In this episode of ThriveCast, we speak with Crystal Carter, Head of AI Search and SEO Communications at Wix and author of the book The New Rules of AI Search, about why a website still matters when buyers are asking ChatGPT, Gemini, and Perplexity instead of typing into Google. Crystal explains why LLMs still treat your domain as the primary source on your own business, how legacy brands end up haunted by old campaigns, and why "indexed" and "cited" are two very different things. If you've been wondering whether your site is still worth the investment now that AI answers questions directly, this one's for you.
Key Insights
Your website is still the source LLMs trust most. Crystal’s take is straightforward: language models need language to pull from, and the cleanest, most controlled version of that language lives on a domain you own. Not a product feed. Not Reddit.
Legacy brands carry information debt. Older or fast-moving companies rack up campaigns, sunset projects, and one-off press hits that keep resurfacing in AI answers long after anyone stopped caring about them.
LLMs will merge everything on your domain into one picture of your brand. Crystal described a business that sold spices but got described by an LLM as also running custom denim and event planning, because a past community campaign lived on the same site and got folded into the AI’s read of the company.
The fix is to separate what doesn’t belong. Move old, unrelated projects to a different domain or subdomain, or mark the pages no-index, so they stop bleeding into how AI describes your core business.
Run your brand through multiple LLMs and compare notes. Ask the same basic questions (what does this brand do, who are its competitors) across different models. If the answers diverge, you’ve found a problem worth fixing.
New businesses have to win retrieval before they win training data. If an LLM doesn’t already “know” you, it falls back on RAG (retrieval augmented generation), and that only works if your site is discoverable in the first place.
“Fresh and spicy” content gets cited. Commodity explainers don’t. Timely news and genuine opinions force a model to go look something up. A basic “what is gelato” post usually doesn’t, because the model already knows the answer and has no reason to send anyone to your page for it.
Citations, mentions, and indexing are three separate things. Being indexed on Google says nothing about whether AI tools cite you. Crystal has seen brands with thousands of indexed pages get cited on only a small fraction of relevant queries.
Context decides who gets cited. Mixture-of-experts architectures (Google calls this “query fan-out”) route a question to a topic-specific cluster of knowledge, so a mention has to actually sit near your topic and geography to help. A random write-up on a tire blog won’t do much for an ice cream shop.
AI crawlers don’t behave like Google. In Crystal’s experience, OpenAI’s bots tend to crawl a site in full looking for anything citable, while Google works within a tighter crawl budget. Check your bot logs, and use robots.txt to steer AI crawlers away from thin or outdated pages.
LLMs hallucinate URLs, and you can correct them. Google sends a dead link to a 404. An LLM will just invent a URL that never existed. Crystal has used the thumbs up and thumbs down feedback in AI chat tools to correct wrong answers, since that feedback feeds directly into how the models get retrained.
A website is becoming a resource, not just a destination. Search bars, clear calls to action, and genuinely useful help docs let both humans and AI agents actually do something on your site, including through integrations like Wix’s built-in MCP, which lets Crystal manage her own sites straight from an AI assistant.
Actionable Takeaways
Audit your domain for old campaigns or sunset projects that AI tools might still be pinning on your current business.
Move off-brand content to a separate domain or subdomain, or set it to no-index, if it no longer represents what you do.
Ask the same basic questions about your brand across several LLMs and compare the answers for inconsistencies.
Publish timely, opinionated content, not just evergreen explainers that LLMs already know how to answer without you.
Check your bot logs to see which AI crawlers show up and how their behavior differs from Google’s.
Use robots.txt to keep AI crawlers away from thin or outdated pages that could muddy how they describe you.
Watch for hallucinated URLs in your AI referral traffic, set up redirects, and use thumbs up and thumbs down feedback to correct wrong information.
Keep your brand description consistent everywhere you show up, PR, blog posts, social, so the phrasing correlates back to you.
Build agentic-friendly features: a search bar, clear CTAs, and real help documentation that both people and AI agents can act on.
Spread your budget across channels instead of betting it all on one, since LLMs pull citations from PR, social, owned content, and paid alike.
Resources Mentioned
The New Rules of AI Search — Crystal Carter’s book on SEO for visibility in ChatGPT, Gemini, and generative search.
Wix AI Search Lab — Wix’s publication on AI search data, insights, and strategy.
NL Web — the open protocol from RV Guha’s team (creators of RSS and Schema.org) that lets AI agents query a website in natural language. Wix is a partner.
Agentic Commerce Protocol — the open standard for AI-agent-driven purchases. Wix is an official signatory.
GPT for Sheets and Docs — the tool Crystal uses to run brand-consistency questions across LLMs at scale.
Wix AI Website Builder — Wix's AI Website Builder generates business-ready websites, pairing that speed with Wix's built-in SEO and GEO tools so the sites it creates are structured to be found by both search engines and AI assistants from day one.
If you're a founder or marketer wondering whether a website still earns its keep in an AI-first world, Crystal's answer is simple: yes, but its job has changed. It's no longer just a place humans land on. It's the primary, controllable source AI systems check before they describe, cite, or transact with your business. Get the basics right, consistent messaging, a clean site structure, and content built for people and agents alike, and you stay the authority on your own brand.
🎧 Loved the episode?
Subscribe to ThriveCast for more behind-the-scenes stories from the builders shaping the future of SaaS.
