Generative Engine Optimization (GEO): rank in ChatGPT and Perplexity

A meaningful share of B2B research now starts in ChatGPT, Claude, Perplexity, and Google’s AI Overviews instead of a plain Google search. For some categories — coding tools, SaaS comparisons, technical decisions — it’s already the majority of early-funnel research. This is the version of SEO most marketing teams aren’t ready for, and the practice that’s emerged to handle it has a new name: GEO, or Generative Engine Optimization.

Here’s what’s actually working and what isn’t.

What GEO is (and isn’t)

GEO is the practice of getting your brand cited or quoted when AI search engines answer questions in your category. It overlaps with SEO but isn’t the same thing. The end goal is the same — be the source of the answer — but the channel rewards different signals.

It is: optimizing for being a cited source inside an AI-generated answer.

It isn’t: just doing regular SEO and hoping AI engines pick you up. Some of regular SEO transfers; much of it doesn’t.

Why the channel matters now

Three signals worth watching:

Google’s AI Overviews appear on a growing share of commercial queries
Perplexity has crossed real revenue milestones for a query-answering product
ChatGPT Search and similar features inside Claude are getting heavy use for early-funnel research

When a buyer asks “what’s the best SEO tool for small teams?” and gets a 200-word AI answer with three product names cited, the click-through from that answer is small but the recall is enormous. Buyers who see your brand cited in that answer arrive at your site already pre-qualified.

How AI engines pick sources

Different engines weight differently, but the patterns across all of them rhyme:

Authority signals — does your site rank for related queries? Do other authoritative sites cite you?
Recency — is the content current?
Structure — is the answer extractable? Headings, lists, tables, FAQs help.
Specificity — vague generalist content gets passed over for specific, data-backed answers.
Schema — structured data helps the engine confidently parse the page.

Note what’s not on this list: keyword density, exact-match phrases, link velocity. Old-school SEO tactics that still move Google sometimes don’t move AI engines at all.

The technical layer

What to make sure is in place:

Crawlable. AI engines crawl with their own user agents (GPTBot, ClaudeBot, PerplexityBot, Google-Extended). Make sure your robots.txt allows them if you want to be cited. Some content sites are now choosing the opposite — blocking AI crawlers. Pick a side deliberately.
Fast. AI crawlers fetch pages at scale and de-prioritize slow ones.
Schema. Article, FAQPage, Product, Organization, Service — same as classic SEO but more important because AI engines lean on structured data to confidently extract facts.
Stable URLs. AI engines build internal source graphs. URLs that change every six months don’t accumulate authority.

The content layer (where the real work is)

This is where GEO diverges from classic SEO. The content that gets cited has consistent properties:

Answers a specific question, completely. Not “the ultimate guide to X” — “What does Y cost in 2026, including the 3 hidden fees most providers don’t mention.”
Includes data, numbers, dates. AI engines love quantifiable facts they can repeat with confidence.
Uses extractable structures. Tables, numbered lists, FAQ blocks. Each block becomes a quotable chunk.
Names things specifically. “Tools to consider: Tool A, Tool B, Tool C” beats “various tools available.”
Is current. A post dated 2026-04 will be cited more than one dated 2022-04, even if both are accurate. Update old posts deliberately.

What we don’t know yet

GEO is a moving target. Things we’d flag as uncertain in 2026:

How much “brand mentions without links” actually counts. Some signals suggest AI engines weight mentions in trusted publications very heavily — more than links. But the data is messy.
Whether llms.txt matters yet. A proposed standard for letting sites declare what’s available for LLM indexing. Adoption is early; impact unclear.
How aggressively engines will start charging or rate-limiting citations. Perplexity has tested revenue-share models with publishers. The economic structure of AI search is still settling.

A practical GEO checklist

Things to do this quarter if you want to start showing up in AI answers:

Allow major AI crawlers in robots.txt (or block them — but decide).
Add Article, FAQPage, and Organization schema to top pages.
Update your top 10 most-trafficked posts with current dates, current data, and explicit numbered answers.
Write one or two “comparison” or “what does X cost” posts targeting questions in your category — these tend to over-index in AI citations.
Get cited by at least one trusted publication in your space. AI engines weight third-party mentions heavily.
Make sure your brand has clean Organization schema and a Wikipedia entry if you’re notable enough.
Test your visibility. Ask the same query in ChatGPT, Perplexity, and Google AI Overviews monthly. Track which sources get cited.

What’s not worth doing

Skip these even if you see them recommended elsewhere:

AI-only content farms. AI-generated content optimized purely for AI engines tends to underperform — both engines now downweight low-effort generated content.
Spammy mention campaigns. Trying to get your brand named in random forum threads or low-quality articles. The trust signal is reversed; you’ll get downweighted.
Stuffing pages with “as an AI language model” prompts. Visible prompt-injection in content gets penalized.

The honest take

Most of GEO is just well-executed SEO with one additional discipline: writing content that’s specifically structured to be quotable in a one-paragraph AI answer. The brands that win will be the ones already producing high-quality, current, well-structured content — and they’ll mostly win because their classic SEO is also strong.

If your SEO basics aren’t in place, GEO won’t save you. If your SEO is solid, GEO is the next layer worth investing in — but it doesn’t replace anything.