The first thing clients want to know is whether their years of SEO work still count. The answer is a useful “partly.” Some of what you built for search carries straight into generative engines. Some of it doesn’t transfer at all. Knowing which is which saves you from both panic and complacency.

What transfers

The fundamentals that made your site legible to a crawler also help a model. None of this work is wasted:

  • Clean technical foundations: fast, crawlable, well-structured HTML.
  • Schema markup: structured data is, if anything, more valuable to a model than to Google.
  • Genuine topical depth: real coverage of a subject reads as authority to both systems.
  • Clear information architecture: logical headings and internal links help extraction.

If you invested in substance over tricks, most of that substance comes with you.

What doesn’t transfer

The parts of SEO that are specifically tuned to Google’s ranking function have little purchase on citation selection:

  • Backlink volume as a primary lever: corroboration and source diversity matter more than raw link counts.
  • Keyword-density tactics: models read meaning, not term frequency.
  • Page-one position itself: ranking is not citation, and the two pools overlap less than you’d expect.

28% of pages cited by ChatGPT have zero visibility in Google. Your search standing does not automatically become your answer standing.

We pulled that number apart in a separate note on the citation pool. The short version is that the citation pool and the search index are assembled by different processes, so a strong search rank is no guarantee of a place in the answer.

The new work GEO adds

Beyond what carries over, generative search introduces requirements SEO never asked of you:

  1. Extraction-first structure: self-contained, attributable claims a model can lift.
  2. Active recency: a refresh loop, because freshness is a live citation signal.
  3. Per-model measurement: you can’t manage a number you don’t track across ChatGPT, Claude, Perplexity, and the rest.
  4. Trusted-pool seeding: mentions in the third-party sources models already pull from.

The takeaway

GEO isn’t a replacement for SEO and it isn’t a rebrand of it. It’s an adjacent discipline that reuses your best technical foundations and discards the tactics that only ever spoke to Google. Keep the substance; retire the tricks; add the new work. Here’s how we sequence it →

// instruments referenced in this research
site_restructuring → visibility_analysis → faq_engine → content_refresh →