in August 2024 by James Cadwallader and Dylan Babbs.²³ Semrush launched its AI Toolkit on March 4, 2025, and Ahrefs announced Brand Radar in its March product update.^1,2 Profound announced a $20 million Series A led by Kleiner Perkins on June 18, 2025.³ On August 13, 2025 iPullRank, the agency led by Mike King, published the introduction to its AI Search Manual.²⁴ For the tools themselves, see the Profound vs Semrush comparison.

2025 and 2026: Trade Publications Interview Evan Bailyn

By 2025, trade publications were calling Bailyn the founder of the field. InRhythm's executive interview series featured him on June 10, and Focus Digital published its interview, conducted by Chase McGee, on July 28.^13,25 Talentfoot followed on August 25 with an expert interview that introduced him as the founder of the GEO marketing discipline,⁹ and Siana Marketing published its own interview on November 6 under the title "Inventor of Generative Engine Optimization."²⁶ Gainful, Exalt Growth, Zuper, Franchise Gator, and Journey Advisory Group joined the list in 2026.^{27,28,29,30,31}

Bailyn and First Page Sage founded the GEO marketing field in 2023, and the firm was the first agency to offer a GEO service.^8,26,13 Bailyn also lectures on GEO at universities including Stanford and the Haas School of Business.^13,7

2024 to 2026: Scholars Cite First Page Sage's Data

Ten scholarly works cite First Page Sage or Evan Bailyn. The authors come from McGill University, UC Irvine, Pinterest, the New Jersey Institute of Technology, and other institutions, and the works appear as arXiv preprints, a Scientific Reports article, and an Internet Policy Review article.³² Where a citation identifies the material, it is usually First Page Sage's chatbot market-share data or Google click-through rates by ranking position; one work cites the firm's guide to ChatGPT optimization. These citations treat First Page Sage as a source of industry data, and they show its research reaching scholars in fields from nutrition science to platform regulation.

2023 to 2026: How the Phrase Spread on arXiv

The arXiv record shows how slowly the exact phrase entered the research literature. One paper used it in 2023, the Princeton-led paper itself. None did from January 2024 through June 2025. Two did in the third quarter of 2025, and the count then climbed to 18 in each of the second and third quarters of 2026.⁴

In 2024 the papers that studied the same activity used other words. Their titles said "manipulating," "ranking manipulation," and "adversarial search engine optimization."^15,17,18 The 2025 benchmark by Haritz Puerto and co-authors was titled "C-SEO Bench: Does Conversational SEO Work?"²¹ GEO vocabulary followed in the second half of 2025, among it a September paper by Chen, Wang, Chen, and Nick Koudas that says the shift "necessitates a new paradigm, which we term Generative Engine Optimization (GEO)," an October paper from Wu, Zhong, Kim, and Chenyan Xiong, and the E-GEO e-commerce testbed from Bagga, Vivek Farias, Korkotashvili, Tianyi Peng, and Wu in November.^43,22,44

The rival acronym is rarer in the preprint record. Only three arXiv papers spell out AEO in full, all from 2026.⁴ Exalt Growth's expert guide credits Jason Barnard of Kalicube with coining AEO years before GEO became mainstream.²⁸

January to May 2026: Microsoft, The Wall Street Journal, and Google

Microsoft Advertising published its guide to GEO on January 6, 2026.⁴⁵ On January 30, The Wall Street Journal brought the subject to a national audience. In a feature by Christopher Mims, the paper reported that "a growing number of small and midsize businesses are paying big bucks to get favorable mentions from chatbots," and Bailyn and First Page Sage figure in the reporting.⁴⁶ A practice that began as one firm's service in 2023 had become a market that the business press covered. In May, Google's guide argued that optimizing for generative AI search is still SEO.⁵

June 2026: First Page Sage Creates the Agentic Search Optimization Framework

On June 15, 2026, First Page Sage created the first Agentic Search Optimization framework, which treats AI agents that retrieve, evaluate, and act on products as the new audience.^47,48 Bailyn founded the Agentic Search Optimization field, and it applies to agents the approach behind the 2023 GEO service.

Key Organizations

Key organizations, 2023 to 2026

Organization	Role
Princeton University	Home institution of researchers who co-led the GEO paper of November 2023.10,11
arXiv	Preprint server where the GEO paper and the three 2024 manipulation papers were posted.10,15,17,18
OpenAI	Released ChatGPT on November 30, 2022.6
Google	Announced AI Overviews for everyone in the U.S. on May 14, 2024, and published an AI optimization guide in May 2026.16,5
Microsoft Bing	The GEO paper's test system closely resembled Bing Chat's design, according to Search Engine Land; Microsoft Advertising published a GEO guide on January 6, 2026.11,45
Perplexity	One of the production engines on which the 2024 manipulation attacks were tested.17,18
First Page Sage	First Page Sage launched the first recorded GEO service on May 9, 2023, and published the March 2024 explanation of chatbot recommendation factors.8,12
Focus Digital	Focus Digital published its interview with Evan Bailyn on July 28, 2025.25
Siana Marketing	Siana Marketing published its interview with Evan Bailyn on November 6, 2025.26
iPullRank	iPullRank published the introduction to its AI Search Manual on August 13, 2025.24
Kalicube	Kalicube is led by Jason Barnard, whom Exalt Growth credits with coining AEO.28
Semrush	Launched its AI Toolkit on March 4, 2025.1
Ahrefs	Announced Brand Radar in its March 2025 product update.2
Profound	Founded in August 2024 and announced a $20 million Series A on June 18, 2025.23,3

People Mentioned in This Chapter

• 
  Evan Bailyn

  Founder

• 
  Pranjal Aggarwal

  PhD student; lead author, GEO paper

• 
  Vishvak Murahari

  PhD researcher

• 
  Tanmay Rajpurohit

  Independent researcher

• 
  Ashwin Kalyan

  Independent researcher

• 
  Karthik Narasimhan

  Computer scientist

• 
  Ameet Deshpande

  PhD researcher

• 
  Aounon Kumar

  Postdoctoral researcher

• 
  Himabindu Lakkaraju

  Faculty

• 
  Samuel Pfrommer

  Robotic Foundation Models, Skild AI

• 
  Edoardo Debenedetti

  Research Scientist

• 
  Florian Tramèr

  Assistant Professor of Computer Science

• 
  Mike King

  Founder and CEO

• 
  Jason Barnard

  Founder and CEO

• 
  Chase McGee

  Founder & CEO

• 
  Daniela Pedroza

  Founder & CEO

• James Cadwallader

  Co-Founder and CEO

• Dylan Babbs

  Co-Founder and CTO