Generative engines pick sources while writing answers
Knowing a brand is one part of the assistant's path to recommending it. The assistant may put that name into a search, find pages, read them and choose which companies to include. We have seen a brand recalled at the start disappear from the searches that followed.
The name GEO comes from Aggarwal and co-authors' paper on arXiv, submitted in 2023 and accepted to KDD 2024. Their subject was improving visibility in generative engine responses.
- Enter consideration. An assistant may remember a company as it begins working out an answer to the buyer's question.
- Supply a search result. A relevant page gives the assistant something to find during its own searches.
- Support a recommendation. What the assistant reads must make a case for naming the company in the answer.
Geo-targeting and generative GEO do unrelated jobs
| Term | What it aims to do | Where it applies |
|---|---|---|
| Geo-targeting | Reach people in, regularly in or interested in selected locations | Ad platforms, including Google Ads |
| SEO | Get a page ranked among search links | Results on Google and Bing |
| Generative engine optimization | Get the company named or its information cited in an answer | Generated responses in ChatGPT, Perplexity, Gemini and Google AI Overviews |
Google describes location targeting in ads. Conductor's GEO entry concerns generative answers. Omnius uses citations to distinguish that work from the rankings associated with SEO.
No company owns most markets' AI answers yet
An assistant's list of suppliers can shape which companies a buyer considers. Buyers often begin a purchase with a shortlist and choose from the names already on it. We see most markets' AI answers as still open: companies have yet to establish a settled position in those recommendations. For a B2B founder, that makes the answers worth examining now, while assistants are still choosing which companies and sources to put before buyers.
GEO work covers answers, pages and outside mentions
The work covers company pages and outside mentions, which we expect to carry more weight in AI answers over time.
Record the answers
Ask 30 buyer questions on three engines, ChatGPT, Perplexity and Gemini, and record the companies each answer names.
Improve the pages
Add evidence to the pages answers should cite, and check schema, llms.txt and markdown copies so the information can be examined.
Develop outside mentions
Work on LinkedIn posts, Reddit threads and directory listings, then repeat the questions to examine what the assistants say after those changes.
Some GEO offers are location ads
- Clarify the offer. Ask whether GEO means choosing where ads appear or working on company mentions in generated answers.
- Check each engine. Assistants retrieve sources differently, so a change aimed at one engine may have little effect on another.
Google's location-targeting guidance explains the advertising usage. Omnius describes differences in retrieval between engines. An answer check records what those assistants returned at that point.