The two layers: what's baked in vs. what's fetched live

Training data is the web snapshot a model learned from — months old by the time you're talking to it. If your name is well-represented there (consistent profiles, published content, mentions), the model "knows" you; if not, you don't exist to it until retrieval saves you. You can't edit training data retroactively — you can only be present enough that the next training run picks you up.

Live retrieval is what happens when the assistant searches the web mid-conversation — which local and "who should I hire" questions almost always trigger, because they're current. This is the layer you can influence this quarter.

Which index feeds which assistant

  • ChatGPT web search draws on Bing's index — which makes Bing Webmaster Tools (free, imports from Search Console in two clicks) quietly one of the highest-leverage registrations in agent marketing.
  • Google AI Overviews / AI Mode draw on Google's index and ranking systems — your classic SEO work feeds them directly.
  • Perplexity maintains its own crawl with an emphasis on citing sources.
  • Gemini draws on Google's index and infrastructure.

Practical consequence: "AI SEO" isn't a separate universe — it's being findable and credible in the two indexes that matter, plus structure that makes you quotable.

What actually gets cited

Across the systems, the pattern of citable sources is consistent: pages that answer a question directly (a clear question-shaped heading with a concise answer up top), entity consistency (your name, brokerage, and markets identical across your site, GBP, and profiles — contradictions read as noise), structured data (schema that makes facts machine-readable — see our schema walkthrough), reviews as third-party evidence, and independent mentions — your own site claims, other sites confirm.

The commercial stakes, from our verified statistics roundup: when an AI Overview appears, brands cited inside it earn about 120% more clicks per impression than brands that aren't — while uncited brands lost 67% of their click-through over 2025 (Seer Interactive).

The realtor checklist

  • Verify your site in Google Search Console AND Bing Webmaster Tools; submit your sitemap to both.
  • Add RealEstateAgent and FAQ schema; keep it valid (test it — even one missing brace makes it unreadable).
  • Publish direct answers to your market's questions, each opening with a liftable summary.
  • Align every profile — GBP, site, socials, directories — on identical name/brokerage/market facts.
  • Build review velocity and pursue genuine third-party mentions (local press, community sites).
  • Then audit: ask each assistant your market's questions monthly and track who gets named — that measurement loop is the core of our AI citations service.