This article is for hospital marketing heads and the SEO or analytics leads who report to them. You have probably been asked a version of the same question in a board or leadership meeting: "Are we showing up in AI answers, and is it doing anything for us?" Until recently the honest answer was a screenshot of a few prompts. That has changed, because both Google and Microsoft now report some AI data inside tools you already have.
The data is partial, and each report measures something different. One counts impressions, the other counts citations, and neither tells you whether a patient booked. This post explains what each report contains, where it stops, and how to build a small citation panel that fills the gap. It assumes you already understand the basics covered in our AI search optimisation guide.
The aim is a single page-level view per service line that your team can refresh monthly and defend in a meeting.
Why a hospital should measure this at all
Google reported in its Q2 2026 results that AI Mode has passed one billion monthly users, and that AI Overviews and AI Mode have been merged into one experience. Whatever your view of the hype, the surface your patients search on has changed.
The effect on clicks depends on whether you are cited. A Seer Interactive study of queries where an AI Overview appeared found organic click-through fell from 1.76% to 0.61%, while brands cited inside the overview earned 35% more organic clicks and 91% more paid clicks than brands that were not. It is a vendor study with its own sample, so treat the direction as useful and the exact figures as indicative. The practical point is that being cited and being absent are different commercial positions, and you cannot manage what you cannot see.
What Search Console shows, and what it does not
Google's documentation on AI features and your website says that traffic from AI Overviews and AI Mode is included in the Performance report under the Web search type. There is no separate AI Mode filter in the standard report. That means your existing clicks and impressions already contain AI-surface activity, mixed in with everything else.
The newer Generative AI performance report isolates part of that. As of Google's help page, it covers AI Overviews and AI Mode, reports one metric (impressions, defined as links to your site shown in a generative AI feature), and is available to all sites worldwide from 31 August 2026. You can group by page, country, date or device, and filter by text or multimodal queries.
Four limits matter when you read it:
- Impressions only. There are no clicks, no click-through rate and no position in this report. You cannot calculate an AI click-through rate from it.
- One impression per response. Several links from your site in one AI response count once on the chart. Do not read the count as "number of links shown".
- Canonical URLs. Page data is assigned to the canonical URL, so a hospital with parameterised doctor or location URLs should check that canonicals are clean before comparing pages.
- Thresholds and row limits. The report may not appear for a site with too few generative AI impressions, and the usual 1,000-row table limit applies. A multi-specialty hospital with thousands of pages should export by page group, not expect one complete table.
A small hospital might see no report at all. That is not a fault; it can simply mean there is too little data.
Reading it by service line
Export the page dimension and tag each URL with a service line (cardiology, oncology, maternity, health checks) and a page type (service hub, condition, treatment, doctor profile, location). In a spreadsheet this is a lookup on the URL path. Then answer three questions per service line:
- Which page types receive AI impressions? If only blog posts do and no service pages, your commercial pages may not be getting retrieved.
- Are the impressions on pages you would want a patient to land on? An old press release collecting AI impressions is not a win.
- Do AI impressions grow while Web clicks to the same pages stay flat? That pattern means you are being shown but not clicked, which is expected in a zero-click surface and is a reason to write for the answer, not just the visit.
What Bing's AI Performance report shows
Microsoft introduced AI Performance in Bing Webmaster Tools as a public preview in February 2026. According to Microsoft's announcement, it reports:
- Total citations: how often your content is displayed as a source in AI answers in the period.
- Average cited pages: the daily average of unique URLs from your site cited.
- Page-level citation activity: citation counts per URL.
- Grounding queries: the phrases the AI used to retrieve your content, described by Microsoft as a sample, not a complete list.
- A timeline of citation activity across supported AI experiences.
Coverage is Microsoft Copilot, AI summaries in Bing and select partner integrations. Microsoft says the data does not show placement, ranking, authority or the role a page played in the answer. Several third-party write-ups describe later additions such as citation share and topic groupings, but I could not confirm those from Microsoft's own pages, so check what your Bing Webmaster Tools account actually shows before promising your team a particular chart.
Why would a hospital in India care about Bing? Because the grounding queries are the only first-party view of how an AI system rewrites a patient's question into retrieval phrases. Those phrases are often longer and more specific than the keywords in your SEO plan. A cardiology page cited for "what tests are done before bypass surgery" tells you something that a keyword tool will not. Register and verify your site in Bing Webmaster Tools, and keep Bing Places details consistent with your Google profile.
