SEO for ChatGPT: The Engine-Specific Optimization Checklist

Generic AI-optimization advice fails on ChatGPT because ChatGPT is not Google: it crawls with its own agents, answers conversationally as often as it searches, and blends retrieval with training knowledge. The general discipline is covered in the AI SEO playbook and how to rank in ChatGPT; this page is the engine-specific checklist — the things that are true for ChatGPT and not necessarily for AI Overviews or Perplexity.

How ChatGPT's pipeline differs

• Its own crawler. Access is governed by OpenAI's user-agents in your robots.txt, independent of Googlebot. A site open to Google can still be closed to ChatGPT by an over-broad AI-agent block.
• Conversational defaults. Many category questions get answered from training knowledge with no live retrieval at all — which means being described correctly in the corpus (consistent facts across many crawlable pages) matters as much as any single page.
• Search mode cites. When ChatGPT does browse, the answer shows clickable sources; those citation lists behave like our AI Overviews study found: comparison pages, roundups and threads dominate.

The checklist

Access and legibility

• robots.txt allows the OpenAI crawler agents on every page you want quoted; no blanket AI-block "just in case".
• /llms.txt live with the canonical description and 5–15 absolute links (templates).
• Question pages answer in the first two sentences, with one table or list; headings phrased as the questions buyers type.

Facts

• One canonical sentence (what/for-whom/what-it-does/price-model) reused verbatim on homepage, pricing, listings and profiles.
• Pricing and limits stated as numbers on a page, not only in images or JS-rendered widgets — retrieval reads HTML.
• Category words you want to own appear literally ("AI video dubbing tool", not just a brand tagline); ChatGPT's conversational recall matches on plain category language.

Mentions

• Present in at least two third-party comparison or roundup pages in your category (the citation pool); pitch the gatekeeper listicles with data.
• A permanent directory listing on indexed domains with dofollow links (submit on aat.ee; networks and DR on /pricing).
• Accurate community presence: the threads where your category is compared, with useful answers rather than links (threads get cited).

Freshness and defense

• Changelog or "last updated" dates that reflect real updates; no stale year stamps on pages you want quoted (we normalize ours at render time for exactly this reason).
• Monthly logged-out prompt audit: your ten category questions, conversational and search mode, logged with the names and sources ChatGPT returns; diff against last month.
• Wrong-description incident plan: when an answer misstates your pricing or category, fix the sources (listings, pages) — not the answer, which you cannot edit.

What is ChatGPT-only vs shared

Shared with every engine: fact consistency, structure, crawl access, third-party mentions. ChatGPT-only emphasis: conversational recall (corpus-level consistency beats any single page) and its independent crawler policy. If you can only do three things this quarter: unblock the crawler, normalize the canonical sentence everywhere, and get into two gatekeeper roundups.

FAQ

Does ChatGPT use Bing's index? Historically partly; today its search mode runs its own retrieval stack. Treat "which index" as unknowable and optimize the invariants: crawlable, consistent, cited.

Should I optimize for conversational or search mode? Both, with the same assets: conversational rewards corpus consistency, search rewards page structure. The checklist above covers each mode in one line.

How do I see whether I'm cited? Ask in search mode and read the source chips; for conversational answers, only the named products are visible — log names weekly. The visibility tools guide covers scaling this.

The fastest route into ChatGPT's corroboration layer is a permanent factual listing: submit your product on aat.ee — canonical description, dofollow, indexed, permanent.