Try Cituna

Glossary September 2026

What is generative engine optimization (GEO)?

Generative engine optimization (GEO) means shaping content so AI engines that write answers from web sources use and cite it more. Where the term came from, what the original study tested and found, and the common misconceptions.

Mark HenryEditor

By Mark Henry, Editor Published Reading time 3 min

Disclosure: VisibilityOS is published by the team that builds Cituna, one of the tools discussed here. Read this as our view, not a neutral review.

Generative engine optimization (GEO) is the practice of shaping web content so that generative engines, the AI systems that find sources and write an answer from them, use that content more often and more prominently in their answers. The term was introduced in a 2023 research paper that defined these engines, proposed ways to measure a source's visibility inside their answers and tested nine ways to raise it.

Where the term comes from

The paper, "GEO: Generative Engine Optimization", is by Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan and Ameet Deshpande. It was first posted to arXiv on 16 November 2023 and accepted to KDD 2024.

Its starting point: generative engines gather information from several sources and summarize it with a large language model, and the sites they draw on have little or no control over when and how their content appears. GEO was their framework for measuring and improving that visibility.

How the study measured GEO

The authors built GEO-bench, a benchmark of 10,000 queries from multiple sources and domains, and scored visibility in two ways:

  • Position-Adjusted Word Count: the share of the answer's words that come from sentences citing a source, weighted so that sentences near the top of the answer count for more.
  • Subjective Impression: a score for how a reader would perceive the citation, covering its relevance, influence, uniqueness and position, and how likely a reader is to click it.

They then rewrote source pages nine ways: a more authoritative tone, added statistics, keyword stuffing, cited sources, added quotations, simpler language, smoother fluency, unique words and technical terms.

What the study found

  • Citing sources, adding quotations and adding statistics worked best, with relative improvements of 30 to 40 percent on Position-Adjusted Word Count and 15 to 30 percent on Subjective Impression.
  • The paper's published prompts let the rewriting model invent that material: quotes "even though fake and artificial", statistics "even if hypothetical", and sources it "may invent". Its test engine was built on GPT-3.5. So the gain shows what quote-, statistic- and citation-shaped text did to that engine, not that real evidence has the same effect on today's engines.
  • Keyword stuffing, a classic SEO tactic, gave little to no improvement.
  • Lower-ranked pages benefited more. Citing sources raised visibility by 115.1 percent for sites ranked fifth in the search results, while the top-ranked site's visibility fell by 30.3 percent on average.
  • Results varied by domain, which led the authors to argue for domain-specific methods.

An example: the test on a live engine

The authors repeated the experiment on Perplexity.ai, a deployed engine with a large user base. Adding quotations was again the best method on Position-Adjusted Word Count, with a 22 percent gain over the unmodified baseline, and keyword stuffing did 10 percent worse than the baseline.

The benchmark and live figures differ, so treat any single GEO number as a direction, not a forecast.

Why GEO matters

GEO measures something a ranking report does not: how much of the written answer comes from your page. That share depends on the engine. In our scans (how we measured), Perplexity cited 20.2 third-party sources per answer and Gemini 3.5, so a single citation is a far larger slice of a Gemini answer than of a Perplexity one. Tactics with the evidence behind each are in 10 ways to get cited by AI answer engines.

For Google's own features, Google's guide to generative AI search (last updated 10 July 2026) says that from Google Search's point of view, optimizing for generative AI search is still SEO.

Common misconceptions

  • "GEO is keyword work." Keyword stuffing gave little to no gain on the benchmark and did worse than doing nothing on Perplexity.ai.
  • "One tactic works everywhere." The study found effects varied by domain, and the lab and live results differed.
  • "GEO means rewriting pages for machines." Google's guide says rewriting content just for AI systems is unnecessary, because they understand synonyms and general meaning. How to write content that AI engines cite covers the practical side.
  • "GEO and AEO are separate jobs." Google's guide treats both as names for work on visibility in AI search. AEO vs GEO vs SEO covers where they differ in emphasis.

Correction

23 September 2026: added what the GEO paper's published prompts allowed (invented quotes, statistics and sources) and the GPT-3.5 test engine, so the 30 to 40 percent figure is not read as proof that real evidence works the same way.

Sources

  1. arXiv 2311.09735, Generative Engine Optimization (Aggarwal et al.), abstract. Retrieved .
  2. arXiv 2311.09735, Generative Engine Optimization (Aggarwal et al.), full text v3. Retrieved .
  3. GEO-optim, GEO source code, src/geo_functions.py. Retrieved .
  4. Google Search Central, Optimizing your website for generative AI features on Google Search. Retrieved .
All terms