How ChatGPT Search works — and what it changes for your AI visibility strategy
ChatGPT with live search changed the rules: it now reaches the web in real time. Here's how ChatGPT Search decides which sources to use — and what that changes for companies that want to show up in the answers.
For a long time, ChatGPT was strictly a trained-knowledge model — its answers came from what it had learned during training, with a knowledge cutoff that kept it from speaking about recent events. The version with live search changed that: ChatGPT can now search the web in real time, similar to what Perplexity has done from the start.
That shift has direct implications for AI visibility strategy. ChatGPT with live search behaves differently from pure-knowledge ChatGPT — and understanding that difference is essential to calibrate the right strategy for each mode.
How ChatGPT Search selects sources
ChatGPT Search uses Microsoft Bing as its underlying search engine (thanks to the OpenAI–Microsoft partnership). That carries an important technical implication: the index behind ChatGPT Search is Bing's — not Google's.
To show up in ChatGPT Search answers, your site needs to be indexed and ranking well on Bing — not just Google. Plenty of companies pour 100% of their SEO effort into Google and ignore Bing, but for ChatGPT Search visibility, Bing is the critical channel.
The recommended setup is to verify your site in Bing Webmaster Tools, make sure your sitemap is submitted and indexed on Bing, and check that your key pages show up when searched directly on Bing.
What ChatGPT Search looks for in a source
Based on observed citation patterns, ChatGPT Search favors:
- Domain authority on Bing — domains with an indexing history, inbound links, and content relevant to the topic
- Freshness — recent content has an edge for queries on topics that move (news, prices, trends)
- Semantic match — a match between the query and the page's content that goes beyond the exact keyword
- Clear answer structure — pages with clean H1/H2s, concise paragraphs, and the answer up front (BLUF)
What changes in your strategy once ChatGPT can search
For evergreen content (guides, deep-dive articles, definitions): the strategy is the same as Bing SEO — authority, structure, answer density. Content that already performs well on Google probably performs well in ChatGPT Search.
For timely content (recent data, industry trends, company news): ChatGPT Search changes the game. Posts about recent data from your company or industry — surveys, reports, announcements — can surface in ChatGPT Search far faster than in the static training model.
For local presence: because ChatGPT Search uses Bing, and Bing integrates with Bing Places (the equivalent of Google Business Profile), companies with a well-optimized local presence on Bing Places show up in ChatGPT Search for local queries.
When it searches vs. when it answers from memory
One confusing part for marketing managers is that ChatGPT sometimes searches and sometimes answers from memory — depending on the query, the user, and the settings. For queries clearly about recent events, ChatGPT Search kicks in. For general concepts, it may answer from memory.
The robust play is to optimize for both modes at once: build knowledge presence (content that made it into training) AND optimize for live search (indexed on Bing, current content, answer structure).
How Crowly can help you track your presence in ChatGPT Search
Crowly monitors your presence in ChatGPT (including the live-search version when relevant), so you can see how your score moves after optimization work aimed at Bing and at timely content.
Measure your presence in ChatGPT, with and without live search. Free diagnostic. Analyze my brand →
Sources:
