Is Copilot for Health Crawling Your Website?
A research paper dropped on arXiv this week that should get the attention of anyone publishing health content online. Microsoft Copilot is already handling health conversations at serious scale — and we don't fully know how it's sourcing what it tells people.
Published 20 April 2026.
What We Know About Copilot for Health
"Copilot for Health" isn't a separate product, at least not officially. Based on the arXiv paper published this week, it refers to health-related use of Microsoft Copilot — the conversational AI already baked into Windows, Microsoft 365, and Bing. Researchers analysed over 500,000 de-identified health conversations from January 2026 to understand what people are actually asking it.
That's half a million conversations. In one month.
The study developed a "hierarchical intent taxonomy of 12 primary categories using privacy-preserving LLM-based classification validated against expert human annotation," according to the paper. In plain terms: they built a structured map of why people turn to Copilot for health questions, then verified it with human experts. That's a serious methodology, not a back-of-napkin analysis.
So what are people asking? The paper applies LLM-driven topic clustering to surface the most common themes within each intent category. The full breakdown isn't in the abstract, but the scope — 12 distinct intent categories across 500k conversations — suggests this covers everything from symptom checking to medication queries to mental health support.
Does It Crawl the Web? What User Agent Does It Use?
Honestly, this is where things get thin. The arXiv paper is a research study characterising user behaviour, not technical documentation about Copilot's crawling infrastructure. We couldn't confirm from this source whether a dedicated health-focused crawler exists, what user agent string it uses, or whether it operates separately from Bingbot.
Microsoft Copilot is known to draw on Bing's index in general use. Whether health queries trigger different retrieval behaviour — say, preferencing authoritative medical sources over standard web results — we can't say from this paper alone. No official documentation exists yet that addresses this specifically.
Does It Support LLMs.txt?
No information available yet. The paper makes no reference to LLMs.txt or any structured content discovery protocol. Given that LLMs.txt adoption is still patchy across the industry, this isn't surprising — but it's a gap worth watching.
Is There a Submission or Indexing Process?
There is no submission process confirmed in the source material. As of 20 April 2026, no official documentation exists describing a dedicated indexing pathway for Copilot's health-related responses. If Copilot is pulling from Bing's standard index, then standard Bing Webmaster Tools submission is your most logical starting point — but that's an inference, not a confirmed fact.
What Type of Content Does Copilot for Health Appear to Favour?
The paper doesn't detail citation behaviour or content preferences directly. What it does tell us is that the conversation analysis was scoped to "health-related conversations," which signals that content touching symptoms, conditions, treatments, medications, and mental health is actively being surfaced in responses. Whether Copilot privileges peer-reviewed sources, NHS or CDC-style institutional pages, or general health publishers is not confirmed here.
Fair point to raise: a taxonomy validated against "expert human annotation" suggests the researchers — and presumably Microsoft — care about accuracy and intent alignment. That's at least a soft signal that authoritative, clearly structured health content has an edge.
What Should Website Owners Do Right Now?
A few practical moves worth making this week:
1. Get your Bing presence in order. If Copilot indexes via Bing, your Bing Webmaster Tools setup matters more than most people treat it. Submit your sitemap if you haven't already.
2. Structure your health content for intent, not just keywords. The 12-category intent taxonomy in this research maps user goals, not search terms. Content that clearly addresses a specific health intent — symptoms, treatment options, medication questions — is better positioned than generic SEO filler.
3. Add schema markup. MedicalCondition, Drug, and FAQPage schema help any AI system understand what your content is actually about. This is low-effort, high-signal work.
4. Track whether you're being cited. This is the part most publishers skip. If Copilot starts surfacing health answers at scale, knowing whether your content appears in those responses is genuinely useful intelligence. Uptrue's AI Visibility tracking monitors exactly this — which AI systems are citing your content and how that changes over time. Worth setting up before you need it, not after.
5. Consider an LLMs.txt file. No confirmation it helps with Copilot today, but adoption cost is low and the direction of travel is clear. Uptrue's tools section has resources to help you get started.
FAQ
Is Copilot for Health a separate Microsoft product? Based on available research as of April 2026, "Copilot for Health" appears to describe health-related use of Microsoft Copilot rather than a distinct standalone product — no official separate product page has been confirmed.
How many health conversations does Microsoft Copilot handle? According to a peer-reviewed arXiv paper (arXiv:2604.15331), researchers analysed over 500,000 de-identified health conversations with Microsoft Copilot from January 2026 alone.
What user agent does Copilot for Health use to crawl websites? We couldn't confirm a specific user agent string for Copilot's health-related crawling from available sources as of 20 April 2026.
Should I submit my health website to Copilot for indexing? No confirmed submission process exists for Copilot health content as of April 2026. Ensuring your site is indexed by Bing via Bing Webmaster Tools is the most defensible step available right now.
Does Copilot for Health support LLMs.txt? No information is available yet confirming LLMs.txt support for Microsoft Copilot's health responses.