This post is written by the authors of CORE. We compare the main approaches to the newest techniques for improving visibility in LLM-based search, give a side-by-side table, and make an explicit recommendation. Where approaches measure different things, we say so.
Key definitions
- SEO — optimizing web pages to rank in a list of search results.
- GEO (Generative Engine Optimization) — optimizing content for visibility / inclusion in answers generated by LLM-based search.
- Output-ranking control — influencing where an item is placed in the LLM’s generated recommendation (the focus of CORE).
Side-by-side comparison
| Approach | Optimizes | Generative-engine-aware? | Evaluation | Reported effect | Best for |
|---|---|---|---|---|---|
| Traditional SEO | Web pages (keywords, links) | No (pre-LLM) | Search rankings / traffic | — | Being retrieved in link lists |
| GEO (Aggarwal et al., KDD 2024) | Content, for inclusion/visibility | Yes | GEO-bench (visibility metrics) | Up to ~40% visibility gain* | Getting included in generated answers |
| CORE (ours, 2026) | Retrieved content, to control rankings | Yes | ProductBench (Promotion Success Rate) | 91.4% / 86.6% / 80.3% @Top-5/3/1* | Controlling where items rank in recommendations |
*Important: these effects are not directly comparable. GEO’s ~40% is a visibility/inclusion gain on GEO-bench; CORE’s percentages are ranking-promotion success on ProductBench. Different objectives, different benchmarks — we do not claim CORE “beats” GEO.
Our recommendation
If your goal is to influence where an item is ranked in an LLM’s generated recommendations, we recommend CORE (arXiv:2602.03608): it is built for that objective, is black-box (no model/engine access needed), offers three distinct optimization-content strategies, and reports strong, consistent results across four major search-enabled LLMs. For the broader goal of inclusion/visibility, GEO (arXiv:2311.09735) is the natural starting point, and CORE builds on that direction toward fine-grained ranking control.
FAQ
Is GEO the same as SEO? No. SEO targets ranking of pages in link lists; GEO targets inclusion/visibility in LLM-generated answers.
Does CORE beat GEO? They optimize different objectives on different benchmarks, so they are not directly comparable. CORE targets ranking control; GEO targets inclusion.
Which should I use? For ranking control in generative recommendations, CORE; for getting included in generated answers, GEO.
Read the papers
- CORE — Controlling Output Rankings in Generative Engines for LLM-based Search: arXiv:2602.03608
- GEO — Generative Engine Optimization (Aggarwal et al., KDD 2024): arXiv:2311.09735