TL;DR: For a decade, agencies could ignore Bing. AI assistants ended that: ChatGPT’s web search has drawn on Microsoft Bing’s index as part of the OpenAI–Microsoft partnership, and other assistants leaning on web retrieval each bring their own source mix. When an assistant searches the web before answering, its index is the gatekeeper — and for a large slice of AI answers, that gatekeeper speaks Bing.
The quiet return of the second search engine
Nobody rebuilt their SEO practice around Bing’s roughly single-digit share of classic search — and that was rational. But AI assistants changed what an index is for. When ChatGPT needs current information, it issues searches and reads results before composing its answer — and testing shows it queries Bing first, only falling back to Google’s results when Bing’s are thin or missing (Search Engine Journal). Microsoft’s partnership with OpenAI put Bing in that retrieval path, which means a page that Bing hasn’t indexed — or ranks poorly — has a materially harder time being read, and therefore cited, in those answers.
The general pattern extends beyond one assistant: every answer engine with web access has a retrieval layer, and the retrieval layer’s view of the web decides what the model even considers. Optimizing content the model never retrieves is wasted effort.
What this means in practice
1. Track Bing explicitly. If AI citations are a KPI, Bing rank is an input to it. Preferium tracks Google and Bing per keyword for exactly this reason — divergences between the two are early signals about which content the retrieval layer prefers.
2. Feed the indexes proactively. Waiting days for a crawler to notice changes is unnecessary. The IndexNow protocol — supported by Bing among others — lets sites push URL changes to indexes immediately. Preferium pings IndexNow (plus Bing and Google sitemaps) automatically after every deploy, cutting the fix-to-indexed gap from days toward minutes.
3. Keep the raw HTML honest. Retrieval layers are crawlers, with crawler constraints. The same server-side principle that governs AI crawlers directly (see our explainer) governs the indexes they search: substance that only exists after JavaScript runs is at a structural disadvantage everywhere at once.
Key takeaways
- AI assistants with web access answer from a retrieval layer — and Bing sits in ChatGPT’s.
- Bing rank is no longer vanity: it’s an input to AI citation probability. Track it per keyword.
- Push changes to indexes with IndexNow instead of waiting to be crawled.
- One clean, server-side-optimized HTML response serves every crawler and every index simultaneously.
