A generative engine reads several pages and answers in its own sentences, with the sources inside the answer, where an earlier web engine would have handed back a list [1] [2]. Presence in that answer is its own kind of visibility: a source cited early, and woven through more of the sentences, occupies more of what a person actually reads [1]. Aggarwal, Murahari, and their colleagues called the craft of preparing for this generative engine optimization. On a benchmark of 10,000 queries, the stronger preparations raised visibility by as much as 40 percent [1].
The edits that helped were small enough to miss. Adding a citation from a reliable source, a credible quotation, or a relevant statistic improved a position-weighted word count by about 30 to 40 percent, and a judged impression by about 15 to 30 percent [1]. On Perplexity, a system already open to the public, quotations raised the position-weighted count by 22 percent, and statistics raised the judged impression by as much as 37 percent [1].
One result cuts against a very common habit. Adding more of the question’s own keywords, long treated as basic search work, did little or nothing on their test engine [1]. On Perplexity it lowered the position-weighted count by about 10 percent, even while a subjective score ticked upward [1]. Another surprise sits beside it. A more authoritative tone, which many of us would expect to help, did not show a clear gain across queries as a whole, though it fit debate and history better than other subjects [1]. Citations suited factual questions, and numbers suited law, debate, and opinion [1]. Fluency paired with statistics beat either change alone, by a little over 5 percent, in a smaller follow-up of 200 examples [1].
The finding we return to is about who moves. For the citation method, a page ranked fifth in ordinary search gained about 115 percent in visibility, while the page already ranked first gave up about 30 percent [1]. The authors’ reading is that a generated answer depends on the text it was shown, so the links a smaller site has not yet earned need not decide the whole of its chances [1]. The same paper says the engines will change, and the useful edits may have to change with them [1]. Across four live generative engines, only about 51.5 percent of sentences were fully supported by their citations, and only about 74.5 percent of citations actually supported the sentence they were tied to [3]. The answers that felt most helpful were often the ones whose citations were least precise, a correlation of about −0.96 [3].
What did the 2024 edits change?
The study did not find one edit that helped every subject in the same way.
| Edit | What the study reported |
|---|---|
| Citation, quotation, or a relevant statistic | About 30 to 40 percent on a position-weighted word count, and about 15 to 30 percent on a judged impression. |
| Quotations on Perplexity | About 22 percent on the position-weighted count. |
| Statistics on Perplexity | Up to about 37 percent on the judged impression. |
| Keyword stuffing | Little or nothing on the authors’ engine. On Perplexity the position-weighted count fell by about 10 percent, while a subjective score rose. |
| Authoritative tone | No clear gain across queries as a whole. A better fit for debate and history. |
| Fluency together with statistics | A little over 5 percent above either change alone, on a follow-up of 200 examples. |
What is generative engine optimization?
It is the craft of preparing a page for an engine that answers in its own sentences, with the sources inside the answer.
Which small edits raised visibility in the 2024 study?
Adding a citation, a credible quotation, or a relevant statistic raised a position-weighted word count by about 30 to 40 percent.
Does stuffing a question’s keywords help a generated answer?
On the authors’ test engine it did little or nothing, and on Perplexity the position-weighted count fell by about 10 percent.
Does an authoritative tone help?
It did not show a clear gain across queries as a whole, though it fit debate and history better than other subjects.
Who gained when citations were added?
A page ranked fifth in ordinary search gained about 115 percent in visibility, while the page already ranked first gave up about 30 percent.
How often did a citation support its sentence?
Across four live engines, about 74.5 percent of citations supported the sentence they were tied to.
References
- Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., & Deshpande, A. (2024). GEO: Generative engine optimization. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. https://doi.org/10.1145/3637528.3671900
- Brin, S., & Page, L. (1998). The anatomy of a large-scale hypertextual Web search engine. Computer Networks and ISDN Systems, 30(1–7), 107–117. https://doi.org/10.1016/S0169-7552(98)00110-X
- Liu, N. F., Zhang, T., & Liang, P. (2023). Evaluating verifiability in generative search engines. Findings of the Association for Computational Linguistics: EMNLP 2023. https://aclanthology.org/2023.findings-emnlp.467/